Data verification method and device, electronic equipment and automatic driving vehicle

Abstract

The invention provides a data verification method and device, electronic equipment and an automatic driving vehicle, and relates to the technical field of automatic driving, in particular to the application fields of obstacle recognition, obstacle tracking, sensor data association, sensor data fusion and the like. According to the specific implementation scheme, the method comprises the steps of obtaining current frame data collected by a first target sensor in a plurality of sensors; wherein the plurality of sensors are mounted on a target object; taking each of the plurality of sensors except the first target sensor as a second target sensor to obtain a plurality of reference data for representing similarity between the current frame data and historical frame data collected by the second target sensor; wherein the historical frame data comes from global obstacle tracking information of the target object; fusing the plurality of reference data to obtain a verification result for the current frame data; wherein the verification result is used for representing whether the current frame data can be associated with the historical frame data.

Description

Technical Field

[0001] The present disclosure relates to the field of autonomous driving technology, and in particular to application fields such as obstacle recognition, obstacle tracking, sensor data association, and sensor data fusion, and specifically to a data verification method, device, electronic equipment, and an autonomous driving vehicle. Background Art

[0002] In the field of autonomous driving technology, sensor data association is a crucial step that aims to improve the autonomous driving system's ability to identify and track obstacles to ensure that the autonomous driving system can make relatively accurate driving decisions. Summary of the invention

[0003] The present invention provides a data verification method, device, electronic equipment and an autonomous driving vehicle.

[0004] According to a first aspect of the present disclosure, there is provided a data verification method, comprising:

[0005] Acquire current frame data collected by a first target sensor among multiple sensors; wherein the multiple sensors are installed on the target object;

[0006] Each sensor except the first target sensor among the multiple sensors is used as a second target sensor to obtain multiple reference data for representing the similarity between the current frame data and the historical frame data collected by the second target sensor; wherein the historical frame data is derived from the global obstacle tracking information of the target object;

[0007] Multiple reference data are fused to obtain a verification result for the current frame data; wherein the verification result is used to characterize whether the current frame data can be associated with the historical frame data.

[0008] According to a second aspect of the present disclosure, there is provided a data verification device, comprising:

[0009] A first data acquisition unit, used to acquire current frame data collected by a first target sensor among the multiple sensors; wherein the multiple sensors are installed on the target object;

[0010] a second data acquisition unit, configured to use each sensor of the plurality of sensors except the first target sensor as a second target sensor, to acquire a plurality of reference data for characterizing the similarity between the current frame data and the historical frame data acquired by the second target sensor; wherein the historical frame data is derived from the global obstacle tracking information of the target object;

[0011] The data fusion unit is used to fuse multiple reference data to obtain a verification result for the current frame data; wherein the verification result is used to indicate whether the current frame data can be associated with the historical frame data.

[0012] According to a third aspect of the present disclosure, there is provided an electronic device, including:

[0013] at least one processor;

[0014] a memory communicatively coupled to the at least one processor;

[0015] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method provided in the first aspect of the present disclosure.

[0016] According to a fourth aspect of the present disclosure, an autonomous driving vehicle is provided, comprising the electronic device provided by the third aspect of the present disclosure.

[0017] According to a fifth aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause the computer to execute the method provided according to the first aspect of the present disclosure.

[0018] According to a sixth aspect of the present disclosure, a computer program product is provided, comprising a computer program, which implements the method provided according to the first aspect of the present disclosure when executed by a processor.

[0019] The present disclosure can improve the accuracy of sensor data association.

[0020] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.

[0022] Figure 1 A flowchart of a data verification method provided by an embodiment of the present disclosure;

[0023] Figure 2 A schematic diagram of a probability distribution of a width difference between a first bounding box and a second bounding box provided in an embodiment of the present disclosure;

[0024] Figure 3 A schematic diagram of a probability distribution of a height difference between a first bounding box and a second bounding box provided in an embodiment of the present disclosure;

[0025] Figure 4 A schematic diagram of probability distribution of reference data used to characterize the similarity between current frame data and historical frame data under a shape evaluation dimension provided by an embodiment of the present disclosure;

[0026] Figure 5 A schematic diagram of a probability partial situation of a first overflow amount of a plurality of feature points relative to a third bounding box in a first direction provided by an embodiment of the present disclosure;

[0027] Figure 6 A schematic diagram of another partial probability situation of a first overflow amount of a plurality of feature points relative to a third bounding box in a first direction provided by an embodiment of the present disclosure;

[0028] Figure 7 A schematic diagram of a partial probability situation of a second overflow amount of a plurality of feature points relative to a third bounding box in a second direction provided by an embodiment of the present disclosure;

[0029] Figure 8 A schematic diagram of another partial probability situation of a second overflow amount of a plurality of feature points relative to a third bounding box in a second direction provided by an embodiment of the present disclosure;

[0030] Fig. 9 A schematic diagram of probability distribution of reference data used to characterize the similarity between current frame data and historical frame data under a location evaluation dimension provided by an embodiment of the present disclosure;

[0031] Fig.10 A schematic diagram of a probability distribution of a first initial distance between a fifth obstacle and a sixth obstacle in a two-dimensional plane provided in an embodiment of the present disclosure;

[0032] Fig.11 A schematic diagram of a probability distribution of a second initial distance between a fifth obstacle and a sixth obstacle in a three-dimensional space provided in an embodiment of the present disclosure;

[0033] Fig.12 A schematic diagram of a probability distribution of a third initial distance between a seventh obstacle and an eighth obstacle in a three-dimensional space provided in an embodiment of the present disclosure;

[0034] Fig.13 A schematic diagram of another probability distribution of a third initial distance between a seventh obstacle and an eighth obstacle in a three-dimensional space provided in an embodiment of the present disclosure;

[0035] Fig.14 A schematic diagram of probability distribution of reference data used to characterize the similarity between current frame data and historical frame data under a second three-dimensional distance evaluation dimension provided by an embodiment of the present disclosure;

[0036] Fig.15 A schematic diagram of a probability distribution of an initial speed difference between a first driving speed and a second driving speed provided in an embodiment of the present disclosure;

[0037] Fig.16A schematic diagram of probability distribution of reference data used to characterize the similarity between current frame data and historical frame data under a speed evaluation dimension provided by an embodiment of the present disclosure;

[0038] Fig.17 A schematic diagram of a probability distribution of a fourth initial distance provided in an embodiment of the present disclosure;

[0039] Fig.18 A schematic diagram of probability distribution of reference data used to characterize the similarity between current frame data and historical frame data under a first projection position evaluation dimension provided by an embodiment of the present disclosure;

[0040] Fig.19 A schematic diagram of a probability distribution of a fifth initial distance provided in an embodiment of the present disclosure;

[0041] Fig. 20 A schematic diagram of probability distribution of reference data used to characterize the similarity between current frame data and historical frame data under a second projection position evaluation dimension provided in an embodiment of the present disclosure;

[0042] Fig.21 A schematic diagram of a scenario of a data verification method provided by an embodiment of the present disclosure;

[0043] Fig. 22 A schematic structural block diagram of a data verification device provided in an embodiment of the present disclosure;

[0044] Fig.23 A schematic structural block diagram of an electronic device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0045] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope of the present disclosure. Similarly, for the sake of clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0046] As described in the background technology, in the field of autonomous driving technology, sensor data association is a crucial step, which aims to improve the ability of the autonomous driving system to identify and track obstacles to ensure that the autonomous driving system can make relatively accurate driving decisions.

[0047] However, currently, sensor data association is usually achieved by manually designing (Hand-Craft) features, similar to the following expression of "data collected by sensor A - data collected by sensor B".

[0048] Among them, sensor B can be used to characterize each sensor installed on the autonomous vehicle except sensor A. For example, if sensor A is a lidar, sensor B can be used to characterize a camera and a radar; for another example, if sensor A is a camera, sensor B can be used to characterize lidar and radar; for another example, if sensor A is a radar, sensor B can be used to characterize lidar and a camera.

[0049] In addition, the data collected by sensor A can specifically be the current frame data collected by sensor A, that is, the sensor data most recently collected by sensor A, which is used to characterize the obstacle state most recently perceived by sensor A; the data collected by sensor B can specifically be the historical frame data collected by sensor B, that is, the sensor data historically collected by sensor B, which is used to characterize the obstacle state historically perceived by sensor B. Among them, the data collected by sensor B comes from the global obstacle tracking information (Global_Track) of the autonomous driving vehicle. Since the global obstacle tracking information of the autonomous driving vehicle stores the sensor data that have been associated, it can be determined that the data collected by sensor B is also the sensor data that has been associated.

[0050] Furthermore, when sensor A is a laser radar, it can be determined whether the data collected by sensor A (that is, the data collected by the laser radar) can be associated with the data collected by sensor B (including the data collected by the camera and the data collected by the radar) based on the following method:

[0051] (1) Data collected by LiDAR - Data collected by camera

[0052] When the shape similarity between the obstacle represented by the data collected by the lidar and the obstacle represented by the data collected by the camera is greater than a first preset similarity threshold, the first evaluation result is determined to be true; when the shape similarity between the obstacle represented by the data collected by the lidar and the obstacle represented by the data collected by the camera is less than or equal to the first preset similarity threshold, the first evaluation result is determined to be false.

[0053] When the data collected by the lidar (three-dimensional point cloud data) completely falls within the bounding box of the obstacle represented by the data collected by the camera, the second evaluation result is determined to be true; when the data collected by the lidar does not completely fall within the bounding box of the obstacle represented by the data collected by the camera, the second evaluation result is determined to be false.

[0054] (2) Data collected by LiDAR - Data collected by radar

[0055] When the plane distance between the obstacle represented by the data collected by the lidar and the obstacle represented by the data collected by the radar is less than the first preset distance threshold, the third evaluation result is determined to be true; when the plane distance between the obstacle represented by the data collected by the lidar and the obstacle represented by the data collected by the radar is greater than or equal to the first preset distance threshold, the third evaluation result is determined to be false.

[0056] When the speed difference between the obstacle represented by the data collected by the lidar and the obstacle represented by the data collected by the radar is less than the first preset speed difference threshold, the fourth evaluation result is determined to be true; when the speed difference between the obstacle represented by the data collected by the lidar and the obstacle represented by the data collected by the radar is greater than or equal to the first preset speed difference threshold, the fourth evaluation result is determined to be false.

[0057] When the first evaluation result, the second evaluation result, the third evaluation result and the fourth evaluation result are all true at the same time, it is determined that the data collected by sensor A can be associated with the data collected by sensor B; when at least one of the first evaluation result, the second evaluation result, the third evaluation result and the fourth evaluation result is false, it is determined that the data collected by sensor A cannot be associated with the data collected by sensor B.

[0058] When sensor A is a camera, it can be determined whether the data collected by sensor A (that is, the data collected by the camera) can be associated with the data collected by sensor B (including the data collected by the laser radar and the data collected by the radar) based on the following method:

[0059] (1) Data collected by the camera - Data collected by the lidar

[0060] When the shape similarity between the obstacle represented by the data collected by the lidar and the obstacle represented by the data collected by the camera is greater than a first preset similarity threshold, the first evaluation result is determined to be true; when the shape similarity between the obstacle represented by the data collected by the lidar and the obstacle represented by the data collected by the camera is less than or equal to the first preset similarity threshold, the first evaluation result is determined to be false.

[0061] When the data collected by the lidar falls completely within the bounding box of the obstacle represented by the data collected by the camera, the second evaluation result is determined to be true; when the data collected by the lidar does not fall completely within the bounding box of the obstacle represented by the data collected by the camera, the second evaluation result is determined to be false.

[0062] (2) Data collected by camera - Data collected by radar

[0063] When the three-dimensional distance between the obstacle represented by the data collected by the camera and the obstacle represented by the data collected by the radar is less than the second preset distance threshold, the fifth evaluation result is determined to be true; when the three-dimensional distance between the obstacle represented by the data collected by the camera and the obstacle represented by the data collected by the radar is greater than or equal to the second preset distance threshold, the fifth evaluation result is determined to be false.

[0064] When the speed difference between the obstacle represented by the data collected by the camera and the obstacle represented by the data collected by the radar is less than the second preset speed difference threshold, the sixth evaluation result is determined to be true; when the speed difference between the obstacle represented by the data collected by the camera and the obstacle represented by the data collected by the radar is greater than or equal to the second preset speed difference threshold, the sixth evaluation result is determined to be false.

[0065] When the distance between the obstacle represented by the data collected by the camera and the obstacle represented by the data collected by the radar in the first direction is less than the third preset distance threshold, the seventh evaluation result is determined to be true; when the distance between the obstacle represented by the data collected by the camera and the obstacle represented by the data collected by the radar in the first direction is greater than or equal to the third distance threshold, the seventh evaluation result is determined to be false.

[0066] When the distance between the obstacle represented by the data collected by the camera and the obstacle represented by the data collected by the radar in the second direction is less than the fourth distance threshold, the eighth evaluation result is determined to be true; when the distance between the obstacle represented by the data collected by the camera and the obstacle represented by the data collected by the radar in the second direction is greater than or equal to the third distance threshold, the eighth evaluation result is determined to be false.

[0067] When the first evaluation result, the second evaluation result, the fifth evaluation result, the sixth evaluation result, the seventh evaluation result and the eighth evaluation result are all true at the same time, it is determined that the data collected by sensor A can be associated with the data collected by sensor B; when at least one of the first evaluation result, the second evaluation result, the fifth evaluation result, the sixth evaluation result, the seventh evaluation result and the eighth evaluation result is false, it is determined that the data collected by sensor A cannot be associated with the data collected by sensor B.

[0068] When sensor A is a radar, it can be determined whether the data collected by sensor A (that is, the data collected by the radar) can be associated with the data collected by sensor B (including the data collected by the laser radar and the camera) based on the following method:

[0069] (1) Data collected by radar - Data collected by lidar

[0070] When the plane distance between the obstacle represented by the data collected by the lidar and the obstacle represented by the data collected by the radar is less than the first preset distance threshold, the third evaluation result is determined to be true; when the plane distance between the obstacle represented by the data collected by the lidar and the obstacle represented by the data collected by the radar is greater than or equal to the first preset distance threshold, the third evaluation result is determined to be false.

[0071] When the speed difference between the obstacle represented by the data collected by the lidar and the obstacle represented by the data collected by the radar is less than the first preset speed difference threshold, the fourth evaluation result is determined to be true; when the speed difference between the obstacle represented by the data collected by the lidar and the obstacle represented by the data collected by the radar is greater than or equal to the first preset speed difference threshold, the fourth evaluation result is determined to be false.

[0072] (2) Data collected by radar - Data collected by camera

[0073] When the three-dimensional distance between the obstacle represented by the data collected by the camera and the obstacle represented by the data collected by the radar is less than the second preset distance threshold, the fifth evaluation result is determined to be true; when the three-dimensional distance between the obstacle represented by the data collected by the camera and the obstacle represented by the data collected by the radar is greater than or equal to the second preset distance threshold, the fifth evaluation result is determined to be false.

[0074] When the speed difference between the obstacle represented by the data collected by the camera and the obstacle represented by the data collected by the radar is less than the second preset speed difference threshold, the sixth evaluation result is determined to be true; when the speed difference between the obstacle represented by the data collected by the camera and the obstacle represented by the data collected by the radar is greater than or equal to the second preset speed difference threshold, the sixth evaluation result is determined to be false.

[0075] When the distance between the obstacle represented by the data collected by the camera and the obstacle represented by the data collected by the radar in the first direction is less than the third preset distance threshold, the seventh evaluation result is determined to be true; when the distance between the obstacle represented by the data collected by the camera and the obstacle represented by the data collected by the radar in the first direction is greater than or equal to the third distance threshold, the seventh evaluation result is determined to be false.

[0076] When the distance between the obstacle represented by the data collected by the camera and the obstacle represented by the data collected by the radar in the second direction is less than the fourth distance threshold, the eighth evaluation result is determined to be true; when the distance between the obstacle represented by the data collected by the camera and the obstacle represented by the data collected by the radar in the second direction is greater than or equal to the third distance threshold, the eighth evaluation result is determined to be false.

[0077] When the third evaluation result, the fourth evaluation result, the fifth evaluation result, the sixth evaluation result, the seventh evaluation result and the eighth evaluation result are all true at the same time, it is determined that the data collected by sensor A can be associated with the data collected by sensor B; when at least one of the third evaluation result, the fourth evaluation result, the fifth evaluation result, the sixth evaluation result, the seventh evaluation result and the eighth evaluation result is false, it is determined that the data collected by sensor A cannot be associated with the data collected by sensor B.

[0078] However, the inventors have found that the above method actually uses a simple conditional superposition rule to determine whether the data collected by sensor A can be associated with the data collected by sensor B. Specifically, when sensor A is a laser radar, only when the first evaluation result, the second evaluation result, the third evaluation result and the fourth evaluation result are true at the same time, it is determined that the data collected by sensor A can be associated with the data collected by sensor B; when sensor A is a camera, only when the first evaluation result, the second evaluation result, the fifth evaluation result, the sixth evaluation result, the seventh evaluation result and the eighth evaluation result are true at the same time, it is determined that the data collected by sensor A can be associated with the data collected by sensor B; when sensor A is a radar, only when the third evaluation result, the fourth evaluation result, the fifth evaluation result, the sixth evaluation result, the seventh evaluation result and the eighth evaluation result are true at the same time, it is determined that the data collected by sensor A can be associated with the data collected by sensor B. In this process, as long as one evaluation result is wrong, it will lead to the occurrence of mis-association or missed association, so the accuracy of sensor data association cannot be ensured.

