Target fusion tracking method and device, electronic equipment and storage medium

Abstract

The invention relates to the technical field of intelligent sensing, and discloses a target fusion tracking method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining a target sensing result; the target sensing result comprises a first sensing result and at least one second sensing result; performing target association according to the first sensing result and the at least one second sensing result, and determining a target association result; according to the attribute perception information of each laser radar target in the first perception result and the target association result, identifying whether the point cloud data has an abnormal reflection point; and if it is identified that the abnormal reflection point exists in the point cloud data, attribute perception information of each reference target in the second perception result is used for carrying out attribute updating on a fusion target in the target association result, and an attribute tracking result of each fusion target at the target time point is obtained. According to the invention, the accuracy and effectiveness of attribute tracking can be improved.

Description

Technical Field

[0001] The present application relates to the field of intelligent perception technology, and in particular to a target fusion tracking method, device, electronic device and storage medium. Background Art

[0002] At present, multi-target tracking systems in the field of intelligent driving are generally implemented by integrating multiple sensor devices. Multiple sensor devices usually include LiDAR devices. However, in some complex acquisition environments, the properties of the target reflected in the point cloud data collected by the LiDAR device may be quite different from the actual properties, which leads to inaccurate results of LiDAR target perception and target attribute tracking based on point cloud data. Summary of the invention

[0003] In view of this, an embodiment of the present application proposes a target fusion tracking method, device, electronic device and storage medium to solve the problem of inaccurate results of target attribute tracking.

[0004] The embodiment of the present application is implemented by adopting the following technical solutions:

[0005] In a first aspect, an embodiment of the present application provides a target fusion tracking method, comprising:

[0006] Target perception is performed according to the detection data collected at the target time point to determine the target perception result; the target perception result includes a first perception result and at least one second perception result; the first perception result is obtained by performing laser radar target perception on the point cloud data in the detection data;

[0007] Performing target association according to the first perception result and the at least one second perception result to determine a target association result;

[0008] According to the attribute perception information of each laser radar target in the first perception result and the target association result, identifying whether there is an abnormal reflection point in the point cloud data;

[0009] If abnormal reflection points are identified in the point cloud data, the attribute perception information of each reference target in the at least one second perception result is used to update the attributes of the fused target in the target association result to obtain the attribute tracking result of each fused target at the target time point.

[0010] In a second aspect, an embodiment of the present application provides a target fusion tracking device, comprising:

[0011] A target perception module is used to perform target perception based on the detection data collected at the target time point and determine a target perception result; the target perception result includes a first perception result and at least one second perception result; the first perception result is obtained by performing laser radar target perception on the point cloud data in the detection data;

[0012] a target association module, configured to perform target association according to the first perception result and the at least one second perception result, and determine a target association result;

[0013] an identification module, configured to identify whether the point cloud data has an abnormal reflection point according to the attribute perception information of each laser radar target in the first perception result and the target association result;

[0014] The attribute update module is used to use the attribute perception information of each reference target in the at least one second perception result to update the attributes of the fused target in the target association result if abnormal reflection points are identified in the point cloud data, so as to obtain the attribute tracking results of each fused target at the target time point.

[0015] In a third aspect, an embodiment of the present application provides an electronic device, comprising: a processor; a memory, wherein computer instructions are stored in the memory, and when the computer instructions are executed by the processor, the above-mentioned target fusion tracking method is implemented.

[0016] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, which stores computer instructions. When the computer instructions are executed by a processor, the above-mentioned target fusion tracking method is implemented.

[0017] In a fifth aspect, an embodiment of the present application provides a computer program product, including computer instructions, which, when executed by a processor, implement the above-mentioned target fusion tracking method.

[0018] In the present application, after performing laser radar target perception on the point cloud data in the detection data to obtain a first perception result, and performing perception on the data collected by other sensing devices to obtain at least one second perception result, target association is performed based on the first perception result and the at least one second perception result to determine the target association result; and based on the attribute perception information of each laser radar target in the first perception result and the target association result, it is identified whether there are abnormal reflection points in the point cloud data. When it is identified that there are abnormal reflection points in the point cloud data, the attribute perception information of each reference target in the at least one second perception result is used to update the attributes of the fused target in the target association result, and the attribute tracking result of each fused target at the target time point is obtained, instead of using the first perception result to update the attributes of the fused target. In this way, it is possible to avoid the inaccuracy of the first perception result due to the presence of abnormal reflection points in the point cloud data, thereby affecting the attribute tracking result. Therefore, the present application can improve the accuracy and effectiveness of the attribute tracking performed.