[0079] In view of the above problems, the present disclosure provides a data verification method, which can be applied to an electronic device. The electronic device can be an automatic driving controller. Figure 1 The flowchart diagram shown in the figure illustrates a data verification method provided by an embodiment of the present disclosure. It should be noted that although the logical order is shown in the flowchart diagram, in some cases, the steps shown or described in the flowchart can also be performed in other orders.

[0080] Step S101: acquiring current frame data collected by a first target sensor among multiple sensors.

[0081] Among them, multiple sensors are installed on the target object, such as an autonomous vehicle, an aircraft, a ship, a robot, etc. Here, the multiple sensors may include only a laser radar and a camera, or only a laser radar and a radar, or only a radar and a camera; the multiple sensors may also include a laser radar, a camera, and a radar at the same time.

[0082] In the embodiment of the present disclosure, the first target sensor may be any one of a plurality of sensors. For example, when the plurality of sensors include a laser radar, a camera, and a radar at the same time, the first target sensor may be any one of the laser radar, the camera, and the radar, and the current frame data may be the sensor data most recently collected by the first target sensor, which is used to characterize the obstacle state most recently perceived by the first target sensor. In addition, it is to be understood that in the embodiment of the present disclosure, when the first sensor is a laser radar, the current frame data collected by it may be three-dimensional point cloud data; when the first sensor is a camera, the current frame data collected by it may be image data; when the first sensor is a radar, the current frame data collected by it may include distance information, speed information, angle information, etc.

[0083] Step S102 : taking each sensor except the first target sensor among the multiple sensors as a second target sensor, to obtain a plurality of reference data for characterizing the similarity between the current frame data and the historical frame data collected by the second target sensor.

[0084] The historical frame data may be sensor data historically collected by the second target sensor, which is used to represent the obstacle state historically perceived by the second target sensor.

[0085] Exemplarily, the current frame data is the sensor data collected by the first target sensor at time T, which is used to characterize the obstacle state perceived by the first target sensor at time T; the historical frame data may be the sensor data collected by the second target sensor at time T-1, which is used to characterize the obstacle state perceived by the second target sensor at time T-1.

[0086] In addition, it should be noted that in the embodiment of the present disclosure, the historical frame data is derived from the global obstacle tracking information of the target object. Since the global obstacle tracking information of the target object stores the sensor data that have been associated, it can be determined that the historical frame data is also the sensor data that has been associated. Exemplarily, the historical frame data is the sensor data collected by the second target sensor at time T-1, then, the historical frame data is the sensor data that has been associated, which can be understood as: the historical frame data has been associated with the sensor data collected by each third target sensor among the multiple sensors except the second target sensor at time T-2. Among them, the sensor data collected by each third sensor among the multiple sensors except the second target sensor at time T-2 also comes from the global obstacle tracking information of the target object.

[0087] It should also be noted that, in the embodiment of the present disclosure, multiple reference data can characterize the similarity between the current frame data and the historical frame data collected by the second target sensor from different evaluation dimensions. Based on this, it can be understood that, in the embodiment of the present disclosure, for each reference data in the multiple reference data, it can be a scalar calculated based on the current frame data and the historical frame data under a certain evaluation dimension, so as to characterize the similarity between the current frame data and the historical frame data under the evaluation dimension.

[0088] Step S103: fuse multiple reference data to obtain a verification result for the current frame data.

[0089] The verification result is used to indicate whether the current frame data can be associated with the historical frame data.

[0090] In one example, when the verification result is greater than the overall similarity threshold, a first verification result for the current frame data can be obtained to indicate that the current frame data can be associated with the historical frame data; when the verification result is less than or equal to the overall similarity threshold, a second verification result for the current frame data can be obtained to indicate that the current frame data cannot be associated with the historical frame data. The overall similarity threshold can be set according to actual application requirements, and the embodiments of the present disclosure are not limited to this.

[0091] By adopting the data verification method provided by the embodiment of the present disclosure, the current frame data collected by the first target sensor among multiple sensors can be obtained; each sensor except the first target sensor among the multiple sensors is used as the second target sensor to obtain multiple reference data for characterizing the similarity between the current frame data and the historical frame data collected by the second target sensor; the multiple reference data are fused to obtain the verification result for the current frame data. Among them, the key is to obtain multiple reference data for characterizing the similarity between the current frame data and the historical frame data, and to fuse the multiple reference data to obtain the verification result for the current frame data, so as to characterize whether the current frame data can be associated with the historical frame data. Based on this, by adopting the data verification method provided by the embodiment of the present disclosure, even if a small amount of reference data is wrong, it will not necessarily lead to the occurrence of mis-association or missed association. Compared with the method of using simple conditional superposition rules in the prior art to judge whether the current frame data can be associated with the historical frame data, the fault tolerance is higher, and therefore, the accuracy of sensor data association can be improved.

[0092] In some optional implementations, the "obtaining a plurality of reference data for characterizing the similarity between the current frame data and the historical frame data acquired by the second target sensor" in step S102 may include:

[0093] determining a plurality of evaluation dimensions based on the first target sensor and the second target sensor;

[0094] Each evaluation dimension in the multiple evaluation dimensions is used as a target evaluation dimension to obtain reference data for characterizing the similarity between the current frame data and the historical frame data acquired by the second target sensor under the target evaluation dimension.

[0095] In one example, when the first target sensor and the second target sensor are any combination of a lidar and a camera (for example, the first target sensor is a lidar and the second target sensor is a camera; or the first target sensor is a camera and the second target sensor is a lidar), multiple evaluation dimensions including a shape evaluation dimension and a position evaluation dimension can be determined; when the first target sensor and the second target sensor are any combination of a lidar and a radar (for example, the first target sensor is a lidar and the second target sensor is a radar; or the first target sensor is a radar and the second target sensor is a lidar), multiple evaluation dimensions including a plane distance evaluation dimension and a first three-dimensional distance evaluation dimension can be determined; when the first target sensor and the second target sensor are any combination of a radar and a camera (for example, the first target sensor is a radar and the second target sensor is a camera; or the first target sensor is a camera and the second target sensor is a radar), multiple evaluation dimensions including a second three-dimensional distance evaluation dimension, a speed evaluation dimension, a first projection position evaluation dimension, and a second projection position evaluation dimension can be determined.

[0096] Among them, the reference data used to characterize the similarity between the current frame data collected by the first target sensor and the historical frame data collected by the second target sensor under the shape evaluation dimension can be specifically used to characterize the shape similarity between the first obstacle represented by the current frame data and the second obstacle represented by the historical frame data.

[0097] The reference data used to characterize the similarity between the current frame data collected by the first target sensor and the historical frame data collected by the second target sensor under the position evaluation dimension can be specifically used to characterize the position similarity between the third obstacle represented by the first perception data and the fourth obstacle represented by the second perception data. Here, the first perception data is the data collected by the laser radar in the current frame data and the historical frame data; the second perception data is the data collected by the camera in the current frame data and the historical frame data.

[0098] The reference data used to characterize the similarity between the current frame data collected by the first target sensor and the historical frame data collected by the second target sensor under the plane distance evaluation dimension can be specifically used to characterize the plane distance similarity between the fifth obstacle represented by the current frame data and the sixth obstacle represented by the historical frame data. Here, the larger the plane distance between the fifth obstacle and the sixth obstacle, the smaller the similarity of the plane distance between the fifth obstacle and the sixth obstacle; the smaller the plane distance between the fifth obstacle and the sixth obstacle, the larger the similarity of the plane distance between the fifth obstacle and the sixth obstacle.

[0099] The reference data used to characterize the similarity between the current frame data collected by the first target sensor and the historical frame data collected by the second target sensor under the first three-dimensional distance evaluation dimension can be specifically used to characterize the three-dimensional distance similarity between the fifth obstacle represented by the current frame data and the sixth obstacle represented by the historical frame data. Here, the larger the three-dimensional distance between the fifth obstacle and the sixth obstacle, the smaller the three-dimensional distance similarity between the fifth obstacle and the sixth obstacle; the smaller the three-dimensional distance between the fifth obstacle and the sixth obstacle, the larger the three-dimensional distance similarity between the fifth obstacle and the sixth obstacle.

[0100] The reference data used to characterize the similarity between the current frame data collected by the first target sensor and the historical frame data collected by the second target sensor under the second three-dimensional distance evaluation dimension can be specifically used to characterize the three-dimensional distance similarity between the seventh obstacle represented by the current frame data and the eighth obstacle represented by the historical frame data. Here, the larger the three-dimensional distance between the seventh obstacle and the eighth obstacle, the smaller the three-dimensional distance similarity between the seventh obstacle and the eighth obstacle; the smaller the three-dimensional distance between the seventh obstacle and the eighth obstacle, the larger the three-dimensional distance similarity between the seventh obstacle and the eighth obstacle.

[0101] The reference data used to characterize the similarity between the current frame data collected by the first target sensor and the historical frame data collected by the second target sensor under the speed evaluation dimension can be specifically used to characterize the speed similarity between the ninth obstacle represented by the third perception data and the tenth obstacle represented by the fourth perception data. Here, the third perception data is the data collected by the radar in the current frame data and the historical frame data; the fourth perception data is the data collected by the camera in the current frame data and the historical frame data.

[0102] The reference data used to characterize the similarity between the current frame data collected by the first target sensor and the historical frame data collected by the second target sensor under the first projection position evaluation dimension can be specifically used to characterize the similarity of the one-way distance in the first direction (for example, the X-axis direction of the world coordinate system) between the seventh obstacle represented by the current frame data and the eighth obstacle represented by the historical frame data. Here, the larger the one-way distance between the seventh obstacle and the eighth obstacle in the first direction, the smaller the one-way distance between the seventh obstacle and the eighth obstacle in the first direction; the smaller the one-way distance between the seventh obstacle and the eighth obstacle in the first direction, the greater the similarity of the one-way distance between the seventh obstacle and the eighth obstacle in the first direction.

[0103] The reference data for representing the similarity between the current frame data collected by the first target sensor and the historical frame data collected by the second target sensor under the second projection position evaluation dimension can be specifically used to represent the similarity of the one-way distance in the second direction (for example, the Y-axis direction of the world coordinate system) between the seventh obstacle represented by the current frame data and the eighth obstacle represented by the historical frame data. Here, the larger the one-way distance between the seventh obstacle and the eighth obstacle in the second direction, the smaller the one-way distance between the seventh obstacle and the eighth obstacle in the second direction; the smaller the one-way distance between the seventh obstacle and the eighth obstacle in the second direction, the greater the similarity of the one-way distance between the seventh obstacle and the eighth obstacle in the second direction.

[0104] Further, in the embodiment of the present disclosure, each of the multiple evaluation dimensions has a corresponding evaluation strategy to indicate how to obtain reference data for characterizing the similarity between the current frame data and the historical frame data collected by the second target sensor under the evaluation dimension. Based on this, in one example, after determining multiple evaluation dimensions based on the first target sensor and the second target sensor, each of the multiple evaluation dimensions can be used as a target evaluation dimension, and the reference data for characterizing the similarity between the current frame data and the historical frame data collected by the second target sensor under the target evaluation dimension can be obtained in the following manner:

[0105] Determine the evaluation strategy corresponding to the target evaluation dimension;

[0106] According to the evaluation strategy corresponding to the target evaluation dimension, the current frame data and the historical frame data collected by the second target sensor are mapped to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the target evaluation dimension.

[0107] The mapping of the current frame data and the historical frame data collected by the second target sensor may be performed by comparing the current frame data with the historical frame data. For example, the current frame data and the historical frame data may be compared after spatial mapping.

[0108] Through the above method, in the embodiment of the present disclosure, multiple evaluation dimensions can be determined based on the first target sensor and the second target sensor; each of the multiple evaluation dimensions is used as a target evaluation dimension to obtain reference data for characterizing the similarity between the current frame data and the historical frame data collected by the second target sensor under the target evaluation dimension. That is to say, in the embodiment of the present disclosure, the multiple reference data for characterizing the similarity between the current frame data and the historical frame data are obtained from different evaluation dimensions, so the current frame data and the historical frame data can be analyzed in a more fine-grained manner, and these evaluation dimensions have relatively reliable similarity characterization capabilities, so it can ensure that more accurate multiple reference data for characterizing the similarity between the current frame data and the historical frame data are obtained, thereby further improving the accuracy of sensor data association.

[0109] In addition, since each of the multiple evaluation dimensions has a corresponding evaluation strategy to indicate how to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the evaluation dimension. Therefore, after determining the multiple evaluation dimensions based on the first target sensor and the second target sensor, each of the multiple evaluation dimensions can be used as the target evaluation dimension, and the evaluation strategy corresponding to the target evaluation dimension can be determined. According to the evaluation strategy corresponding to the target evaluation dimension, the current frame data and the historical frame data are mapped to obtain the reference data for characterizing the similarity between the current frame data and the historical frame data under the target evaluation dimension. In this way, the efficiency of obtaining the reference data for characterizing the similarity between the current frame data and the historical frame data under the target evaluation dimension can be improved, thereby improving the execution efficiency of the data verification method.

[0110] In an optional implementation, the first target sensor and the second target sensor are any combination of a laser radar and a camera, and the target evaluation dimension is a shape evaluation dimension, and “mapping the current frame data and the historical frame data collected by the second target sensor to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the target evaluation dimension” may include:

[0111] Mapping the current frame data and the historical frame data collected by the second target sensor to obtain a shape difference between a first obstacle represented by the current frame data and a second obstacle represented by the historical frame data;

[0112] Based on the shape difference, reference data for characterizing the similarity between the current frame data and the historical frame data under the shape evaluation dimension is obtained.

[0113] The mapping of the current frame data and the historical frame data collected by the second target sensor may be performed by comparing the current frame data and the historical frame data after spatial mapping. For example, the current frame data and the historical frame data may be mapped to the same world coordinate system and then compared.

[0114] Further, in one example, “mapping the current frame data and the historical frame data collected by the second target sensor to obtain a shape difference between the first obstacle represented by the current frame data and the second obstacle represented by the historical frame data” may include:

[0115] Mapping the current frame data and the historical frame data collected by the second target sensor to obtain a width difference between the first bounding box and the second bounding box, and a height difference between the first bounding box and the second bounding box; wherein the first bounding box is a bounding box of a first obstacle represented by the current frame data; and the second bounding box is a bounding box of a second obstacle represented by the historical frame data;

[0116] According to the conversion method from Gaussian distribution to chi-square distribution, the width difference and the height difference are standardized to obtain the shape difference between the first obstacle represented by the current frame data and the second obstacle represented by the historical frame data.

[0117] The current frame data and the historical frame data collected by the second target sensor are mapped, and the width difference between the first bounding box and the second bounding box can be represented as:

[0118] Box_diff(w)=(w1-w2) / ((wl+wc) / 2)

[0119] Here, Box_diff(w) is used to represent the width difference between the first bounding box and the second bounding box; w1 is used to represent the width of the first bounding box; and w2 is used to represent the width of the second bounding box.

[0120] In addition, it should be noted that, in the embodiment of the present disclosure, when considering the theoretical jitter range of the width difference between the first bounding box and the second bounding box, the width difference between the first bounding box and the second bounding box obeys Gaussian distribution. Specifically, in an actual application example, the probability distribution of the width difference between the first bounding box and the second bounding box obeys Gaussian distribution, which can be specifically described as follows: Figure 2 As shown ( Figure 2 In the figure, the horizontal axis is used to represent the width difference (in m), and the vertical axis is used to represent the probability density). The theoretical jitter range of the width difference between the first bounding box and the second bounding box can be determined according to experimental values, and the embodiments of the present disclosure are not limited thereto.

[0121] The current frame data and the historical frame data collected by the second target sensor are mapped, and the height difference between the first bounding box and the second bounding box can be represented as:

[0122] Box_diff(h)=(h1-h2) / ((hl+hc) / 2)

[0123] Here, Box_diff(h) is used to represent the height difference between the first bounding box and the second bounding box; h1 is used to represent the height of the first bounding box; and h2 is used to represent the height of the second bounding box.

[0124] In addition, it should be noted that, in the embodiment of the present disclosure, when considering the theoretical jitter range of the height difference between the first bounding box and the second bounding box, the height difference between the first bounding box and the second bounding box obeys the Gaussian distribution. Specifically, in an actual application example, the probability distribution of the height difference between the first bounding box and the second bounding box obeys the Gaussian distribution, which can be specifically described as follows: Figure 3 As shown ( Figure 3 In the figure, the horizontal axis is used to represent the height difference (in m), and the vertical axis is used to represent the probability density). The theoretical jitter range of the height difference between the first bounding box and the second bounding box can be determined according to experimental values, and the embodiments of the present disclosure are not limited to this.