[0019] These and other aspects of the present application will become more clearly understood in the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0021] Figure 1 It is a flowchart of a target fusion tracking method according to an embodiment of the present application.

[0022] Figure 2 It is a flowchart of step 130 shown in one embodiment of the present application.

[0023] Figure 3 This is a flowchart of determining a first perception result shown in an embodiment of the present application.

[0024] Figure 4 This is a flow chart of regional division of a collection environment shown in one embodiment of the present application.

[0025] Figure 5 is a flowchart of a target fusion tracking method according to another embodiment of the present application.

[0026] Figure 6 It is a block diagram of a target fusion tracking device according to an embodiment of the present application.

[0027] Figure 7It is a block diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0028] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application, and cannot be understood as limiting the present application.

[0029] In order to enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of the present application.

[0030] In the following description, the terms "first\second" and the like are used to distinguish similar objects and do not represent a specific ordering of the objects. It can be understood that "first\second" can be interchanged with a specific order or sequence where permitted, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein.

[0031] The "plurality" mentioned in this article refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. In the following description, it involves "some embodiments or some embodiments", which describe a subset of all possible embodiments, but it can be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments, and can be combined with each other without conflict.

[0032] Figure 1 1 is a flow chart of a target fusion tracking method according to an embodiment of the present application. The method of the present application can be executed by an electronic device, such as a server, or other terminals with processing capabilities, such as vehicles, etc. Figure 1 As shown, the method includes steps 110 to 140:

[0033] Step 110, performing target perception based on the detection data collected at the target time point, and determining the target perception result; the target perception result includes a first perception result and at least one second perception result; the first perception result is obtained by performing laser radar target perception on the point cloud data in the detection data.

[0034] In the present application, the detection data includes data collected by at least two different sensor devices. The at least two different sensor devices include a laser radar device and other sensor devices other than the laser radar device. The laser radar device can be a mechanical laser radar device, a semi-solid laser radar device, or a solid-state laser radar device.

[0035] For ease of description, the other sensing devices except the laser radar device among the at least two different sensing devices are referred to as reference sensing devices. The target identified from the sensing data collected by the reference sensing device is referred to as the reference target. Correspondingly, at least one second sensing result is a target sensing result obtained by performing reference target sensing on the sensing data collected by the reference sensing device.

[0036] Target perception refers to identifying targets and their attribute information from data (e.g., point cloud data, sensor data collected by reference sensor equipment). Therefore, the first perception result and at least one second perception result both include attribute perception information of targets identified from the corresponding data. For ease of distinction, targets identified from point cloud data collected by a LiDAR device are referred to as LiDAR targets.

[0037] Attribute perception information may include the target's identification, the target's category (i.e., semantic category, such as vehicle, pedestrian, building, etc.), the target's location information, the target's size parameters (such as horizontal width, vertical length, height), motion information (such as speed, acceleration, heading angle, etc.), etc. Among them, the horizontal width of the target refers to the width in the direction perpendicular to the target's motion, and the vertical length of the target refers to the length of the target in the direction of the target's motion.

[0038] In some embodiments, the reference sensing device includes at least one of an image acquisition device and a millimeter wave radar device. For example, the reference sensing device includes only an image acquisition device, or only a millimeter wave radar device, or both an image acquisition device and a millimeter wave radar device. For example, multiple cameras, millimeter wave radar devices, and lidar devices can be deployed on a vehicle for data acquisition, and subsequently target fusion tracking can be performed according to the method of the present application for subsequent intelligent driving decision-making.