[0125] Based on the above, it can be understood that in the embodiment of the present disclosure, after obtaining the width difference between the first bounding box and the second bounding box, and the height difference between the first bounding box and the second bounding box, the width difference and the height difference can be standardized according to the conversion method from Gaussian distribution to chi-square distribution to obtain the shape difference between the first obstacle represented by the current frame data and the second obstacle represented by the historical frame data. This process can be characterized as:

[0126] Box_diff(shape)=

[0127] [Box_diff(w) 2 / dev(Box_diff(w)) 2 ]+[Box_diff(h) 2 / dev(Box_diff(h)) 2 ]

[0128] Among them, Box_diff(shape) is used to characterize the shape difference between the first obstacle represented by the current frame data and the second obstacle represented by the historical frame data; Box_diff(w) is used to characterize the width difference; dev(Box_diff(w)) is used to characterize the standard deviation of the width difference; Box_diff(h) is used to characterize the height difference; dev(Box_diff(h)) is used to characterize the standard deviation of the height difference.

[0129] Here, the standard deviation of the width difference between the first bounding box and the second bounding box can be obtained by calculation, for example, after obtaining the probability distribution of the width difference (for example, Figure 2 After that, the standard deviation of the width difference can be calculated based on the probability distribution of the width difference, which will not be described in detail here; similarly, the standard deviation of the height difference can be obtained by calculation, for example, after obtaining the probability distribution of the height difference (for example, Figure 3 After that, the standard deviation of the height difference can be calculated based on the probability distribution of the height difference, which is also not described in detail here.

[0130] Further, in one example, after obtaining the shape difference between the first obstacle represented by the current frame data and the second obstacle represented by the historical frame data, reference data for representing the similarity between the current frame data and the historical frame data in the shape evaluation dimension can be obtained based on the shape difference in the following manner:

[0131] Get the first chi-square statistic for shape differences;

[0132] Based on the first chi-square statistic, reference data for characterizing the similarity between the current frame data and the historical frame data in the shape evaluation dimension is obtained.

[0133] In a specific example, the above process can be characterized as follows:

[0134] Rd1=1-ChiSquaredCdf2TableFun(Box_diff(shape)

[0135] Among them, Rd1 is used to characterize the reference data used to characterize the similarity between the current frame data and the historical frame data under the shape evaluation dimension; ChiSquaredCdf2TableFun(Box_diff(shape)) is used to characterize the first chi-square statistic for shape difference obtained by looking up from the chi-square distribution table with a degree of freedom of 2 based on the shape difference (that is, Box_diff(shape)). In one example, the first chi-square statistic for shape difference can be obtained by looking up from the chi-square distribution table with a degree of freedom of 2 based on the shape difference through interpolation method to improve the search efficiency of the first chi-square statistic.

[0136] In addition, in an actual application example, the probability distribution of the width difference between the first bounding box and the second bounding box is as follows: Figure 2 As shown, the probability distribution of the height difference between the first bounding box and the second bounding box is as follows Figure 3 In the case shown in FIG. 1 , the probability distribution of the reference data used to characterize the similarity between the current frame data and the historical frame data under the shape evaluation dimension can be as follows: Figure 4 As shown ( Figure 4 In , the horizontal axis is used to represent the reference data used to represent the similarity between the current frame data and the historical frame data under the shape evaluation dimension, and the vertical axis is used to represent the probability density).

[0137] Through the above method, in the embodiment of the present disclosure, the current frame data and the historical frame data collected by the second target sensor can be mapped to obtain the shape difference between the first obstacle represented by the current frame data and the second obstacle represented by the historical frame data, and based on the shape difference, the reference data used to characterize the similarity between the current frame data and the historical frame data under the shape evaluation dimension is obtained. This process does not involve a complex data processing flow, and therefore, the efficiency of obtaining the reference data used to characterize the similarity between the current frame data and the historical frame data under the shape evaluation dimension can be improved, thereby improving the execution efficiency of the data verification method. Moreover, the above process involves the standardized processing of related data (for example, width difference and height difference), and therefore, it can also improve the accuracy of the reference data used to characterize the similarity between the current frame data and the historical frame data under the shape evaluation dimension, thereby further improving the accuracy of sensor data association.

[0138] In an optional implementation, the first target sensor and the second target sensor are any combination of a laser radar and a camera, and the target evaluation dimension is a position evaluation dimension, and “mapping the current frame data and the historical frame data collected by the second target sensor to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the target evaluation dimension” may include:

[0139] Determine the data collected by the laser radar in the current frame data and the historical frame data collected by the second target sensor as the first perception data, and determine the data collected by the camera in the current frame data and the historical frame data as the second perception data;

[0140] Mapping the first perception data and the second perception data to obtain a first plane distance between a third obstacle represented by the first perception data and a fourth obstacle represented by the second perception data in a two-dimensional plane;

[0141] Based on the first plane distance, reference data for characterizing the similarity between the current frame data and the historical frame data in the position evaluation dimension is obtained.

[0142] The mapping of the first perception data and the second perception data may be performed by comparing the first perception data and the second perception data after spatial mapping. For example, the first perception data and the second perception data may be mapped to the same world coordinate system and then compared.

[0143] Further, in an example, “mapping the first perception data and the second perception data to obtain a first plane distance between a third obstacle represented by the first perception data and a fourth obstacle represented by the second perception data in a two-dimensional plane” may include:

[0144] Mapping the first perception data and the second perception data to determine a plurality of feature points in the first perception data that are outside a third bounding box; wherein the third bounding box is a bounding box of a fourth obstacle represented by the second perception data;

[0145] Acquire a first overflow amount of the plurality of feature points in the first direction relative to the third bounding box, and a second overflow amount of the plurality of feature points in the second direction relative to the third bounding box;

[0146] The first overflow amount and the second overflow amount are standardized according to a conversion method from a Gaussian distribution to a chi-square distribution, so as to obtain a first plane distance between the third obstacle and the fourth obstacle in a two-dimensional plane represented by the first perception data.

[0147] In a specific example, after determining multiple feature points in the first perception data that are outside the third bounding box, the first shortest distance of each of the multiple feature points in the first direction (for example, the X-axis direction of the world coordinate system) relative to the third bounding box can be obtained to obtain multiple first shortest distances corresponding to the multiple feature points one by one, and the average of the multiple first shortest distances is used as the first overflow amount of the multiple feature points in the first direction relative to the third bounding box.

[0148] In addition, it should be noted that, in the embodiment of the present disclosure, while considering the theoretical jitter range of the first overflow amount of multiple feature points in the first direction relative to the third bounding box, the probability distribution of the first overflow amount of multiple feature points in the first direction relative to the third bounding box is symmetrically processed to obtain a new probability distribution that obeys the Gaussian distribution. Specifically, in an actual application example, the probability partial situation of the first overflow amount of multiple feature points in the first direction relative to the third bounding box can be as follows Figure 5 As shown ( Figure 5 In the equation, the horizontal axis is used to represent the first overflow amount (in m), and the vertical axis is used to represent the probability density). After symmetric processing, we can get the following: Figure 6 The new probability distribution that follows the Gaussian distribution is shown as ( Figure 6 , the horizontal axis is used to represent the first overflow amount (in m), and the vertical axis is used to represent the probability density). The theoretical jitter range of the first overflow amount of the plurality of feature points in the first direction relative to the third bounding box can be determined according to experimental values, and the embodiments of the present disclosure are not limited thereto.

[0149] Similarly, after determining the multiple feature points in the first perception data that are outside the third bounding box, the second shortest distance of each of the multiple feature points in the second direction (for example, the Y-axis direction of the world coordinate system) relative to the third bounding box can be obtained to obtain multiple second shortest distances corresponding to the multiple feature points one by one, and the average of the multiple second shortest distances is used as the second overflow amount of the multiple feature points in the second direction relative to the third bounding box.

[0150] In addition, it should be noted that, in the embodiment of the present disclosure, while considering the theoretical jitter range of the second overflow amount of multiple feature points in the second direction relative to the third bounding box, the probability distribution of the second overflow amount of multiple feature points in the second direction relative to the third bounding box is symmetrically processed to obtain a new probability distribution that obeys the Gaussian distribution. Specifically, in an actual application example, the probability partial situation of the second overflow amount of multiple feature points in the second direction relative to the third bounding box can be as follows Figure 7 As shown ( Figure 7 In the equation, the horizontal axis is used to represent the second overflow (in m), and the vertical axis is used to represent the probability density). After symmetric processing, we can get the following: Figure 8 The new probability distribution that follows the Gaussian distribution is shown as ( Figure 8 , the horizontal axis is used to represent the first overflow amount (in m), and the vertical axis is used to represent the probability density). The theoretical jitter range of the second overflow amount of the plurality of feature points in the second direction relative to the third bounding box can be determined according to experimental values, and the embodiments of the present disclosure are not limited thereto.

[0151] Based on the above, it can be understood that in the embodiment of the present disclosure, after obtaining the first overflow amount of multiple feature points in the first direction relative to the third bounding box, and the second overflow amount of multiple feature points in the second direction relative to the third bounding box, the first overflow amount and the second overflow amount can be standardized according to the conversion method from Gaussian distribution to chi-square distribution to obtain the first plane distance between the third obstacle and the fourth obstacle represented by the first perception data in the two-dimensional plane. This process can be characterized as:

[0152] Box_diff(Location)=

[0153] [Box_diff(x) 2 / dev(Box_diff(x)) 2 ]+[Box_diff(y) 2 / dev(Box_diff(y)) 2 ]

[0154] Among them, Box_diff(Location) is used to characterize the first plane distance between the third obstacle and the fourth obstacle represented by the first perception data in the two-dimensional plane; Box_diff(x) is used to characterize the first overflow amount; dev(Box_diff(x)) is used to characterize the standard deviation of the first overflow amount; Box_diff(y) is used to characterize the second overflow amount; dev(Box_diff(y)) is used to characterize the standard deviation of the second overflow amount.

[0155] Here, the standard deviation of the first overflow amount can be obtained by calculation, for example, in the case of obtaining the probability part of the first overflow amount (for example, Figure 6 After the probability partial situation shown in FIG. 1 is obtained, the standard deviation of the first overflow amount can be calculated based on the probability partial situation with the first overflow amount, which will not be described in detail here; similarly, after the probability partial situation with the second overflow amount is obtained (for example, Figure 8 After that, the standard deviation of the second overflow amount can be calculated based on the probability partial situation of the second overflow amount, which is also not described in detail here.

[0156] Further, in one example, after obtaining the first plane distance between the third obstacle and the fourth obstacle represented by the first perception data in the two-dimensional plane, reference data for representing the similarity between the current frame data and the historical frame data in the position evaluation dimension can be obtained based on the first plane distance in the following manner:

[0157] Obtain a second chi-square statistic for the first plane distance;

[0158] Based on the second chi-square statistic, reference data for characterizing the similarity between the current frame data and the historical frame data in the position evaluation dimension is obtained.

[0159] In a specific example, the above process can be characterized as follows:

[0160] Rd2=1-ChiSquaredCdf2TableFun(Box_diff(Location)

[0161] Among them, Rd2 is used to characterize the reference data used to characterize the similarity between the current frame data and the historical frame data under the location evaluation dimension; ChiSquaredCdf2TableFun(Box_diff(Location)) is used to characterize the second chi-square statistic for the first plane distance obtained by looking up from the chi-square distribution table with a degree of freedom of 2 based on the first plane distance (that is, Box_diff(Location)). In one example, the second chi-square statistic for the first plane distance can be obtained by looking up from the chi-square distribution table with a degree of freedom of 2 based on the first plane distance through interpolation to improve the search efficiency of the second chi-square statistic.

[0162] In addition, in an actual application example, the first overflow amount of the plurality of feature points relative to the third bounding box in the first direction is as follows: Figure 6 As shown, the probability of the second overflow amount of the plurality of feature points in the second direction relative to the third bounding box is as follows: Figure 8 In the case shown in FIG. 1 , the probability distribution of the reference data used to characterize the similarity between the current frame data and the historical frame data under the location evaluation dimension can be as follows: Fig. 9 As shown ( Fig. 9 In , the horizontal axis is used to represent the reference data used to represent the similarity between the current frame data and the historical frame data under the position evaluation dimension, and the vertical axis is used to represent the probability density).

[0163] Through the above method, in the embodiment of the present disclosure, the data collected by the laser radar in the current frame data and the historical frame data collected by the second target sensor can be determined as the first perception data, and the data collected by the camera in the current frame data and the historical frame data can be determined as the second perception data, and the first perception data and the second perception data are mapped to obtain the first plane distance in the two-dimensional plane between the third obstacle represented by the first perception data and the fourth obstacle represented by the second perception data, and then based on the first plane distance, the reference data used to characterize the similarity between the current frame data and the historical frame data under the position evaluation dimension is obtained. This process does not involve a complex data processing flow, and therefore, the efficiency of obtaining the reference data used to characterize the similarity between the current frame data and the historical frame data under the position evaluation dimension can be improved, thereby improving the execution efficiency of the data verification method. Moreover, the above process involves the standardized processing of related data (for example, the first overflow amount and the second overflow amount), and therefore, it can also improve the accuracy of the reference data used to characterize the similarity between the current frame data and the historical frame data under the position evaluation dimension, thereby further improving the accuracy of sensor data association.

[0164] In an optional implementation, the first target sensor and the second target sensor are any combination of a laser radar and a radar, and the target evaluation dimension is a plane distance evaluation dimension, and “mapping the current frame data and the historical frame data collected by the second target sensor to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the target evaluation dimension” may include:

[0165] Mapping the current frame data and the historical frame data collected by the second target sensor to obtain a first initial distance between a fifth obstacle represented by the current frame data and a sixth obstacle represented by the historical frame data in a two-dimensional plane;

[0166] The first initial distance is normalized to obtain a second plane distance;

[0167] Based on the second plane distance, reference data for characterizing the similarity between the current frame data and the historical frame data under the plane distance evaluation dimension is obtained.

[0168] The mapping of the current frame data and the historical frame data collected by the second target sensor may be performed by comparing the current frame data and the historical frame data after spatial mapping. For example, the current frame data and the historical frame data may be mapped to the same world coordinate system and then compared.

[0169] Further, in an example, “mapping the current frame data and the historical frame data collected by the second target sensor to obtain a first initial distance between a fifth obstacle represented by the current frame data and a sixth obstacle represented by the historical frame data in a two-dimensional plane” may include:

[0170] Determine a first target vertex on a fourth bounding box that is closest to the reference position; wherein the fourth bounding box is a bounding box of a fifth obstacle represented by the current frame data;

[0171] Determine a second target vertex on a fifth bounding box that is closest to the reference position; wherein the fifth bounding box is a bounding box of a sixth obstacle represented by historical frame data collected by the second target sensor;

[0172] The first target vertex and the second target vertex are mapped to obtain a spacing distance between the first target vertex and the second target vertex, which is used as a first initial distance between the fifth obstacle and the sixth obstacle in the two-dimensional plane.

[0173] The reference position may be a ground plane.

[0174] After determining the first target vertex closest to the reference position on the fourth bounding box and determining the second target vertex closest to the reference position on the fifth bounding box, the Euclidean distance between the first target vertex and the second target vertex can be calculated as the interval distance between the first target vertex and the second target vertex obtained by mapping the first target vertex and the second target vertex, that is, as the first initial distance between the fifth obstacle and the sixth obstacle in the two-dimensional plane.

[0175] In addition, it should be noted that, in the embodiment of the present disclosure, considering the theoretical jitter range of the first initial distance between the fifth obstacle and the sixth obstacle in the two-dimensional plane, in an actual application example, the probability distribution of the first initial distance between the fifth obstacle and the sixth obstacle in the two-dimensional plane can be as follows: Fig.10 As shown ( Fig.10 , the horizontal axis is used to represent the first initial distance (in m), and the vertical axis is used to represent the probability density). The theoretical jitter range of the first initial distance between the fifth obstacle and the sixth obstacle in the two-dimensional plane can be determined according to experimental values, and the embodiments of the present disclosure are not limited to this.

[0176] After obtaining the first initial distance between the fifth obstacle represented by the current frame data and the sixth obstacle represented by the historical frame data in the two-dimensional plane, the first initial distance can be normalized to obtain the second plane distance. This process can be characterized as:

[0177] SPd(Postion21)=Pd(Postion21) / Lnv21

[0178] Among them, SPd(Position21) is used to characterize the second plane distance; Pd(Position21) is used to characterize the first initial distance; Lnv21 is used to characterize the first standardized parameter. Here, the first standardized parameter can be a value greater than the first initial distance, which can be set according to actual application requirements, and the embodiments of the present disclosure are not limited to this.

[0179] After the first initial distance is normalized to obtain the second plane distance, reference data for characterizing the similarity between the current frame data and the historical frame data under the plane distance evaluation dimension can be obtained based on the second plane distance. This process can be characterized as follows:

[0180] Rd3=1-SPd(Position21)

[0181] Among them, Rd3 is used to represent the reference data used to represent the similarity between the current frame data and the historical frame data under the plane distance evaluation dimension; SPd (Position21) is used to represent the second plane distance.

[0182] Through the above method, in the embodiment of the present disclosure, the current frame data and the historical frame data collected by the second target sensor can be mapped to obtain the first initial distance between the fifth obstacle represented by the current frame data and the sixth obstacle represented by the historical frame data in the two-dimensional plane, and the first initial distance is standardized to obtain the second plane distance, and then based on the second plane distance, the reference data used to characterize the similarity between the current frame data and the historical frame data under the plane distance evaluation dimension is obtained. This process does not involve a complex data processing flow, and therefore, the efficiency of obtaining the reference data used to characterize the similarity between the current frame data and the historical frame data under the plane distance evaluation dimension can be improved, thereby improving the execution efficiency of the data verification method. Moreover, the above process involves the standardized processing of related data (for example, the first initial distance), and therefore, it can also improve the accuracy of the reference data used to characterize the similarity between the current frame data and the historical frame data under the plane distance evaluation dimension, thereby further improving the accuracy of sensor data association.

[0183] In an optional implementation, the first target sensor and the second target sensor are any combination of a laser radar and a radar, and the target evaluation dimension is a first three-dimensional distance evaluation dimension, and “mapping the current frame data and the historical frame data collected by the second target sensor to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the target evaluation dimension” may include:

[0184] Mapping the current frame data and the historical frame data collected by the second target sensor to obtain a second initial distance in the three-dimensional space between a fifth obstacle represented by the current frame data and a sixth obstacle represented by the historical frame data;

[0185] The second initial distance is normalized to obtain a first three-dimensional distance;

[0186] Based on the first three-dimensional distance, reference data for characterizing the similarity between the current frame data and the historical frame data under the first three-dimensional distance evaluation dimension is obtained.

[0187] The mapping of the current frame data and the historical frame data collected by the second target sensor may be performed by comparing the current frame data and the historical frame data after spatial mapping. For example, the current frame data and the historical frame data may be mapped to the same world coordinate system and then compared.

[0188] Further, in one example, “mapping the current frame data and the historical frame data collected by the second target sensor to obtain a second initial distance in the three-dimensional space between the fifth obstacle represented by the current frame data and the sixth obstacle represented by the historical frame data” may include:

[0189] Determine a first target center point of a fifth obstacle represented by the current frame data;

[0190] Determine a second target center point of a sixth obstacle represented by historical frame data collected by a second target sensor;

[0191] The first target center point and the second target center point are mapped to obtain a spacing distance between the first target center point and the second target center point as a second initial distance between the fifth obstacle and the sixth obstacle in the three-dimensional space.

[0192] In a specific example, the center point of the obstacle represented by the data collected by the laser radar in the current frame data and the historical frame data can be determined by using at least one of a voxel grid filter (VGF) algorithm, a clustering algorithm (e.g., a K-Means clustering algorithm, a density-based spatial clustering application and noise (DBS CAN) clustering algorithm), a random sample consensus (RANS AC) algorithm, a principal component analysis (PCA) algorithm, etc. In another specific example, the center point of the obstacle represented by the data collected by the radar in the current frame data and the historical frame data can be determined by using at least one of a target detection algorithm, a Kalman filter algorithm, a clustering algorithm, a fusion algorithm, etc.

[0193] After determining the first target center point of the fifth obstacle represented by the current frame data and determining the second target center point of the sixth obstacle represented by the historical frame data collected by the second target sensor, the Euclidean distance between the first target center point and the second target center point can be calculated as a mapping of the first target center point and the second target center point to obtain the interval distance between the first target center point and the second target center point, that is, as the second initial distance between the fifth obstacle and the sixth obstacle in the three-dimensional space.

[0194] In addition, it should be noted that, in the embodiment of the present disclosure, considering the theoretical jitter range of the second initial distance between the fifth obstacle and the sixth obstacle in the three-dimensional space, in an actual application example, the probability distribution of the second initial distance between the fifth obstacle and the sixth obstacle in the three-dimensional space can be as follows: Fig.11 As shown ( Fig.11 , the horizontal axis is used to represent the second initial distance (in m), and the vertical axis is used to represent the probability density). The theoretical jitter range of the second initial distance between the fifth obstacle and the sixth obstacle in the three-dimensional space can be determined according to experimental values, and the embodiments of the present disclosure are not limited thereto.

[0195] After obtaining the second initial distance between the fifth obstacle represented by the current frame data and the sixth obstacle represented by the historical frame data in the three-dimensional space, the second initial distance can be normalized to obtain the first three-dimensional distance. This process can be characterized as:

[0196] SPd(Postion22)=Pd(Postion22) / Lnv22

[0197] Among them, SPd(Position22) is used to characterize the first three-dimensional distance; Pd(Position22) is used to characterize the second initial distance; Lnv22 is used to characterize the second normalized parameter. Here, the second normalized parameter can be a value greater than the second initial distance, which can be set according to actual application requirements, and the embodiments of the present disclosure are not limited to this.

[0198] After the second initial distance is normalized to obtain the first three-dimensional distance, reference data for characterizing the similarity between the current frame data and the historical frame data under the first three-dimensional distance evaluation dimension can be obtained based on the first three-dimensional distance. This process can be characterized as:

[0199] Rd4=1-SPd(Position22)

[0200] Among them, Rd4 is used to characterize the reference data used to characterize the similarity between the current frame data and the historical frame data under the first three-dimensional distance evaluation dimension; SPd (Position22) is used to characterize the first three-dimensional distance.

[0201] Through the above method, in the embodiment of the present disclosure, the current frame data and the historical frame data collected by the second target sensor can be mapped to obtain the second initial distance in the three-dimensional space between the fifth obstacle represented by the current frame data and the sixth obstacle represented by the historical frame data, and the second initial distance is standardized to obtain the first three-dimensional distance, and then based on the first three-dimensional distance, the reference data used to characterize the similarity between the current frame data and the historical frame data under the first three-dimensional distance evaluation dimension is obtained. This process does not involve a complex data processing flow, and therefore, the efficiency of obtaining the reference data used to characterize the similarity between the current frame data and the historical frame data under the first three-dimensional distance evaluation dimension can be improved, thereby improving the execution efficiency of the data verification method. Moreover, the above process involves the standardized processing of related data (for example, the second initial distance), and therefore, it can also improve the accuracy of the reference data used to characterize the similarity between the current frame data and the historical frame data under the first three-dimensional distance evaluation dimension, thereby further improving the accuracy of sensor data association.

[0202] In an optional implementation, the first target sensor and the second target sensor are any combination of a radar and a camera, and the target evaluation dimension is a second three-dimensional distance evaluation dimension, and “mapping the current frame data and the historical frame data collected by the second target sensor to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the target evaluation dimension” may include:

[0203] Mapping the current frame data and the historical frame data collected by the second target sensor to obtain a third initial distance in the three-dimensional space between a seventh obstacle represented by the current frame data and an eighth obstacle represented by the historical frame data;

[0204] The third initial distance is normalized to obtain a second three-dimensional distance;

[0205] Based on the second three-dimensional distance, reference data for characterizing the similarity between the current frame data and the historical frame data under the second three-dimensional distance evaluation dimension is obtained.

[0206] The mapping of the current frame data and the historical frame data collected by the second target sensor may be performed by comparing the current frame data and the historical frame data after spatial mapping. For example, the current frame data and the historical frame data may be mapped to the same world coordinate system and then compared.

[0207] Further, in one example, “mapping the current frame data and the historical frame data collected by the second target sensor to obtain a third initial distance between the seventh obstacle represented by the current frame data and the eighth obstacle represented by the historical frame data in the three-dimensional space” may include:

[0208] Determine a third target center point of a seventh obstacle represented by the current frame data;

[0209] Determine a fourth target center point of an eighth obstacle represented by the historical frame data collected by the second target sensor;

[0210] The third target center point and the fourth target center point are mapped to obtain a spacing distance between the third target center point and the fourth target center point as a third initial distance between the seventh obstacle and the eighth obstacle in the three-dimensional space.

[0211] In a specific example, at least one of a target detection algorithm, a Kalman filter algorithm, a clustering algorithm, a fusion algorithm, etc. can be used to determine the center point of the obstacle represented by the data collected by the radar in the current frame data and the historical frame data; in another specific example, at least one of contour detection and shape analysis, a target detection algorithm, a semantic segmentation algorithm, etc. can be used to determine the center point of the obstacle represented by the data collected by the camera in the current frame data and the historical frame data.

[0212] After determining the third target center point of the seventh obstacle represented by the current frame data and determining the fourth target center point of the eighth obstacle represented by the historical frame data collected by the second target sensor, the Euclidean distance between the third target center point and the fourth target center point can be calculated as the interval distance between the third target center point and the fourth target center point obtained by mapping the third target center point and the fourth target center point, that is, as the third initial distance between the seventh obstacle and the eighth obstacle in the three-dimensional space.

[0213] In addition, it should be noted that, in the embodiment of the present disclosure, considering the theoretical jitter range of the third initial distance between the seventh obstacle and the eighth obstacle in the three-dimensional space, in an actual application example, the probability distribution of the third initial distance between the seventh obstacle and the eighth obstacle in the three-dimensional space can be as follows: Fig.12 As shown ( Fig.12 In , the horizontal axis is used to represent the third initial distance (in meters), and the vertical axis is used to represent the probability density). Fig.12 The probability distribution shown is trimmed (for example, Fig.12 In the above example, the part with the third initial distance greater than or equal to 6.5m is cut off and the probability is redistributed), and then we can get Fig.13 , which is used to characterize the probability distribution of the third initial distance between the seventh obstacle and the eighth obstacle in the three-dimensional space ( Fig.13, the horizontal axis is used to represent the third initial distance (in meters), and the vertical axis is used to represent the probability density). The theoretical jitter range of the third initial distance between the seventh obstacle and the eighth obstacle in the three-dimensional space can be determined according to experimental values, and the embodiments of the present disclosure are not limited thereto.

[0214] After obtaining the third initial distance in the three-dimensional space between the seventh obstacle represented by the current frame data and the eighth obstacle represented by the historical frame data, the third initial distance can be standardized to obtain the second three-dimensional distance. This process can be characterized as:

[0215] SPd(Postion31)=Pd(Postion31) / Lnv31

[0216] Among them, SPd(Position31) is used to characterize the second three-dimensional distance; Pd(Position31) is used to characterize the third initial distance; Lnv31 is used to characterize the third normalized parameter. Here, the third normalized parameter can be a value greater than the second three-dimensional distance, which can be set according to actual application requirements, and the embodiments of the present disclosure are not limited to this.

[0217] After the third initial distance is normalized to obtain the second three-dimensional distance, it is necessary to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the second three-dimensional distance evaluation dimension based on the second three-dimensional distance. In one example, "obtaining reference data for characterizing the similarity between the current frame data and the historical frame data under the second three-dimensional distance evaluation dimension based on the second three-dimensional distance" may include:

[0218] When the second three-dimensional distance is less than the first distance threshold, determining the first similarity representation value as the first intermediate data;

[0219] Alternatively, when the second three-dimensional distance is greater than or equal to the first distance threshold, the second three-dimensional distance is processed by a Welsh loss function to obtain first intermediate data;

[0220] Based on the first intermediate data, reference data for characterizing the similarity between the current frame data and the historical frame data in a second three-dimensional distance evaluation dimension is obtained.

[0221] Among them, the first distance threshold and the first similarity characterization value can be set according to actual application requirements. For example, the first distance threshold can be set to 0.05m, and the first similarity characterization value can be set to a probability value close to 1. For example, the first similarity characterization value can be set to 1-(1e-6), and the embodiments of the present disclosure are not limited to this.

[0222] When the second three-dimensional distance is greater than or equal to the first distance threshold, the second three-dimensional distance can be processed by the Welsh loss function to obtain the first intermediate data. This process can be characterized as:

[0223] ID1=Exp[(SPd(Postion31)-Dth31) / Dsc31]

[0224] Among them, ID1 is used to characterize the first intermediate data obtained by processing the second three-dimensional distance through the Welsh loss function; SPd (Position 31) is used to characterize the second three-dimensional distance; Dth31 is used to characterize the first distance threshold; Dsc31 is used to characterize the first step length parameter. Here, the first step length parameter can be set according to actual application requirements. For example, the first step length parameter can be set to 0.05m, which is not limited in the embodiments of the present disclosure.

[0225] In the above example, when the second three-dimensional distance is less than the first distance threshold, the first similarity characterization value (that is, a probability value close to 1) can be used as the first intermediate data to characterize that the current frame data and the historical frame data have high similarity under the second three-dimensional distance evaluation dimension; when the second three-dimensional distance is greater than or equal to the first distance threshold, the second three-dimensional distance is processed by the Welsh loss function to obtain the first intermediate data. Specifically, the second three-dimensional distance can be made to decay sharply in an exponential decay manner to obtain a smaller first intermediate data to characterize that the current frame data and the historical frame data have low similarity under the second three-dimensional distance evaluation dimension.

[0226] After obtaining the first intermediate data, reference data for characterizing the similarity between the current frame data and the historical frame data in the second three-dimensional distance evaluation dimension may be obtained based on the first intermediate data in the following manner:

[0227] When the first intermediate data is less than or equal to the first similarity threshold, determining the first intermediate data as reference data for characterizing the similarity between the current frame data and the historical frame data under the second three-dimensional distance evaluation dimension;

[0228] Alternatively, when the first intermediate data is greater than the first similarity threshold, the first intermediate data is compressed by a scaling function to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under a second three-dimensional distance evaluation dimension.

[0229] The first similarity threshold may be set according to actual application requirements. For example, the first distance threshold may be set to 0.5, which is not limited in the embodiments of the present disclosure.

[0230] Further, in a specific example, “when the first intermediate data is less than or equal to the first similarity threshold, determining the first intermediate data as reference data for characterizing the similarity between the current frame data and the historical frame data under the second three-dimensional distance evaluation dimension” can be characterized as:

[0231] Rd5=ID1

[0232] Among them, Rd5 is used to represent the reference data used to represent the similarity between the current frame data and the historical frame data in the second three-dimensional distance evaluation dimension; ID1 is used to represent the first intermediate data.

[0233] “When the first intermediate data is greater than the first similarity threshold, the first intermediate data is compressed by a scaling function to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the second three-dimensional distance evaluation dimension” can be characterized as:

[0234] Rd5=[(ID1-Pth1) / (1-Pth1)](Pmax-Pth1)+Pth1

[0235] Among them, Rd5 is used to characterize the reference data used to characterize the similarity between the current frame data and the historical frame data under the second three-dimensional distance evaluation dimension; ID1 is used to characterize the first intermediate data; Pth1 is used to characterize the first zoom threshold; Pmax1 is used to characterize the first maximum zoom probability value. Here, the first zoom threshold and the first maximum zoom probability value can be set according to actual application requirements. For example, the first zoom threshold can be set to 0.5, and the first maximum zoom probability value can be set to 0.7. The embodiments of the present disclosure are not limited to this.

[0236] In addition, in an actual example, the probability distribution of the third initial distance between the seventh obstacle represented by the current frame data and the eighth obstacle represented by the historical frame data in the three-dimensional space is as follows: Fig.13 In the case shown in FIG. 1 , the probability distribution of the reference data used to characterize the similarity between the current frame data and the historical frame data under the second three-dimensional distance evaluation dimension can be as follows: Fig.14 As shown ( Fig.14 In the figure, the horizontal axis is used to represent the reference data used to represent the similarity between the current frame data and the historical frame data in the second three-dimensional distance evaluation dimension under the position evaluation dimension, and the vertical axis is used to represent the probability density).

[0237] Through the above method, in the embodiment of the present disclosure, the current frame data and the historical frame data collected by the second target sensor can be mapped to obtain the third initial distance in the three-dimensional space between the seventh obstacle represented by the current frame data and the eighth obstacle represented by the historical frame data, and the third initial distance is standardized to obtain the second three-dimensional distance, and then based on the second three-dimensional distance, the reference data used to characterize the similarity between the current frame data and the historical frame data under the second three-dimensional distance evaluation dimension is obtained. This process does not involve a complex data processing flow, and therefore, the efficiency of obtaining the reference data used to characterize the similarity between the current frame data and the historical frame data under the second three-dimensional distance evaluation dimension can be improved, thereby improving the execution efficiency of the data verification method. Moreover, the above process involves the standardized processing of related data (for example, the third initial distance), and therefore, it can also improve the accuracy of the reference data used to characterize the similarity between the current frame data and the historical frame data under the second three-dimensional distance evaluation dimension, thereby further improving the accuracy of sensor data association.

[0238] In addition, through the above method, in the embodiment of the present disclosure, when obtaining reference data for characterizing the similarity between current frame data and historical frame data under the second three-dimensional distance evaluation dimension based on the second three-dimensional distance, the Welsh loss function and the scaling function are successively used to optimize the relevant data (for example, the second three-dimensional distance and the first intermediate data), and finally, the reference data for characterizing the similarity between current frame data and historical frame data under the second three-dimensional distance evaluation dimension is obtained. Therefore, the robustness and rationality of the reference data for characterizing the similarity between current frame data and historical frame data under the second three-dimensional distance evaluation dimension can be improved.