[0039] In the case where the reference sensing device includes an image acquisition device, the detection data includes image data acquired by the image acquisition device. The image acquisition device may be a camera. Of course, one or more cameras may be provided for image acquisition. The multiple cameras may acquire images facing different actual viewing angles. Correspondingly, the image data may include view image data acquired by multiple different cameras at corresponding viewing angles. Correspondingly, at least one second perception result includes performing target perception on the image data to obtain a visual target perception result. Correspondingly, the reference target includes a target identified from the image data. For ease of distinction, the target identified from the image data acquired by the image acquisition device is referred to as a visual target.

[0040] In some embodiments, the target perception model can be used to perform target perception on the image data or point cloud data, and the corresponding target perception result can be output. For example, the first target perception model is used to perform target perception on the image data, and the second target perception model is used to perform target perception on the point cloud data. The first target perception model and the second target perception model are different. The first target perception model and the second target perception model are models constructed by one or more neural networks, such as convolutional neural networks, fully connected neural networks, etc.

[0041] In the case where the reference sensing device includes a millimeter-wave radar device, the detection data includes millimeter-wave radar data collected by the millimeter-wave radar device. In some embodiments, the millimeter-wave radar data may be a list of millimeter-wave radar targets output by the millimeter-wave radar device based on data collection and target perception, wherein the millimeter-wave radar target list includes attribute perception information of each millimeter-wave radar target identified by the millimeter-wave radar device. Correspondingly, the reference target includes a target identified from data collected by the millimeter-wave radar device. For ease of distinction, the target identified by data collected by the millimeter-wave radar device is referred to as a millimeter-wave radar target. Correspondingly, at least one second perception result includes a millimeter-wave radar target perception result. For example, the millimeter-wave radar target list output by the millimeter-wave radar device may be used as a millimeter-wave radar target perception result.

[0042] Considering that a single sensor device may not have high accuracy in target perception results in some scenarios due to its own application characteristics, for example, the camera is greatly affected by external conditions (light, shadow), and the target perception effect is not good in scenes with poor light; the laser radar is relatively sensitive to the environment, such as the noise generated by rain; and the millimeter wave radar has a strong ability to penetrate fog, smoke, and dust, but its ability to perceive targets is relatively weak. In the case where the reference sensor device includes an image acquisition device and a millimeter wave radar device, the advantages of various sensor devices can be comprehensively utilized, which is conducive to enhancing the stability and accuracy of target perception.

[0043] The target time point refers to the collection time point of the detection data to be processed at present, and the detection data collected at different collection time points can be processed according to the method of the present application. The at least two different sensing devices as above can be deployed on the same device (for example, called the target device), and the at least two different sensing devices deployed on the target device can collect data in real time. In this way, the target perception can be performed by the sensing data collected by the at least two different sensing devices deployed on the target device (for example, point cloud data, sensing data collected by the reference sensing device), etc., to accurately identify other objects around the target device. For example, the target device can be a vehicle, and the at least two different sensing devices deployed on the vehicle can collect data in real time while the vehicle is moving. The corresponding target that needs to be perceived can be obstacles around the target device that may affect the movement of the target device, such as other vehicles, pedestrians, buildings, etc.

[0044] Step 120, performing target association based on the first perception result and at least one second perception result to determine a target association result.

[0045] Target association refers to associating the lidar target in the first perception result with the reference target in at least one second perception result to determine the lidar target and the reference target representing the same target in the physical environment space. In this application, the target represented by the lidar target and the reference target in the physical environment space is called a fusion target.

[0046] In some embodiments, if the target time point is the first acquisition time point, the laser radar target in the first perception result and the reference target in at least one second perception result may be target associated to obtain a target association result. In this case, the target association result may indicate the laser radar target and the reference target representing the same target in the physical environment space, and the target association result may include assigning an identifier to each fused target.

[0047] In some embodiments, if the target time point is not the first acquisition time point, the attribute tracking results of each fused target tracked and determined at the previous time point of the target time point can also be obtained, and the fused target tracked and determined at the previous time point is associated with the lidar target in the first perception result at the target time point and the reference target in at least one second perception result. For example, the attribute information of each fusion target tracked and determined at a time point before the target time point can be used to perform attribute similarity calculation with the attribute perception information of the lidar target in the first perception result to obtain the first attribute similarity; the attribute information of each fusion target tracked and determined at a time point before the target time point can be used to perform attribute similarity calculation with the attribute perception information of the reference target in at least one second perception result to obtain the second attribute similarity; thereafter, if the attribute similarity difference between the first attribute similarity between the same fusion target (assumed to be fusion target K1) and a lidar target (assumed to be lidar target A) and the second attribute similarity between the same fusion target (assumed to be reference target B) is less than a specified threshold, and both the first attribute similarity and the second attribute similarity are greater than the first similarity threshold, the fusion target represented by the lidar target A and the reference target B can be regarded as the fusion target K1.