[0239] In an optional implementation, the first target sensor and the second target sensor are any combination of a radar and a camera, and the target evaluation dimension is a speed evaluation dimension, and “mapping the current frame data and the historical frame data collected by the second target sensor to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the target evaluation dimension” may include:

[0240] Determine the data collected by the radar in the current frame data and the historical frame data collected by the second target sensor as the third perception data, and determine the data collected by the camera in the current frame data and the historical frame data as the fourth perception data;

[0241] Acquire a first driving speed of a ninth obstacle represented by the third perception data;

[0242] Acquire a second driving speed of a tenth obstacle represented by the fourth perception data;

[0243] The first driving speed and the second driving speed are mapped to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the speed evaluation dimension.

[0244] Among them, mapping the first driving speed and the second driving speed can be to determine the correlation between the first driving speed and the second driving speed, so as to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the speed evaluation dimension based on the correlation between the first driving speed and the second driving speed.

[0245] Further, in an example, “mapping the first driving speed and the second driving speed to obtain reference data for characterizing the similarity between the current frame data and the historical frame data in the speed evaluation dimension” may include:

[0246] (1) Mapping the first driving speed and the second driving speed so that, when it is determined that the first driving speed and the second driving speed meet a first preset relationship, a second similarity characterization value is determined as reference data for characterizing the similarity between the current frame data and the historical frame data under the speed evaluation dimension.

[0247] Among them, the first preset relationship can be that the first driving speed is less than the first speed threshold, and the second driving speed is greater than the second speed threshold, and the first speed threshold is less than the second speed threshold. Here, the first speed threshold and the second speed threshold can be set according to actual application requirements. For example, the first speed threshold can be set to 0.1m / s, and the second speed threshold can be set to 1m / s. The embodiment of the present disclosure is also not limited to this. In addition, in the embodiment of the present disclosure, the second similarity characterization value can be set according to actual application requirements. For example, the first speed threshold can be set to 0.00001. The embodiment of the present disclosure is also not limited to this.

[0248] Further, in a specific example, “when it is determined that the first driving speed and the second driving speed meet the first preset relationship, determining the second similarity characterization value as reference data for characterizing the similarity between the current frame data and the historical frame data under the speed evaluation dimension” can be characterized as:

[0249] Rd6=SCV2

[0250] Among them, Rd6 is used to represent the reference data used to represent the similarity between the current frame data and the historical frame data under the speed evaluation dimension; SCV2 is used to represent the second similarity representation value.

[0251] (2) Mapping the first driving speed and the second driving speed to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the speed evaluation dimension based on an initial speed difference between the first driving speed and the second driving speed when it is determined that the first driving speed and the second driving speed meet a second preset relationship.

[0252] The second preset relationship may be that the second driving speed is greater than the first speed threshold, and at least one of the first driving speed and the second driving speed is greater than the third speed threshold, and the third speed threshold is greater than the second speed threshold. Here, the third speed threshold can be set according to actual application requirements, for example, the third speed threshold can be set to 2m / s, and the embodiments of the present disclosure are also not limited to this.

[0253] Further, in a specific example, “obtaining reference data for characterizing the similarity between the current frame data and the historical frame data in the speed evaluation dimension based on the initial speed difference between the first driving speed and the second driving speed” may include:

[0254] normalizing an initial speed difference between the first driving speed and the second driving speed to obtain a target speed difference between the first driving speed and the second driving speed;

[0255] Based on the target speed difference, reference data for characterizing the similarity between the current frame data and the historical frame data in the speed evaluation dimension is obtained.

[0256] In a more specific example, “normalizing the initial speed difference between the first driving speed and the second driving speed to obtain a target speed difference between the first driving speed and the second driving speed” can be represented as:

[0257] Target_diff(Speed)=M_diff(Speed) / Max(Speed)

[0258] Among them, Target_diff(Speed) is used to represent the target speed difference between the first driving speed and the second driving speed; M_diff(Speed) is used to represent the initial speed difference between the first driving speed and the second driving speed; Max(Speed) is used to represent the driving speed with the larger value between the first driving speed and the second driving speed.

[0259] In addition, it should be noted that, in the embodiment of the present disclosure, considering the theoretical jitter range of the initial speed difference between the first driving speed and the second driving speed, in an actual application example, the probability distribution of the initial speed difference between the first driving speed and the second driving speed can be as follows: Fig.15 As shown ( Fig.15, the horizontal axis is used to represent the initial speed difference (in m / s), and the vertical axis is used to represent the probability density). The theoretical jitter range of the initial speed difference between the first driving speed and the second driving speed can be determined according to experimental values, and the embodiments of the present disclosure are not limited to this.

[0260] It should also be noted that in the embodiment of the present disclosure, based on the probability distribution of the initial speed difference between the first driving speed and the second driving speed, the probability distribution of the target speed difference between the first driving speed and the second driving speed can be obtained, which will not be elaborated here.

[0261] In a more specific example, “obtaining reference data for characterizing the similarity between the current frame data and the historical frame data in the speed evaluation dimension based on the target speed difference” may include:

[0262] Obtain the third chi-square statistic for the target speed difference;

[0263] Based on the third chi-square statistic, second intermediate data for characterizing the similarity between the current frame data and the historical frame data under the speed evaluation dimension is obtained;

[0264] When the second intermediate data is less than or equal to the second similarity threshold, the second intermediate data is determined as reference data for characterizing the similarity between the current frame data and the historical frame data under the speed evaluation dimension;

[0265] Alternatively, when the second intermediate data is greater than the second similarity threshold, the second intermediate data is compressed by a scaling function to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the speed evaluation dimension.

[0266] Among them, “obtaining a third chi-square statistic for the target speed difference, and based on the third chi-square statistic, obtaining second intermediate data for characterizing the similarity between the current frame data and the historical frame data under the speed evaluation dimension” can be characterized as:

[0267] ID2=

[0268] 1-ChiSquaredCdf1TableFun[Target_diff(Speed) 2 / dev(Target_diff(Speed)) 2 ]

[0269] Among them, ID2 is the second intermediate data used to characterize the similarity between the current frame data and the historical frame data under the speed evaluation dimension; Target_diff(Speed) is used to characterize the target speed difference; dev(Target_diff(Speed)) is used to characterize the standard deviation of the target speed difference; ChiSquaredCdf1TableF un[Target_diff(Speed) 2 / dev(Target_diff(Speed)) 2 ] is used to characterize the final speed difference (ie, Target_diff(Speed) 2 / dev(Target_diff(Speed)) 2 ), and a third chi-square statistic for the final speed is obtained by searching from a chi-square distribution table with a degree of freedom of 1. For example, the third chi-square statistic for the final speed can be obtained by searching from a chi-square distribution table with a degree of freedom of 1 based on the final speed difference through interpolation to improve the search efficiency of the third chi-square statistic.

[0270] The standard deviation of the above target speed difference can be obtained by calculation. For example, after obtaining the probability distribution of the target speed difference, the standard deviation of the target speed difference can be calculated based on the probability distribution of the target speed difference, which will not be described in detail here.

[0271] Thereafter, when the second intermediate data is less than or equal to the second similarity threshold, the second intermediate data can be determined as reference data for characterizing the similarity between the current frame data and the historical frame data under the speed evaluation dimension. This process can be characterized as:

[0272] Rd6=ID2

[0273] Among them, Rd6 is used to represent the reference data used to represent the similarity between the current frame data and the historical frame data under the speed evaluation dimension; ID2 is used to represent the second intermediate data.

[0274] When the second intermediate data is greater than the second similarity threshold, the second intermediate data can be compressed by a scaling function to obtain reference data for characterizing the similarity between the current frame data and the historical frame data in the speed evaluation dimension. This process can be characterized as:

[0275] Rd6=[(ID2-Pth2) / (1-Pth2)](Pmax2-Pth2)+Pth2

[0276] Among them, Rd6 is used to characterize the reference data used to characterize the similarity between the current frame data and the historical frame data under the speed evaluation dimension; ID2 is used to characterize the second intermediate data; Pth2 is used to characterize the second zoom threshold; Pmax2 is used to characterize the second maximum zoom probability value. Here, the second zoom threshold and the second maximum zoom probability value can be set according to actual application requirements. For example, the second zoom threshold can be set to 0.5, and the second maximum zoom probability value can be set to 0.7. The embodiments of the present disclosure are not limited to this.

[0277] (3) Mapping the first driving speed and the second driving speed so that, when the first driving speed and the second driving speed do not conform to the first preset relationship and the second preset relationship, a third similarity characterization value is determined as reference data for characterizing the similarity between the current frame data and the historical frame data under the speed evaluation dimension.

[0278] The third similarity characterization value is greater than the second similarity characterization value. Here, the third similarity characterization value can be set according to actual application requirements, for example, the third similarity characterization value can be set to 0.5, which is not limited in the embodiment of the present disclosure.

[0279] In a specific example, “when the first driving speed and the second driving speed do not meet the first preset relationship and the second preset relationship, determining the third similarity characterization value as reference data for characterizing the similarity between the current frame data and the historical frame data under the speed evaluation dimension” can be characterized as:

[0280] Rd7=SCV3

[0281] Among them, Rd7 is used to represent the reference data used to represent the similarity between the current frame data and the historical frame data under the speed evaluation dimension; SCV3 is used to represent the third similarity representation value.

[0282] In addition, in a practical example, the probability distribution of the initial speed difference between the first driving speed and the second driving speed is as follows: Fig.15 In the case shown in FIG. 1 , the reference data used to characterize the similarity between the current frame data and the historical frame data under the speed evaluation dimension can be as follows: Fig.16 Shown Fig.16 In , the horizontal axis is used to represent the reference data used to represent the similarity between the current frame data and the historical frame data under the speed evaluation dimension, and the vertical axis is used to represent the probability density).

[0283] Through the above method, in the embodiment of the present disclosure, the data collected by the radar in the current frame data and the historical frame data collected by the second target sensor can be determined as the third perception data, and the data collected by the camera in the current frame data and the historical frame data can be determined as the fourth perception data, and the first driving speed of the ninth obstacle represented by the third perception data is obtained, and the second driving speed of the tenth obstacle represented by the fourth perception data is obtained, and then the first driving speed and the second driving speed are mapped to obtain the reference data for characterizing the similarity between the current frame data and the historical frame data under the speed evaluation dimension. This process does not involve a complex data processing flow, so the acquisition efficiency of the reference data for characterizing the similarity between the current frame data and the historical frame data under the speed evaluation dimension can be improved, thereby improving the execution efficiency of the data verification method. Moreover, when the first driving speed and the second driving speed are mapped to obtain the reference data for characterizing the similarity between the current frame data and the historical frame data under the speed evaluation dimension, the initial speed difference between the first driving speed and the second driving speed is also standardized, so that the accuracy of the reference data for characterizing the similarity between the current frame data and the historical frame data under the speed evaluation dimension can be improved, thereby further improving the accuracy of sensor data association.

[0284] In an optional implementation, the first target sensor and the second target sensor are any combination of a radar and a camera, and the target evaluation dimension is the first projection position evaluation dimension, and “mapping the current frame data and the historical frame data collected by the second target sensor to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the target evaluation dimension” may include:

[0285] Mapping the current frame data and the historical frame data collected by the second target sensor, obtaining a fourth initial distance in the first direction (for example, the X-axis direction of the world coordinate system) between the seventh obstacle represented by the current frame data and the eighth obstacle represented by the historical frame data collected by the second target sensor as the first unidirectional distance;

[0286] Based on the first one-way distance, reference data for characterizing the similarity between the current frame data and the historical frame data under the first projection position evaluation dimension is obtained.

[0287] The mapping of the current frame data and the historical frame data collected by the second target sensor may be performed by comparing the current frame data and the historical frame data after spatial mapping. For example, the current frame data and the historical frame data may be mapped to the same world coordinate system and then compared.

[0288] Further, in an example, “mapping the current frame data and the historical frame data collected by the second target sensor to obtain a fourth initial distance in the first direction between the seventh obstacle represented by the current frame data and the eighth obstacle represented by the historical frame data collected by the second target sensor” may include:

[0289] Determine a third target center point of a seventh obstacle represented by the current frame data;

[0290] Determine a fourth target center point of an eighth obstacle represented by the historical frame data collected by the second target sensor;

[0291] The third target center point and the fourth target center point are mapped to obtain a fourth initial distance between the seventh obstacle and the eighth obstacle in the first direction.

[0292] In a specific example, at least one of a target detection algorithm, a Kalman filter algorithm, a clustering algorithm, a fusion algorithm, etc. can be used to determine the center point of the obstacle represented by the data collected by the radar in the current frame data and the historical frame data; in another specific example, at least one of contour detection and shape analysis, a target detection algorithm, a semantic segmentation algorithm, etc. can be used to determine the center point of the obstacle represented by the data collected by the camera in the current frame data and the historical frame data.

[0293] After determining the third target center point of the seventh obstacle represented by the current frame data and determining the fourth target center point of the eighth obstacle represented by the historical frame data collected by the second target sensor, the third target center point and the fourth target center point can be mapped to obtain a fourth initial distance between the seventh obstacle and the eighth obstacle in the first direction as the first one-way distance.

[0294] In addition, it should be noted that, considering the theoretical jitter range of the fourth initial distance, in an actual application example, the probability distribution of the fourth initial distance obeys the Gaussian distribution, which can be specifically described as follows: Fig.17 As shown ( Fig.17 In the figure, the horizontal axis is used to represent the fourth initial distance (in meters), and the vertical axis is used to represent the probability density). The theoretical jitter range of the fourth initial distance can be determined according to experimental values, and the embodiment of the present disclosure does not limit this.

[0295] In one example, “obtaining reference data for characterizing the similarity between current frame data and historical frame data in a first projection position evaluation dimension based on the first one-way distance” may include:

[0296] When the first one-way distance is less than or equal to the second distance threshold, determining the fourth similarity representation value as the third intermediate data;

[0297] Alternatively, when the first one-way distance is greater than the second distance threshold, the first one-way distance is processed by the Welsh loss function to obtain third intermediate data;

[0298] Based on the third intermediate data, reference data for characterizing the similarity between the current frame data and the historical frame data under the first projection position evaluation dimension is obtained.

[0299] Among them, the second distance threshold and the fourth similarity characterization value can be set according to actual application requirements. For example, the second distance threshold can be set to 0.05m, and the fourth similarity characterization value can be set to a probability value close to 1. For example, the fourth similarity characterization value can be set to 1-(1e-6). The embodiment of the present disclosure is also not limited to this.

[0300] When the first one-way distance is greater than the second distance threshold, the first one-way distance can be processed by the Welsh loss function to obtain the third intermediate data. This process can be characterized as:

[0301] ID3=Exp[(SPd(Postion33)-Dth33) / Dsc33]

[0302] Among them, ID3 is used to characterize the third intermediate data obtained by processing the first one-way distance through the Welsh loss function; SPd (Position 33) is used to characterize the first one-way distance; Dth33 is used to characterize the second distance threshold; Dsc33 is used to characterize the second step length parameter. Here, the second step length parameter can be set according to actual application requirements. For example, the second step length parameter can be set to 0.05m, which is not limited in the embodiments of the present disclosure.

[0303] In the above example, when the first unidirectional distance is less than the second distance threshold, the fourth similarity characterization value (that is, a probability value close to 1) can be used as the third intermediate data to characterize the first projection position evaluation dimension for characterizing that the current frame data has a high similarity with the historical frame data; when the first unidirectional distance is greater than or equal to the second distance threshold, the first unidirectional distance is processed by the Welsh loss function to obtain the third intermediate data. Specifically, the first unidirectional distance can be made to decay sharply in an exponential decay manner to obtain a smaller third intermediate data to characterize the first projection position evaluation dimension for characterizing that the current frame data has a low similarity with the historical frame data.

[0304] After obtaining the third intermediate data, reference data for characterizing the similarity between the current frame data and the historical frame data under the first projection position evaluation dimension may be obtained based on the third intermediate data in the following manner:

[0305] When the third intermediate data is less than or equal to the third similarity threshold, the third intermediate data is determined as reference data for characterizing the similarity between the current frame data and the historical frame data under the first projection position evaluation dimension;

[0306] Alternatively, when the third intermediate data is greater than the third similarity threshold, the third intermediate data is compressed by a scaling function to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the first projection position evaluation dimension.

[0307] The third similarity threshold may be set according to actual application requirements. For example, the third distance threshold may be set to 0.5, which is not limited in the embodiments of the present disclosure.

[0308] Further, in a specific example, “when the third intermediate data is less than or equal to the third similarity threshold, determining the third intermediate data as reference data for characterizing the similarity between the current frame data and the historical frame data under the first projection position evaluation dimension” can be characterized as:

[0309] Rd7=ID3

[0310] Among them, Rd7 is used to represent the reference data used to represent the similarity between the current frame data and the historical frame data under the first projection position evaluation dimension; ID3 is used to represent the third intermediate data.

[0311] “When the third intermediate data is greater than the third similarity threshold, the third intermediate data is compressed by a scaling function to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the first projection position evaluation dimension” can be characterized as:

[0312] Rd7=[(ID3-Pth3) / (1-Pth3)](Pmax3-Pth3)+Pth3

[0313] Among them, Rd7 is used to characterize the reference data used to characterize the similarity between the current frame data and the historical frame data under the first projection position evaluation dimension; ID3 is used to characterize the third intermediate data; Pth3 is used to characterize the third zoom threshold; Pmax3 is used to characterize the third maximum zoom probability value. Here, the third zoom threshold and the third maximum zoom probability value can be set according to actual application requirements. For example, the third zoom threshold can be set to 0.5, and the third maximum zoom probability value can be set to 0.7. The embodiments of the present disclosure are not limited to this.

[0314] In addition, in a practical example, the probability distribution of the fourth initial distance is as follows Fig.17In the case shown in FIG. 1 , the probability distribution of the reference data used to characterize the similarity between the current frame data and the historical frame data under the first projection position evaluation dimension can be as follows: Fig.18 As shown ( Fig.18 In the figure, the horizontal axis is used to represent the reference data used to represent the similarity between the current frame data and the historical frame data under the first projection position evaluation dimension, and the vertical axis is used to represent the probability density).

[0315] Through the above method, in the embodiment of the present disclosure, the current frame data and the historical frame data collected by the second target sensor can be mapped to obtain the fourth initial distance in the first direction between the seventh obstacle represented by the current frame data and the eighth obstacle represented by the historical frame data collected by the second target sensor as the first unidirectional distance, and based on the first unidirectional distance, obtain the reference data for characterizing the similarity between the current frame data and the historical frame data under the first projection position evaluation dimension. This process does not involve a complex data processing flow, and therefore, can improve the efficiency of obtaining the reference data for characterizing the similarity between the current frame data and the historical frame data under the first projection position evaluation dimension, thereby improving the execution efficiency of the data verification method.

[0316] In addition, through the above method, in the embodiment of the present disclosure, when obtaining the reference data for characterizing the similarity between the current frame data and the historical frame data under the first projection position evaluation dimension based on the first unidirectional distance, the Welsh loss function and the scaling function are successively used to optimize the relevant data (for example, the first unidirectional distance and the third intermediate data), and finally, the reference data for characterizing the similarity between the current frame data and the historical frame data under the first projection position evaluation dimension is obtained. Therefore, the robustness and rationality of the reference data for characterizing the similarity between the current frame data and the historical frame data under the first projection position evaluation dimension can be improved.

[0317] In an optional implementation, the first target sensor and the second target sensor are any combination of a radar and a camera, and the target evaluation dimension is a second projection position evaluation dimension, and “mapping the current frame data and the historical frame data collected by the second target sensor to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the target evaluation dimension” may include:

[0318] Mapping the current frame data and the historical frame data collected by the second target sensor to obtain a fifth initial distance in the second direction (for example, the Y-axis direction of the world coordinate system) between the seventh obstacle represented by the current frame data and the eighth obstacle represented by the historical frame data collected by the second target sensor;

[0319] The fifth initial distance is normalized to obtain a second one-way distance;

[0320] Based on the second one-way distance, reference data for characterizing the similarity between the current frame data and the historical frame data under the second projection position evaluation dimension is obtained.

[0321] The mapping of the current frame data and the historical frame data collected by the second target sensor may be performed by comparing the current frame data and the historical frame data after spatial mapping. For example, the current frame data and the historical frame data may be mapped to the same world coordinate system and then compared.

[0322] Further, in an example, “mapping the current frame data and the historical frame data collected by the second target sensor to obtain a fifth initial distance in the second direction between the seventh obstacle represented by the current frame data and the eighth obstacle represented by the historical frame data collected by the second target sensor” may include:

[0323] Determine a third target center point of a seventh obstacle represented by the current frame data;

[0324] Determine a fourth target center point of an eighth obstacle represented by the historical frame data collected by the second target sensor;

[0325] The third target center point and the fourth target center point are mapped to obtain a fifth initial distance between the seventh obstacle and the eighth obstacle in the second direction.

[0326] In a specific example, at least one of a target detection algorithm, a Kalman filter algorithm, a clustering algorithm, a fusion algorithm, etc. can be used to determine the center point of the obstacle represented by the data collected by the radar in the current frame data and the historical frame data; in another specific example, at least one of contour detection and shape analysis, a target detection algorithm, a semantic segmentation algorithm, etc. can be used to determine the center point of the obstacle represented by the data collected by the camera in the current frame data and the historical frame data.

[0327] After determining the third target center point of the seventh obstacle represented by the current frame data and determining the fourth target center point of the eighth obstacle represented by the historical frame data collected by the second target sensor, the third target center point and the fourth target center point can be mapped to obtain a fifth initial distance between the seventh obstacle and the eighth obstacle in the second direction.

[0328] In addition, it should be noted that, considering the theoretical jitter range of the fifth initial distance, in an actual application example, the probability distribution of the fifth initial distance can be as follows: Fig.19 As shown ( Fig.19 In the figure, the horizontal axis is used to represent the fifth initial distance (in meters), and the vertical axis is used to represent the probability density). The theoretical jitter range of the fifth initial distance can be determined according to experimental values, and the embodiment of the present disclosure does not limit this.

[0329] In one example, “normalizing the fifth initial distance to obtain a second normalized one-way distance” may include:

[0330] The fifth initial distance is standardized according to the conversion method from Gaussian distribution to chi-square distribution to obtain the second one-way distance.

[0331] In a specific example, the above process can be characterized as follows:

[0332] D_diff(Position)=[D_diff(yd) 2 / dev(D_diff(yd)) 2 ]

[0333] Among them, D_diff(Position) is used to characterize the second one-way distance; D_diff(yd) is used to characterize the fifth initial distance; dev(D_diff(yd)) is used to characterize the standard deviation of the fifth initial distance.

[0334] Here, the standard deviation of the fifth initial distance can be obtained by calculation. For example, after the probability distribution of the fifth initial distance is obtained, the standard deviation of the fifth initial distance can be calculated based on the probability distribution of the fifth initial distance, which is not elaborated here.

[0335] Further, in one example, after obtaining the second one-way distance, reference data for characterizing the similarity between the current frame data and the historical frame data under the second projection position evaluation dimension can be obtained based on the second one-way distance in the following manner:

[0336] Get the fourth chi-square statistic for the second one-way distance;

[0337] Based on the fourth chi-square statistic, reference data for characterizing the similarity between the current frame data and the historical frame data under the second projection position evaluation dimension is obtained.

[0338] In a specific example, the above process can be characterized as follows:

[0339] Rd8=1-ChiSquaredCdf1TableFun(D_diff(Postion))

[0340] Among them, Rd8 is used to characterize the reference data used to characterize the similarity between the current frame data and the historical frame data under the second projection position evaluation dimension; ChiSquaredCdf1TableFun(D_diff(Po stion)) is used to characterize the fourth chi-square statistic for the second one-way distance obtained by searching from the chi-square distribution table with a degree of freedom of 1 based on the second one-way distance (that is, D_diff(Po stion)). In one example, the fourth chi-square statistic for the second one-way distance can be obtained by searching from the chi-square distribution table with a degree of freedom of 1 based on the second one-way distance through interpolation to improve the search efficiency of the fourth chi-square statistic.

[0341] In addition, in a practical example, the probability distribution of the fifth initial distance is as follows: Fig.19 In the case shown in FIG. 1 , the probability distribution of the reference data used to characterize the similarity between the current frame data and the historical frame data under the second projection position evaluation dimension can be as follows: Fig. 20 As shown ( Fig. 20 In the figure, the horizontal axis is used to represent the reference data used to represent the similarity between the current frame data and the historical frame data under the second projection position evaluation dimension, and the vertical axis is used to represent the probability density).

[0342] Through the above method, in the embodiment of the present disclosure, the current frame data and the historical frame data collected by the second target sensor can be mapped to obtain the fifth initial distance in the second direction between the seventh obstacle represented by the current frame data and the eighth obstacle represented by the historical frame data collected by the second target sensor, and the fifth initial distance is standardized to obtain the second unidirectional distance, and then based on the second unidirectional distance, the reference data used to characterize the similarity between the current frame data and the historical frame data under the second projection position evaluation dimension is obtained. This process does not involve a complex data processing flow, and therefore, the efficiency of obtaining the reference data used to characterize the similarity between the current frame data and the historical frame data under the first projection position evaluation dimension can be improved, thereby improving the execution efficiency of the data verification method.

[0343] In some optional implementations, step S103, that is, "merging multiple reference data to obtain a verification result for the current frame data" may include:

[0344] Transforming each reference data among the multiple reference data to obtain multiple logarithmic probabilities corresponding to the multiple reference data one by one;

[0345] Calculate the sum of multiple log-probabilities as the overall log-probability;

[0346] The overall log probability is converted into a probability parameter as a verification result for the current frame data.

[0347] In one example, “converting each reference data in a plurality of reference data into a logarithmic probability to obtain a plurality of logarithmic probabilities” can be characterized as:

[0348] Log oddsi=log(Rdi / (1-Rdi) wherein Log oddsi is used to characterize the logarithmic probability obtained by converting the i-th reference data among the multiple reference data; and Rdi is used to characterize the i-th reference data among the multiple reference data.

[0349] In one example, after obtaining multiple logarithmic probabilities corresponding to multiple reference data, the sum of the multiple logarithmic probabilities can be calculated as the overall logarithmic probability and recorded as Log odds, and the overall logarithmic probability is converted into a probability parameter as a verification result for the current frame data. Among them, "converting the logarithmic probability value into a probability parameter as a verification result for the current frame data" can be characterized as:

[0350] RD=e -Log odds / (e -Log odds +1)

[0351] Among them, RD is used to represent the verification result for the current frame data; Log odds is used to represent the overall logarithmic probability.

[0352] Through the above method, in the embodiment of the present disclosure, each of the multiple reference data can be converted to obtain multiple logarithmic probabilities corresponding to the multiple reference data, and the sum of the multiple logarithmic probabilities is calculated as the overall logarithmic probability, and then the logarithmic probability value is converted into a probability parameter as the verification result for the current frame data. In this way, not only can the verification result for the current frame data be limited to a probability value within the interval [0, 1] to meet the actual application requirements, but also the complexity of the calculation can be reduced, thereby improving the execution efficiency of the data verification method.

[0353] The following describes step S103 in conjunction with a specific example.

[0354] In this example, the multiple sensors include lidar, camera and radar at the same time, the first target sensor is the lidar, and each sensor except the first target sensor in the multiple sensors is used as the second target sensor, that is, the camera and radar are used as the second target sensor respectively.

[0355] Assuming that the camera is used as the second target sensor, we can get:

[0356] (1) Reference data used to characterize the similarity between the current frame data and the historical frame data under the shape evaluation dimension, denoted as Rd1;

[0357] (2) Reference data used to characterize the similarity between the current frame data and the historical frame data in the location evaluation dimension, denoted as Rd2.

[0358] Assuming that the radar is used as the second target sensor, we can get:

[0359] (1) Reference data used to characterize the similarity between the current frame data and the historical frame data under the plane distance evaluation dimension, denoted as Rd3;

[0360] (2) Reference data used to characterize the similarity between the current frame data and the historical frame data under the first three-dimensional distance evaluation dimension, denoted as Rd4.

[0361] Thereafter, the reference data Rd1 is transformed to obtain the logarithmic probability Log odds1; the reference data Rd2 is transformed to obtain the logarithmic probability Log odds2; the reference data Rd3 is transformed to obtain the logarithmic probability Logodds3; and the reference data Rd4 is transformed to obtain the logarithmic probability Log odds4.

[0362] Next, calculate the sum of multiple log odds as the overall log odds: Log odds = Log

[0363] odds1+Log odds2+Log odds3+Log odds4

[0364] Finally, the overall log probability is converted into a probability parameter as the verification result for the current frame data:

[0365] RD=e -Log odds / (e -Log odds +1)

[0366] After obtaining the verification result for the current frame data, when the verification result is greater than the overall similarity threshold, a first verification result for the current frame data can be obtained to characterize that the current frame data can be associated with the historical frame data; when the verification result is less than or equal to the overall similarity threshold, a second verification result for the current frame data can be obtained to characterize that the current frame data cannot be associated with the historical frame data.

[0367] See also Fig.21 , which is a scenario diagram of a data verification method provided in an embodiment of the present disclosure.

[0368] As mentioned above, the data verification method provided by the embodiment of the present disclosure is applied to an electronic device, wherein the electronic device may be a server, a workbench, a large computer, a conventional computer or other similar computing devices.

[0369] Electronic devices are used for:

[0370] Acquire current frame data collected by a first target sensor among multiple sensors; wherein the multiple sensors are installed on the target object;

[0371] Each sensor except the first target sensor among the multiple sensors is used as a second target sensor to obtain multiple reference data for representing the similarity between the current frame data and the historical frame data collected by the second target sensor; wherein the historical frame data is derived from the global obstacle tracking information of the target object;

[0372] Multiple reference data are fused to obtain a verification result for the current frame data; wherein the verification result is used to characterize whether the current frame data can be associated with the historical frame data.

[0373] It should be noted that in the embodiments of the present disclosure, Figure 4 The scene diagram shown is only illustrative and not restrictive. Those skilled in the art can Figure 4 Various obvious changes and / or substitutions are made to the examples, and the obtained technical solutions still fall within the scope of the disclosure of the embodiments of the present disclosure.

[0374] In order to better implement the data verification method, the embodiment of the present disclosure also provides a data verification device 2200, which can be integrated in an electronic device. The electronic device can be a server, a workbench, a large computer, a conventional computer or other similar computing devices. Fig. 22 The schematic structural block diagram shown illustrates a data verification device 2200 provided in the disclosed embodiment.

[0375] The data verification device 2200 comprises:

[0376] The first data acquisition unit 2201 is used to acquire current frame data collected by a first target sensor among multiple sensors; wherein the multiple sensors are installed on the target object;

[0377] The second data acquisition unit 2202 is used to use each sensor of the multiple sensors except the first target sensor as a second target sensor to obtain multiple reference data for representing the similarity between the current frame data and the historical frame data collected by the second target sensor; wherein the historical frame data is derived from the global obstacle tracking information of the target object;

[0378] The data fusion unit 2203 is used to fuse multiple reference data to obtain a verification result for the current frame data; wherein the verification result is used to indicate whether the current frame data can be associated with the historical frame data.

[0379] In some optional implementations, the second data acquiring unit 2202 is used to:

[0380] determining a plurality of evaluation dimensions based on the first target sensor and the second target sensor;

[0381] Each evaluation dimension in the multiple evaluation dimensions is used as a target evaluation dimension to obtain reference data for characterizing the similarity between the current frame data and the historical frame data acquired by the second target sensor under the target evaluation dimension.

[0382] In some optional implementations, the second data acquisition unit 2202 is used for one of the following:

[0383] In the case where the first target sensor and the second target sensor are any combination of a laser radar and a camera, determining a plurality of evaluation dimensions including a shape evaluation dimension and a position evaluation dimension;

[0384] In the case where the first target sensor and the second target sensor are any combination of a laser radar and a radar, determining the plurality of evaluation dimensions includes a planar distance evaluation dimension and a first three-dimensional distance evaluation dimension;

[0385] In the case where the first target sensor and the second target sensor are any combination of radar and camera, the multiple evaluation dimensions are determined to include a second three-dimensional distance evaluation dimension, a speed evaluation dimension, a first projection position evaluation dimension and a second projection position evaluation dimension.

[0386] In some optional implementations, the second data acquiring unit 2202 is used to:

[0387] Determine the evaluation strategy corresponding to the target evaluation dimension;

[0388] According to the evaluation strategy corresponding to the target evaluation dimension, the current frame data and the historical frame data collected by the second target sensor are mapped to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the target evaluation dimension.

[0389] In some optional implementations, the first target sensor and the second target sensor are any combination of a laser radar and a camera; the target evaluation dimension is a shape evaluation dimension. Based on this, the second data acquisition unit 2202 is used to:

[0390] Mapping the current frame data and the historical frame data collected by the second target sensor to obtain a shape difference between a first obstacle represented by the current frame data and a second obstacle represented by the historical frame data;

[0391] Based on the shape difference, reference data for characterizing the similarity between the current frame data and the historical frame data under the shape evaluation dimension is obtained.

[0392] In some optional implementations, the second data acquiring unit 2202 is used to:

[0393] Mapping the current frame data and the historical frame data collected by the second target sensor to obtain a width difference between the first bounding box and the second bounding box, and a height difference between the first bounding box and the second bounding box; wherein the first bounding box is a bounding box of a first obstacle represented by the current frame data; and the second bounding box is a bounding box of a second obstacle represented by the historical frame data;

[0394] According to the conversion method from Gaussian distribution to chi-square distribution, the width difference and the height difference are standardized to obtain the shape difference between the first obstacle represented by the current frame data and the second obstacle represented by the historical frame data.

[0395] In some optional implementations, the second data acquiring unit 2202 is used to:

[0396] Get the first chi-square statistic for shape differences;

[0397] Based on the first chi-square statistic, reference data for characterizing the similarity between the current frame data and the historical frame data in the shape evaluation dimension is obtained.

[0398] In some optional implementations, the first target sensor and the second target sensor are any combination of a laser radar and a camera; the target evaluation dimension is a position evaluation dimension. Based on this, the second data acquisition unit 2202 is used to:

[0399] Determine the data collected by the laser radar in the current frame data and the historical frame data collected by the second target sensor as the first perception data, and determine the data collected by the camera in the current frame data and the historical frame data as the second perception data;

[0400] Mapping the first perception data and the second perception data to obtain a first plane distance between a third obstacle represented by the first perception data and a fourth obstacle represented by the second perception data in a two-dimensional plane;

[0401] Based on the first plane distance, reference data for characterizing the similarity between the current frame data and the historical frame data in the position evaluation dimension is obtained.