[0048] It can be understood that if the first perception result includes a visual target perception result and a millimeter-wave radar target perception result, the attribute information of each fused target tracked and determined at a time point before the target time point is calculated with the attribute perception information of the reference target in at least one second perception result, including: calculating the attribute similarity between the attribute information of each fused target tracked and determined at a time point before the target time point and the attribute perception information of the visual target in the visual target perception result, and calculating the attribute similarity between the attribute information of each fused target tracked and determined at a time point before the target time point and the attribute perception information of the millimeter-wave radar target in the millimeter-wave radar target perception result.

[0049] In some embodiments, attribute similarity calculation can be performed based on the attribute perception information of each laser radar target in the first perception result and the attribute perception information of each reference target in at least one second perception result to determine the attribute similarity between the laser radar target and the reference target. If the attribute similarity between a laser radar target and a reference target exceeds the second similarity threshold, it can be determined that the laser radar target and the reference target represent the same fused target in the physical environment space. The attribute similarity calculation can be to vectorize the attribute perception information and then calculate the Euclidean distance between the two vectors obtained by vectorization.

[0050] In some embodiments, similarity calculation can be performed for attribute information of the same attribute based on attribute perception information of each laser radar target in the first perception result and attribute perception information of each reference target in at least one second perception result. For example, if the attribute perception information includes position information and speed information, the position similarity between the position information of the laser radar target A (obtained from the first perception result) and the position information of the reference target B (obtained from at least one second perception result) can be calculated, that is, the sub-attribute similarity under the attribute of position; the speed similarity between the speed information of the laser radar target A and the speed information of the reference target B can be calculated, that is, the sub-attribute similarity under the attribute of speed; then, the sub-attribute similarities of the laser radar target and the reference target under multiple attributes are weighted and calculated to obtain the attribute similarity between the laser radar target and the reference target. Among them, the above similarity can be measured by Euclidean distance, Mahalanobis distance, etc. For example, the position similarity can be measured based on Euclidean distance, Mahalanobis distance, etc. In some instances, the speed similarity can be measured based on Cartesian coordinate system or polar coordinate system.

[0051] If at least one second perception result includes only visual target perception results, the target association result may include one or more first target groups, wherein a first target group includes a lidar target, a visual target, and a fused target identifier, and the lidar target and the visual target in the same first target group represent the same fused target in the physical environment space. The fused target identifier in a first target group is the identifier of the fused target represented by the lidar target and the visual target in the first target group.

[0052] Similarly, if at least one second perception result includes only millimeter-wave radar target perception results, the target association result may include one or more second target groups, one second target group includes one lidar target, one millimeter-wave radar target and one fusion target identifier, and the lidar target and millimeter-wave radar target in the same second target group represent the same fusion target in the physical environment space. The fusion target identifier in a second target group is the identifier of the fusion target represented by the lidar target and millimeter-wave radar target in the second target group.

[0053] If at least one second perception result includes a visual target perception result and a millimeter-wave radar target perception result, the target association result may include one or more third target groups, and a third target group includes a laser radar target, a visual target, a millimeter-wave radar target and a fusion target identifier, and the laser radar target, visual target and millimeter-wave radar target in the same third target group represent the same fusion target in the physical environment space. The fusion target identifier in a third target group is the identifier of the fusion target represented by the laser radar target, visual target and millimeter-wave radar target in the third target group.

[0054] Step 130, based on the attribute perception information and target association results of each lidar target in the first perception result, identify whether there are abnormal reflection points in the point cloud data.

[0055] In the scenario of water accumulation on the road (for example, water accumulation caused by rain, or residual water accumulation after the road surface is cleaned, or water accumulation caused by people splashing water on the road surface), for the current self-vehicle, if there is a target vehicle in front of the self-vehicle along the direction of movement of the self-vehicle, the target vehicle will splash water after passing through the water on the road, and the laser emitted by the lidar device on the self-vehicle cannot bypass the splashing water due to the laser wavelength, causing the splashing water to reflect the laser. This will cause the point cloud data collected by the lidar device to include a large number of reflection sampling points reflected by splashing water. This part of the reflection sampling points reflected by the liquid splashed from the road surface are the abnormal reflection points to be identified in this application.

[0056] For a target vehicle passing through water, the liquid mostly splashes to the left and right sides of the vehicle. When the point cloud data includes a large number of reflection sampling points of the liquid splashed on the road surface, subsequent lidar target perception based on the point cloud data will cause the width of the identified lidar target to increase, that is, the width of the perceived lidar target is larger than the actual width of the fused target represented by the lidar target.

[0057] Therefore, in this application, based on this principle, according to the attribute perception information of each laser radar target in the first perception result and the target association result, it is identified whether there are abnormal reflection points in the point cloud data. If there are reflection sampling points in the point cloud data that are reflected by liquid splashed from the road surface, it indicates that the point cloud data was collected when liquid splashed on the road surface.

[0058] In some embodiments, the attribute perception information of the laser radar target includes the lateral width of the laser radar target; Figure 2 As shown, step 130 includes the following steps 210 to 230:

[0059] In step 210, for each lidar target in the first perception result, a reference lateral width of the fused target represented by the lidar target is obtained according to the target association result; the reference lateral width is the lateral width of a reference target representing the same fused target as the lidar target in at least one second perception result, or is the lateral width of the fused target represented by the lidar target tracked and determined at the previous acquisition time point of the target time point.

[0060] That is to say, the lateral width of the fused target represented by the LiDAR target determined by tracking at the last acquisition time point of the target time point can be used as a reference for judging whether the lateral width of the LiDAR target identified by the point cloud data acquired at the target time point has increased. The lateral width of the fused target represented by the LiDAR target determined by tracking at the last acquisition time point of the target time point refers to the lateral width determined by target fusion tracking using the detection data acquired at the last acquisition time point of the target time point.

[0061] Since the target association result indicates the fused target represented by the lidar target, the lateral width of the fused target represented by the current lidar target determined by tracking at the last acquisition time point of the target time point can be determined based on the target association result from the lateral width of each fused target determined by tracking at the last acquisition time point of the target time point.

[0062] The lateral width of the reference target representing the same fused target as the laser radar target in at least one second perception result can also be used as a reference for judging whether the lateral width of the laser radar target identified by the point cloud data collected at the target time point has increased. Because, using two different sensing devices to respectively perceive the data collected for the same fused target, the perceived lateral width for the same fused target should theoretically have a small difference, but if the difference between the two is large, it may be because there are abnormal reflection points in the point cloud data.

[0063] Since the target association result indicates the lidar target and the reference target representing the same fused target, the lateral width of the reference target representing the same fused target as the current lidar target can be obtained from at least one second perception result based on the target association result.

[0064] Step 220, determining the width difference between the lateral width of each lidar target in the first perception result and the reference lateral width of the represented fusion target.

[0065] In some embodiments, the width ratio between the lateral width of the lidar target in the first perception result and the reference lateral width of the represented fusion target can be calculated as the width difference between the two.

[0066] In other embodiments, the width difference between the lateral width of the lidar target in the first perception result and the reference lateral width of the represented fusion target may also be calculated as the width difference between the two.

[0067] Step 230: Identify whether there are abnormal reflection points in the point cloud data according to the width difference.

[0068] In some embodiments, step 230 includes: if it is determined that the first condition is met based on the width difference, determining that there are abnormal reflection points in the point cloud data; wherein the first condition is that there is at least one first lidar target in the first perception result; the first lidar target refers to a lidar target whose width ratio exceeds a first ratio threshold, and the width ratio refers to the ratio of the lateral width in the first perception result to the corresponding reference lateral width; the first ratio threshold is greater than 1.