[0402] In some optional implementations, the second data acquiring unit 2202 is used to:

[0403] Mapping the first perception data and the second perception data to determine a plurality of feature points in the first perception data that are outside a third bounding box; wherein the third bounding box is a bounding box of a fourth obstacle represented by the second perception data;

[0404] Acquire a first overflow amount of the plurality of feature points in the first direction relative to the third bounding box, and a second overflow amount of the plurality of feature points in the second direction relative to the third bounding box;

[0405] The first overflow amount and the second overflow amount are standardized according to a conversion method from a Gaussian distribution to a chi-square distribution, so as to obtain a first plane distance between the third obstacle and the fourth obstacle in a two-dimensional plane represented by the first perception data.

[0406] In some optional implementations, the second data acquiring unit 2202 is used to:

[0407] Obtain a second chi-square statistic for the first plane distance;

[0408] Based on the second chi-square statistic, reference data for characterizing the similarity between the current frame data and the historical frame data in the position evaluation dimension is obtained.

[0409] In some optional implementations, the first target sensor and the second target sensor are any combination of a laser radar and a radar; the target evaluation dimension is a plane distance evaluation dimension. Based on this, the second data acquisition unit 2202 is used to:

[0410] Mapping the current frame data and the historical frame data collected by the second target sensor to obtain a first initial distance between a fifth obstacle represented by the current frame data and a sixth obstacle represented by the historical frame data in a two-dimensional plane;

[0411] The first initial distance is normalized to obtain a second plane distance;

[0412] Based on the second plane distance, reference data for characterizing the similarity between the current frame data and the historical frame data under the plane distance evaluation dimension is obtained.

[0413] In some optional implementations, the second data acquiring unit 2202 is used to:

[0414] Determine a first target vertex on a fourth bounding box that is closest to the reference position; wherein the fourth bounding box is a bounding box of a fifth obstacle represented by the current frame data;

[0415] Determine a second target vertex on a fifth bounding box that is closest to the reference position; wherein the fifth bounding box is a bounding box of a sixth obstacle represented by historical frame data collected by the second target sensor;

[0416] The first target vertex and the second target vertex are mapped to obtain a spacing distance between the first target vertex and the second target vertex, which is used as a first initial distance between the fifth obstacle and the sixth obstacle in the two-dimensional plane.

[0417] In some optional implementations, the first target sensor and the second target sensor are any combination of a laser radar and a radar; the target evaluation dimension is a first three-dimensional distance evaluation dimension. Based on this, the second data acquisition unit 2202 is used to:

[0418] Mapping the current frame data and the historical frame data collected by the second target sensor to obtain a second initial distance in the three-dimensional space between a fifth obstacle represented by the current frame data and a sixth obstacle represented by the historical frame data;

[0419] The second initial distance is normalized to obtain a first three-dimensional distance;

[0420] Based on the first three-dimensional distance, reference data for characterizing the similarity between the current frame data and the historical frame data under the first three-dimensional distance evaluation dimension is obtained.

[0421] In some optional implementations, the second data acquiring unit 2202 is used to:

[0422] Determine a first target center point of a fifth obstacle represented by the current frame data;

[0423] Determine a second target center point of a sixth obstacle represented by historical frame data collected by a second target sensor;

[0424] The first target center point and the second target center point are mapped to obtain a spacing distance between the first target center point and the second target center point as a second initial distance between the fifth obstacle and the sixth obstacle in the three-dimensional space.

[0425] In some optional implementations, the first target sensor and the second target sensor are any combination of a radar and a camera; the target evaluation dimension is a second three-dimensional distance evaluation dimension. Based on this, the second data acquisition unit 2202 is used to:

[0426] Mapping the current frame data and the historical frame data collected by the second target sensor to obtain a third initial distance in the three-dimensional space between a seventh obstacle represented by the current frame data and an eighth obstacle represented by the historical frame data;

[0427] The third initial distance is normalized to obtain a second three-dimensional distance;

[0428] Based on the second three-dimensional distance, reference data for characterizing the similarity between the current frame data and the historical frame data under the second three-dimensional distance evaluation dimension is obtained.

[0429] In some optional implementations, the second data acquiring unit 2202 is used to:

[0430] Determine a third target center point of a seventh obstacle represented by the current frame data;

[0431] Determine a fourth target center point of an eighth obstacle represented by the historical frame data collected by the second target sensor;

[0432] The third target center point and the fourth target center point are mapped to obtain a spacing distance between the third target center point and the fourth target center point as a third initial distance between the seventh obstacle and the eighth obstacle in the three-dimensional space.

[0433] In some optional implementations, the second data acquiring unit 2202 is used to:

[0434] When the second three-dimensional distance is less than the first distance threshold, determining the first similarity representation value as the first intermediate data;

[0435] Alternatively, when the second three-dimensional distance is greater than or equal to the first distance threshold, the second three-dimensional distance is processed by a Welsh loss function to obtain first intermediate data;

[0436] Based on the first intermediate data, reference data for characterizing the similarity between the current frame data and the historical frame data in a second three-dimensional distance evaluation dimension is obtained.

[0437] In some optional implementations, the second data acquiring unit 2202 is used to:

[0438] When the first intermediate data is less than or equal to the first similarity threshold, determining the first intermediate data as reference data for characterizing the similarity between the current frame data and the historical frame data under the second three-dimensional distance evaluation dimension;

[0439] Alternatively, when the first intermediate data is greater than the first similarity threshold, the first intermediate data is compressed by a scaling function to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under a second three-dimensional distance evaluation dimension.

[0440] In some optional implementations, the first target sensor and the second target sensor are any combination of a radar and a camera; the target evaluation dimension is a speed evaluation dimension. Based on this, the second data acquisition unit 2202 is used to:

[0441] Determine the data collected by the radar in the current frame data and the historical frame data collected by the second target sensor as the third perception data, and determine the data collected by the camera in the current frame data and the historical frame data as the fourth perception data;

[0442] Acquire a first driving speed of a ninth obstacle represented by the third perception data;

[0443] Acquire a second driving speed of a tenth obstacle represented by the fourth perception data;

[0444] The first driving speed and the second driving speed are mapped to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the speed evaluation dimension.

[0445] In some optional implementations, the second data acquisition unit 2202 is used for one of the following:

[0446] The first driving speed and the second driving speed are mapped, so that when it is determined that the first driving speed and the second driving speed meet a first preset relationship, a second similarity characterization value is determined as reference data for characterizing the similarity between the current frame data and the historical frame data under the speed evaluation dimension; wherein the first preset relationship is that the first driving speed is less than a first speed threshold, and the second driving speed is greater than a second speed threshold; and the first speed threshold is less than the second speed threshold;

[0447] The first driving speed and the second driving speed are mapped to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the speed evaluation dimension based on the initial speed difference between the first driving speed and the second driving speed when it is determined that the first driving speed and the second driving speed meet the second preset relationship; wherein the second preset relationship is that the second driving speed is greater than the first speed threshold, and at least one of the first driving speed and the second driving speed is greater than the third speed threshold; and the third speed threshold is greater than the second speed threshold;

[0448] The first driving speed and the second driving speed are mapped so that when the first driving speed and the second driving speed do not conform to the first preset relationship and the second preset relationship, a third similarity characterization value is determined as reference data for characterizing the similarity between the current frame data and the historical frame data under the speed evaluation dimension; wherein the third similarity characterization value is greater than the second similarity characterization value.

[0449] In some optional implementations, the second data acquiring unit 2202 is used to:

[0450] normalizing an initial speed difference between the first driving speed and the second driving speed to obtain a target speed difference between the first driving speed and the second driving speed;

[0451] Based on the target speed difference, reference data for characterizing the similarity between the current frame data and the historical frame data in the speed evaluation dimension is obtained.

[0452] In some optional implementations, the second data acquiring unit 2202 is used to:

[0453] Obtain the third chi-square statistic for the target speed difference;

[0454] Based on the third chi-square statistic, second intermediate data for characterizing the similarity between the current frame data and the historical frame data under the speed evaluation dimension is obtained;

[0455] When the second intermediate data is less than or equal to the second similarity threshold, the second intermediate data is determined as reference data for characterizing the similarity between the current frame data and the historical frame data under the speed evaluation dimension;

[0456] Alternatively, when the second intermediate data is greater than the second similarity threshold, the second intermediate data is compressed by a scaling function to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the speed evaluation dimension.

[0457] In some optional implementations, the first target sensor and the second target sensor are any combination of a radar and a camera; the target evaluation dimension is the first projection position evaluation dimension. Based on this, the second data acquisition unit 2202 is used to:

[0458] Mapping the current frame data and the historical frame data collected by the second target sensor to obtain a fourth initial distance in the first direction between the seventh obstacle represented by the current frame data and the eighth obstacle represented by the historical frame data collected by the second target sensor;

[0459] The fourth initial distance is normalized to obtain a first one-way distance;

[0460] Based on the first one-way distance, reference data for characterizing the similarity between the current frame data and the historical frame data under the first projection position evaluation dimension is obtained.

[0461] In some optional implementations, the second data acquiring unit 2202 is used to:

[0462] Determine a third target center point of a seventh obstacle represented by the current frame data;

[0463] Determine a fourth target center point of an eighth obstacle represented by the historical frame data collected by the second target sensor;

[0464] The third target center point and the fourth target center point are mapped to obtain a fourth initial distance between the seventh obstacle and the eighth obstacle in the first direction.

[0465] In some optional implementations, the second data acquiring unit 2202 is used to:

[0466] When the first one-way distance is less than or equal to the second distance threshold, determining the fourth similarity representation value as the third intermediate data;

[0467] Alternatively, when the first one-way distance is greater than the second distance threshold, the first one-way distance is processed by the Welsh loss function to obtain third intermediate data;

[0468] Based on the third intermediate data, reference data for characterizing the similarity between the current frame data and the historical frame data under the first projection position evaluation dimension is obtained.

[0469] In some optional implementations, the second data acquiring unit 2202 is used to:

[0470] When the third intermediate data is less than or equal to the third similarity threshold, the third intermediate data is determined as reference data for characterizing the similarity between the current frame data and the historical frame data under the first projection position evaluation dimension;

[0471] Alternatively, when the third intermediate data is greater than the third similarity threshold, the third intermediate data is compressed by a scaling function to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the first projection position evaluation dimension.

[0472] In some optional implementations, the first target sensor and the second target sensor are any combination of a radar and a camera; the target evaluation dimension is the second projection position evaluation dimension. Based on this, the second data acquisition unit 2202 is used to:

[0473] Mapping the current frame data and the historical frame data collected by the second target sensor to obtain a fifth initial distance in the second direction between the seventh obstacle represented by the current frame data and the eighth obstacle represented by the historical frame data collected by the second target sensor;

[0474] The fifth initial distance is normalized to obtain a second one-way distance;

[0475] Based on the second one-way distance, reference data for characterizing the similarity between the current frame data and the historical frame data under the second projection position evaluation dimension is obtained.

[0476] In some optional implementations, the second data acquiring unit 2202 is used to:

[0477] Determine a third target center point of a seventh obstacle represented by the current frame data;

[0478] Determine a fourth target center point of an eighth obstacle represented by the historical frame data collected by the second target sensor;

[0479] The third target center point and the fourth target center point are mapped to obtain a fifth initial distance between the seventh obstacle and the eighth obstacle in the second direction.

[0480] In some optional implementations, the second data acquiring unit 2202 is used to:

[0481] The fifth initial distance is standardized according to the conversion method from Gaussian distribution to chi-square distribution to obtain the second one-way distance.

[0482] In some optional implementations, the second data acquiring unit 2202 is used to:

[0483] Get the fourth chi-square statistic for the second one-way distance;

[0484] Based on the fourth chi-square statistic, reference data for characterizing the similarity between the current frame data and the historical frame data under the second projection position evaluation dimension is obtained.

[0485] In some optional implementations, the data fusion unit is used to:

[0486] Transforming each reference data among the multiple reference data to obtain multiple logarithmic probabilities corresponding to the multiple reference data one by one;

[0487] Calculate the sum of multiple log-probabilities as the overall log-probability;

[0488] Convert the overall logarithmic probability into a probability parameter as the verification result for the current frame data

[0489] In the embodiment of the present disclosure, the specific functions and examples of each unit in the data verification device 2200 can be found in the relevant description of the corresponding steps in the aforementioned data verification method embodiment, and will not be repeated here.

[0490] In the technical solution disclosed herein, the acquisition, storage and application of user personal information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0491] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, an autonomous driving vehicle, a readable storage medium, and a computer program product.

[0492] Fig.23 A schematic block diagram of an example electronic device 2300 that can be used to implement an embodiment of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as vehicle-mounted computing devices, laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0493] like Fig.23 As shown, the electronic device 2300 includes a computing unit 2301, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 2302 or a computer program loaded from a storage unit 2308 to a random access memory (RAM) 2303. In RAM 2303, various programs and data required for the operation of the electronic device 2300 can also be stored. The computing unit 2301, ROM 2302 and RAM 2303 are connected to each other via a bus 2304. An input / output (I / O) interface 2305 is also connected to the bus 2304.

[0494] A number of components in the electronic device 2300 are connected to the I / O interface 2305, including: an input unit 2306, such as a keyboard, a mouse, etc.; an output unit 2307, such as various types of renderers, speakers, etc.; a storage unit 2308, such as a disk, an optical disk, etc.; and a communication unit 2309, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 2309 allows the electronic device 2300 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0495] The computing unit 2301 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 2301 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSP), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 2301 performs the various methods and processes described above, for example, a data verification method. For example, in some embodiments, the data verification method may be implemented as a computer software program, which is tangibly included in a machine-readable medium, for example, a storage unit 2308. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 2300 via ROM 2302 and / or a communication unit 2309. When the computer program is loaded into RAM 2303 and executed by the computing unit 2301, one or more steps in the data verification method described above may be executed. Alternatively, in other embodiments, the computing unit 2301 may be configured as a data verification method in any other appropriate manner (eg, by means of firmware).

[0496] Various embodiments of the systems and techniques described above herein may be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0497] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine as a stand-alone software package and partially on a remote machine, or entirely on a remote machine or server.

[0498] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a RAM, a ROM, an erasable programmable read-only memory (EPROM) or a flash memory, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0499] To provide interaction with a user, the systems and techniques described herein may be implemented on a computer having: a rendering device (e.g., a cathode ray tube (CRT) renderer or a liquid crystal display (LCD)) for rendering information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices are also used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0500] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: Local Area Network (LAN), Wide Area Network (WAN), and the Internet.

[0501] The computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship between the client and the server is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server can be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0502] The disclosed embodiment also provides an autonomous driving vehicle, including an electronic device 2300.

[0503] The embodiment of the present disclosure also provides a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable a computer to execute a data verification method.

[0504] The embodiment of the present disclosure also provides a computer program product, including a computer program, which implements the data verification method when executed by a processor.

[0505] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in the present disclosure can be executed in parallel, sequentially or in a different order, as long as the desired results of the technical solutions disclosed in the present disclosure can be achieved, and this document is not limited here. In addition, in the present disclosure, relational terms such as "first", "second", "third", etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply the existence of any such actual relationship or order between these entities or operations. In addition, "multiple" in the present disclosure can be understood as at least two.

[0506] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the principles of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A data verification method, comprising: Acquire current frame data collected by a first target sensor among a plurality of sensors; wherein the plurality of sensors are installed on a target object; Using each sensor of the plurality of sensors except the first target sensor as a second target sensor to obtain a plurality of reference data for characterizing the similarity between the current frame data and the historical frame data acquired by the second target sensor; wherein the historical frame data is derived from the global obstacle tracking information of the target object; The multiple reference data are fused to obtain a verification result for the current frame data; wherein the verification result is used to indicate whether the current frame data can be associated with the historical frame data.

2. The method according to claim 1, wherein: The acquiring of a plurality of reference data for characterizing the similarity between the current frame data and the historical frame data acquired by the second target sensor includes: determining a plurality of evaluation dimensions based on the first target sensor and the second target sensor; Each evaluation dimension of the multiple evaluation dimensions is used as a target evaluation dimension to obtain reference data under the target evaluation dimension for characterizing the similarity between the current frame data and the historical frame data collected by the second target sensor.

3. The method according to claim 2, wherein: The determining of multiple evaluation dimensions based on the first target sensor and the second target sensor includes one of the following: In the case where the first target sensor and the second target sensor are any combination of a laser radar and a camera, determining that the multiple evaluation dimensions include a shape evaluation dimension and a position evaluation dimension; In the case where the first target sensor and the second target sensor are any combination of a laser radar and a radar, determining that the multiple evaluation dimensions include a planar distance evaluation dimension and a first three-dimensional distance evaluation dimension; In the case where the first target sensor and the second target sensor are any combination of radar and camera, the multiple evaluation dimensions are determined to include a second three-dimensional distance evaluation dimension, a speed evaluation dimension, a first projection position evaluation dimension and a second projection position evaluation dimension.