[0069] As described above, if the point cloud data is collected in a splashing water environment, it may cause the lateral width of the target identified using the point cloud data to increase. Based on this, if the width difference is present, it is determined that there is at least one laser radar target in the first perception result, and the ratio between its lateral width in the first perception result and the reference lateral width of the laser radar target exceeds the first ratio threshold, indicating that the perceived lateral width of the laser radar target exceeds the actual width of the laser radar target. Therefore, in this case, it can be determined that the point cloud data was collected in a splashing water environment, and correspondingly, it can be determined that the point cloud data includes more sampling points reflected by splashing water. The first ratio threshold can be determined by experiment, for example, the first ratio threshold is 1.2, 1.3, etc., which is not specifically limited here.

[0070] In other embodiments, the detection data is collected by a sensor device installed on the target device; step 230 includes: if it is determined that the first condition is met based on the width difference, and the first perception result satisfies at least one of the second condition, the third condition and the fourth condition, it is determined that there are abnormal reflection points in the point cloud data.

[0071] The first condition is that there is at least one first laser radar target in the first perception result, the first laser radar target refers to a laser radar target whose width ratio exceeds a first ratio threshold, the width ratio refers to the ratio of the lateral width in the first perception result to the corresponding reference lateral width; the first ratio threshold is greater than 1;

[0072] The second condition is that the first laser radar target in the first perception result is located in front of the target device at the target time point and is located in a non-safe interval driving area relative to the target device;

[0073] The third condition is that the speed of the first lidar target at the target time point in the first perception result exceeds the first speed threshold.

[0074] The fourth condition is: the first ratio between the first perception area and the second perception area of ​​the first lidar target at the target time point in the first perception result exceeds the second ratio threshold, or the first ratio between the first perception area and the second perception area is less than the third ratio threshold; wherein the second ratio threshold is greater than 1 and the third ratio threshold is less than 1; the second perception area refers to the perception area determined by tracking the fusion target represented by the first lidar target at the previous acquisition time point of the target time point.

[0075] For the second condition, under normal circumstances, when driving on the road, two traffic participants (such as vehicles, etc.) traveling in front and behind each other are required to have a spacing distance (i.e., longitudinal distance) greater than the safe spacing distance in the direction of movement. If the first laser radar target is located in front of the target device at the target time point and is located in the non-safe spacing driving area relative to the target device, that is, the longitudinal spacing distance between the first laser radar target and the target device is less than the safe spacing distance. If it is determined that the longitudinal distance between the two is less than the safe spacing distance based on the position information of the first laser radar object in the first perception result and the position information of the current target device (the target device may be a vehicle), the possible reason is that the point cloud data includes sampling points reflected by splashing water from the rear of the front vehicle, which causes the longitudinal distance between the identified first laser radar object and the target device to be less than the actual longitudinal spacing distance.

[0076] In some embodiments, since the safety interval distance is related to the road, the safety interval distances required for roads with different maximum allowable speeds are different. Therefore, the safety interval distance above can be the safety interval distance corresponding to the road grade of the road where the target device is located. Correspondingly, the size of the non-safety interval driving area is also related to the road grade of the road where the target device is located. The safety interval distances required for roads of different road grades may be different.

[0077] Among them, the size of the non-safe interval driving area can be set as needed. For example, if the first laser radar target is located in the longitudinal area 5 to 80m in front of the target device at the target time point, and is located in the lateral area -20m to 20m, it is regarded as the first laser radar target is located in the non-safe interval driving area of ​​the target device at the target time point.

[0078] For the third condition, vehicles are usually limited to the corresponding maximum permissible speed on the road. If the speed of the first laser radar target at the target time point is higher than the maximum permissible speed on the road as identified by the point cloud data, it is obvious that the point cloud data is abnormal. In a specific embodiment, the first speed threshold can be set according to the maximum permissible speed on the road, for example, the first speed threshold is not less than the maximum permissible speed on the corresponding road. For example, if the maximum permissible speed on a road is 60km / h, the set first speed threshold can be 20m / s (i.e. 72km / h).

[0079] The first perception area refers to the area of ​​the first lidar target determined according to the size of the first lidar target in the first perception result; the second perception area refers to the perception area determined by tracking the fusion target represented by the first lidar target at the previous acquisition time point of the target time point, that is, the area of ​​the first lidar target determined according to the size determined by tracking the fusion target represented by the first lidar target at the previous acquisition time point of the target time point.