4. The method according to claim 2 or 3, wherein: The obtaining of reference data for characterizing the similarity between the current frame data and the historical frame data collected by the second target sensor under the target evaluation dimension includes: Determining an evaluation strategy corresponding to the target evaluation dimension; According to the evaluation strategy corresponding to the target evaluation dimension, the current frame data and the historical frame data collected by the second target sensor are mapped to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the target evaluation dimension.

5. The method according to claim 4, wherein: The first target sensor and the second target sensor are any combination of a laser radar and a camera; the target evaluation dimension is a shape evaluation dimension; the mapping of the current frame data and the historical frame data collected by the second target sensor to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the target evaluation dimension includes: Mapping the current frame data and the historical frame data collected by the second target sensor to obtain a shape difference between a first obstacle represented by the current frame data and a second obstacle represented by the historical frame data; Based on the shape difference, reference data for characterizing the similarity between the current frame data and the historical frame data under the shape evaluation dimension is obtained.

6. The method according to claim 5, wherein: The mapping of the current frame data and the historical frame data collected by the second target sensor to obtain a shape difference between a first obstacle represented by the current frame data and a second obstacle represented by the historical frame data includes: Mapping the current frame data and the historical frame data collected by the second target sensor to obtain a width difference between a first bounding box and a second bounding box, and a height difference between the first bounding box and the second bounding box; wherein the first bounding box is a bounding box of a first obstacle represented by the current frame data; and the second bounding box is a bounding box of a second obstacle represented by the historical frame data; The width difference and the height difference are standardized according to a conversion method from Gaussian distribution to chi-square distribution, so as to obtain a shape difference between a first obstacle represented by the current frame data and a second obstacle represented by the historical frame data.

7. The method according to claim 6, wherein: The obtaining, based on the shape difference, reference data for characterizing the similarity between the current frame data and the historical frame data under the shape evaluation dimension includes: obtaining a first chi-square statistic for the shape difference; Based on the first chi-square statistic, reference data for characterizing the similarity between the current frame data and the historical frame data under the shape evaluation dimension is obtained.

8. The method according to claim 4, wherein: The first target sensor and the second target sensor are any combination of a laser radar and a camera; the target evaluation dimension is a position evaluation dimension; the mapping of the current frame data and the historical frame data collected by the second target sensor to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the target evaluation dimension includes: Determine the data collected by the laser radar in the current frame data and the historical frame data collected by the second target sensor as the first perception data, and determine the data collected by the camera in the current frame data and the historical frame data as the second perception data; Mapping the first perception data and the second perception data to obtain a first plane distance between a third obstacle represented by the first perception data and a fourth obstacle represented by the second perception data in a two-dimensional plane; Based on the first plane distance, reference data for characterizing the similarity between the current frame data and the historical frame data in the position evaluation dimension is obtained.

9. The method according to claim 8, wherein: The mapping of the first perception data and the second perception data to obtain a first plane distance between a third obstacle represented by the first perception data and a fourth obstacle represented by the second perception data in a two-dimensional plane includes: Mapping the first perception data and the second perception data to determine a plurality of feature points in the first perception data that are outside a third bounding box; wherein the third bounding box is a bounding box of a fourth obstacle represented by the second perception data; Acquire a first overflow amount of the plurality of feature points relative to the third bounding box in a first direction, and a second overflow amount of the plurality of feature points relative to the third bounding box in a second direction; The first overflow amount and the second overflow amount are standardized according to a conversion method from a Gaussian distribution to a chi-square distribution to obtain a first plane distance between the third obstacle and the fourth obstacle represented by the first perception data in a two-dimensional plane.

10. The method according to claim 9, wherein: The obtaining, based on the first plane distance, reference data for characterizing the similarity between the current frame data and the historical frame data in the position evaluation dimension includes: Obtaining a second chi-square statistic for the first plane distance; Based on the second chi-square statistic, reference data for characterizing the similarity between the current frame data and the historical frame data in the position evaluation dimension is obtained.

11. The method according to claim 4, wherein: The first target sensor and the second target sensor are any combination of a laser radar and a radar; the target evaluation dimension is a plane distance evaluation dimension; the mapping of the current frame data and the historical frame data collected by the second target sensor to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the target evaluation dimension includes: Mapping the current frame data and the historical frame data collected by the second target sensor to obtain a first initial distance between a fifth obstacle represented by the current frame data and a sixth obstacle represented by the historical frame data in a two-dimensional plane; Normalizing the first initial distance to obtain a second plane distance; Based on the second plane distance, reference data for characterizing the similarity between the current frame data and the historical frame data under the plane distance evaluation dimension is obtained.

12. The method according to claim 11, wherein: The mapping of the current frame data and the historical frame data collected by the second target sensor to obtain a first initial distance between a fifth obstacle represented by the current frame data and a sixth obstacle represented by the historical frame data in a two-dimensional plane includes: Determine a first target vertex on a fourth bounding box that is closest to a reference position; wherein the fourth bounding box is a bounding box of a fifth obstacle represented by the current frame data; Determine a second target vertex on a fifth bounding box that is closest to the reference position; wherein the fifth bounding box is a bounding box of a sixth obstacle represented by the historical frame data collected by the second target sensor; The first target vertex and the second target vertex are mapped to obtain a spacing distance between the first target vertex and the second target vertex as a first initial distance between the fifth obstacle and the sixth obstacle in a two-dimensional plane.

13. The method according to claim 4, wherein: The first target sensor and the second target sensor are any combination of a laser radar and a radar; the target evaluation dimension is a first three-dimensional distance evaluation dimension; the mapping of the current frame data and the historical frame data collected by the second target sensor to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the target evaluation dimension includes: Mapping the current frame data and the historical frame data collected by the second target sensor to obtain a second initial distance in the three-dimensional space between a fifth obstacle represented by the current frame data and a sixth obstacle represented by the historical frame data; performing normalization processing on the second initial distance to obtain a first three-dimensional distance; Based on the first three-dimensional distance, reference data for characterizing the similarity between the current frame data and the historical frame data under the first three-dimensional distance evaluation dimension is obtained.

14. The method according to claim 13, wherein: The mapping of the current frame data and the historical frame data collected by the second target sensor to obtain a second initial distance in the three-dimensional space between a fifth obstacle represented by the current frame data and a sixth obstacle represented by the historical frame data includes: Determining a first target center point of a fifth obstacle represented by the current frame data; Determine a second target center point of a sixth obstacle represented by the historical frame data collected by the second target sensor; The first target center point and the second target center point are mapped to obtain a spacing distance between the first target center point and the second target center point as a second initial distance between the fifth obstacle and the sixth obstacle in the three-dimensional space.

15. The method according to claim 4, wherein: The first target sensor and the second target sensor are any combination of a radar and a camera; the target evaluation dimension is a second three-dimensional distance evaluation dimension; the mapping of the current frame data and the historical frame data collected by the second target sensor to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the target evaluation dimension includes: Mapping the current frame data and the historical frame data collected by the second target sensor to obtain a third initial distance in the three-dimensional space between a seventh obstacle represented by the current frame data and an eighth obstacle represented by the historical frame data; Normalizing the third initial distance to obtain a second three-dimensional distance; Based on the second three-dimensional distance, reference data for characterizing the similarity between the current frame data and the historical frame data under the second three-dimensional distance evaluation dimension is obtained.

16. The method according to claim 15, wherein: The mapping of the current frame data and the historical frame data collected by the second target sensor to obtain a third initial distance in the three-dimensional space between the seventh obstacle represented by the current frame data and the eighth obstacle represented by the historical frame data includes: Determining a third target center point of a seventh obstacle represented by the current frame data; Determining a fourth target center point of an eighth obstacle represented by the historical frame data collected by the second target sensor; The third target center point and the fourth target center point are mapped to obtain a spacing distance between the third target center point and the fourth target center point as a third initial distance between the seventh obstacle and the eighth obstacle in the three-dimensional space.

17. The method according to claim 15, wherein: The obtaining, based on the second three-dimensional distance, reference data for characterizing the similarity between the current frame data and the historical frame data in the second three-dimensional distance evaluation dimension includes: When the second three-dimensional distance is less than the first distance threshold, determining the first similarity representation value as the first intermediate data; Alternatively, when the second three-dimensional distance is greater than or equal to the first distance threshold, the second three-dimensional distance is processed by a Welsh loss function to obtain first intermediate data; Based on the first intermediate data, reference data for characterizing the similarity between the current frame data and the historical frame data in the second three-dimensional distance evaluation dimension is obtained.

18. The method according to claim 17, wherein: The obtaining, based on the first intermediate data, reference data for characterizing the similarity between the current frame data and the historical frame data in the second three-dimensional distance evaluation dimension includes: In a case where the first intermediate data is less than or equal to a first similarity threshold, determining the first intermediate data as reference data for characterizing the similarity between the current frame data and the historical frame data under the second three-dimensional distance evaluation dimension; Alternatively, when the first intermediate data is greater than a first similarity threshold, the first intermediate data is compressed by a scaling function to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the second three-dimensional distance evaluation dimension.

19. The method according to claim 4, wherein: The first target sensor and the second target sensor are any combination of a radar and a camera; the target evaluation dimension is a speed evaluation dimension; the mapping of the current frame data and the historical frame data collected by the second target sensor to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the target evaluation dimension includes: Determine the data collected by the radar in the current frame data and the historical frame data collected by the second target sensor as third perception data, and determine the data collected by the camera in the current frame data and the historical frame data as fourth perception data; Acquire a first driving speed of a ninth obstacle represented by the third perception data; Acquire a second driving speed of a tenth obstacle represented by the fourth perception data; The first driving speed and the second driving speed are mapped to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the speed evaluation dimension.

20. The method according to claim 19, wherein: The mapping of the first driving speed and the second driving speed to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the speed evaluation dimension includes one of the following: The first driving speed and the second driving speed are mapped, so that when it is determined that the first driving speed and the second driving speed meet a first preset relationship, a second similarity characterization value is determined as reference data for characterizing the similarity between the current frame data and the historical frame data under the speed evaluation dimension; wherein the first preset relationship is that the first driving speed is less than a first speed threshold, and the second driving speed is greater than a second speed threshold; and the first speed threshold is less than the second speed threshold; The first driving speed and the second driving speed are mapped to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the speed evaluation dimension based on an initial speed difference between the first driving speed and the second driving speed when it is determined that the first driving speed and the second driving speed meet a second preset relationship; wherein the second preset relationship is that the second driving speed is greater than a first speed threshold, and at least one of the first driving speed and the second driving speed is greater than a third speed threshold; and the third speed threshold is greater than the second speed threshold; The first driving speed and the second driving speed are mapped so that when the first driving speed and the second driving speed do not conform to the first preset relationship and the second preset relationship, a third similarity representation value is determined as reference data for representing the similarity between the current frame data and the historical frame data under the speed evaluation dimension; wherein the third similarity representation value is greater than the second similarity representation value.

21. The method according to claim 20, wherein: The obtaining, based on the initial speed difference between the first driving speed and the second driving speed, reference data for characterizing the similarity between the current frame data and the historical frame data in the speed evaluation dimension includes: normalizing an initial speed difference between the first driving speed and the second driving speed to obtain a target speed difference between the first driving speed and the second driving speed; Based on the target speed difference, reference data for characterizing the similarity between the current frame data and the historical frame data in the speed evaluation dimension is obtained.

22. The method according to claim 21, wherein: The obtaining, based on the target speed difference, reference data for characterizing the similarity between the current frame data and the historical frame data under the speed evaluation dimension includes: obtaining a third chi-square statistic for the target speed difference; Based on the third chi-square statistic, obtaining second intermediate data for characterizing the similarity between the current frame data and the historical frame data under the speed evaluation dimension; In a case where the second intermediate data is less than or equal to a second similarity threshold, determining the second intermediate data as reference data for characterizing the similarity between the current frame data and the historical frame data under the speed evaluation dimension; Alternatively, when the second intermediate data is greater than a second similarity threshold, the second intermediate data is compressed by a scaling function to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the speed evaluation dimension.

23. The method according to claim 4, wherein: The first target sensor and the second target sensor are any combination of a radar and a camera; the target evaluation dimension is a first projection position evaluation dimension; the mapping of the current frame data and the historical frame data collected by the second target sensor to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the target evaluation dimension includes: Mapping the current frame data and the historical frame data collected by the second target sensor to obtain a fourth initial distance in the first direction between a seventh obstacle represented by the current frame data and an eighth obstacle represented by the historical frame data collected by the second target sensor as a first one-way distance; Based on the first one-way distance, reference data for characterizing the similarity between the current frame data and the historical frame data in the first projection position evaluation dimension is obtained.

24. The method according to claim 23, wherein: The mapping of the current frame data and the historical frame data collected by the second target sensor to obtain a fourth initial distance in the first direction between the seventh obstacle represented by the current frame data and the eighth obstacle represented by the historical frame data collected by the second target sensor includes: Determining a third target center point of a seventh obstacle represented by the current frame data; Determining a fourth target center point of an eighth obstacle represented by the historical frame data collected by the second target sensor; The third target center point and the fourth target center point are mapped to obtain a fourth initial distance between the seventh obstacle and the eighth obstacle in the first direction.

25. The method according to claim 23, wherein: The obtaining, based on the first one-way distance, reference data for characterizing the similarity between the current frame data and the historical frame data in the first projection position evaluation dimension includes: When the first one-way distance is less than or equal to the second distance threshold, determining the fourth similarity characterization value as the third intermediate data; Alternatively, when the first one-way distance is greater than the second distance threshold, the first one-way distance is processed by a Welsh loss function to obtain third intermediate data; Based on the third intermediate data, reference data for characterizing the similarity between the current frame data and the historical frame data in the first projection position evaluation dimension is obtained.

26. The method according to claim 25, wherein: The obtaining, based on the third intermediate data, reference data for characterizing the similarity between the current frame data and the historical frame data under the first projection position evaluation dimension includes: In a case where the third intermediate data is less than or equal to a third similarity threshold, determining the third intermediate data as reference data for characterizing the similarity between the current frame data and the historical frame data under the first projection position evaluation dimension; Alternatively, when the third intermediate data is greater than a third similarity threshold, the third intermediate data is compressed by a scaling function to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the first projection position evaluation dimension.

27. The method according to claim 4, wherein: The first target sensor and the second target sensor are any combination of a radar and a camera; the target evaluation dimension is a second projection position evaluation dimension; the mapping of the current frame data and the historical frame data collected by the second target sensor to obtain reference data for characterizing the similarity between the current frame data and the historical frame data under the target evaluation dimension includes: Mapping the current frame data and the historical frame data collected by the second target sensor to obtain a fifth initial distance in the second direction between a seventh obstacle represented by the current frame data and an eighth obstacle represented by the historical frame data collected by the second target sensor; normalizing the fifth initial distance to obtain a second one-way distance; Based on the second one-way distance, reference data for characterizing the similarity between the current frame data and the historical frame data in the second projection position evaluation dimension is obtained.

28. The method according to claim 27, wherein: The mapping of the current frame data and the historical frame data collected by the second target sensor to obtain a fifth initial distance in the second direction between the seventh obstacle represented by the current frame data and the eighth obstacle represented by the historical frame data collected by the second target sensor includes: Determining a third target center point of a seventh obstacle represented by the current frame data; Determining a fourth target center point of an eighth obstacle represented by the historical frame data collected by the second target sensor; The third target center point and the fourth target center point are mapped to obtain a fifth initial distance between the seventh obstacle and the eighth obstacle in the second direction.

29. The method according to claim 27, wherein: The step of normalizing the fifth initial distance to obtain a second normalized one-way distance includes: The fifth initial distance is standardized according to a conversion method from Gaussian distribution to chi-square distribution to obtain the second one-way distance.

30. The method of claim 29, wherein: The obtaining, based on the second one-way distance, reference data for characterizing the similarity between the current frame data and the historical frame data in the second projection position evaluation dimension includes: Obtaining a fourth chi-square statistic for the second one-way distance; Based on the fourth chi-square statistic, reference data for characterizing the similarity between the current frame data and the historical frame data in the second projection position evaluation dimension is obtained.

31. The method of claim 1, wherein: The fusing the multiple reference data to obtain a verification result for the current frame data includes: Convert each reference data in the plurality of reference data to obtain a plurality of logarithmic probabilities corresponding one-to-one to the plurality of reference data; Calculating the sum of the multiple log-probabilities as the overall log-probability; The overall logarithmic probability is converted into a probability parameter to serve as a verification result for the current frame data.

32. A data verification device, comprising: A first data acquisition unit, configured to acquire current frame data collected by a first target sensor among a plurality of sensors; wherein the plurality of sensors are installed on a target object; a second data acquisition unit, configured to use each sensor of the plurality of sensors except the first target sensor as a second target sensor, to acquire a plurality of reference data for characterizing the similarity between the current frame data and the historical frame data acquired by the second target sensor; wherein the historical frame data is derived from the global obstacle tracking information of the target object; The data fusion unit is used to fuse the multiple reference data to obtain a verification result for the current frame data; wherein the verification result is used to indicate whether the current frame data can be associated with the historical frame data.

33. An electronic device comprising: at least one processor; a memory communicatively coupled to the at least one processor; The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 31.

34. An autonomous driving vehicle comprising the electronic device described in claim 33.

35. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1 to 31.

36. A computer program product, comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 31.

Images