[0080] For the fourth condition, if the first ratio between the first perception area and the second perception area of ​​the first laser radar target at the target time point in the first perception result exceeds the second ratio threshold, it indicates that the possible reason is that the size of the first laser radar target in the first perception result is larger than the actual size. If the first ratio between the first perception area and the second perception area is less than the third ratio threshold, it indicates that the possible reason is that the size of the first laser radar target in the first perception result is smaller than the actual size. Both cases indicate that the identified first laser radar target is abnormal.

[0081] Among them, the second ratio threshold and the third ratio threshold can be set according to actual needs. The second ratio threshold can be the same as the first ratio threshold mentioned above, or it can be different. It is required to be greater than 1. For example, the first ratio threshold and the second ratio threshold are both set to 1.2, and the third ratio threshold is 0.8. Of course, it is not limited to this.

[0082] In the above embodiment, the second condition, the third condition and the fourth condition are used as auxiliary recognition conditions. When the first condition is met, if the second condition, the third condition and the fourth condition are also met, the accuracy of the determined recognition result (that is, the point cloud data is the recognition result collected in a splashing environment) can be guaranteed to be higher.

[0083] Step 140, if abnormal reflection points are identified in the point cloud data, the attribute perception information of each reference target in at least one second perception result is used to update the attributes of the fused target in the target association result to obtain the attribute tracking result of each fused target at the target time point.

[0084] Updating the attributes of the fusion target refers to determining the final attribute information of each attribute of the fusion target at the target time point, such as determining the position, size, and speed light at the target time point. Correspondingly, the attribute tracking result of each fusion target at the target time point indicates the final attribute information of each attribute of the fusion target at the target time point.

[0085] If there are many abnormal reflection points in the point cloud data, for example, there are many sampling points reflected by splashing water in the point cloud data, and the sampling points reflected by splashing water are not the actual sampling points reflected by the target, therefore, if the first perception result obtained by performing target perception on the point cloud data is used to update the attributes of the fused target, it will cause the attribute tracking result of the determined fused target at the target time point (that is, the attribute information of each attribute of the fused target at the target time point) to be inaccurate. Therefore, in the present application, if it is identified that there are abnormal reflection points in the point cloud data, the first perception result is not used to update the attributes of the fused target, and the attribute perception information of each reference target in at least one second perception result is used to update the attributes of the fused target in the target association result, so as to ensure the attribute tracking result of the fused target at the target time point.

[0086] In some embodiments, the attribute perception information of each reference target in at least one second perception result may be used as the attribute information of the fusion target represented by the reference target at the target time point.

[0087] In some embodiments, if at least one second perception result includes a visual target perception result and a millimeter-wave radar target perception result, in step 140, for the same attribute of each fused target, the attribute information of the fused target for the attribute in the visual target perception result and the attribute information of the fused target for the attribute in the millimeter-wave radar target perception result can be comprehensively determined to determine the final attribute information of the fused target for the attribute at the target time point.

[0088] In the present application, after performing laser radar target perception on the point cloud data in the detection data to obtain a first perception result, and performing perception on the data collected by other sensing devices to obtain at least one second perception result, target association is performed based on the first perception result and at least one second perception result to determine the target association result; and based on the attribute perception information of each laser radar target in the first perception result and the target association result, it is identified whether the point cloud data is an abnormal reflection point. When the abnormal reflection point of the point cloud data is identified, the attribute perception information of each reference target in the at least one second perception result is used to update the attributes of the fused target in the target association result, and the attribute tracking result of each fused target at the target time point is obtained, instead of using the first perception result to update the attributes of the fused target. In this way, it is possible to avoid the inaccuracy of the first perception result due to the abnormal reflection point in the point cloud data, thereby affecting the attribute tracking result. Therefore, the present application can improve the accuracy of the attribute tracking performed.

[0089] In some embodiments, the detection data includes first sensing data collected by a reference sensing device and point cloud data collected by a laser radar device; step 110 includes a process of determining a first sensing result and at least one second sensing result, wherein the first sensing data may be subjected to reference target sensing to obtain at least one second sensing result. The process of determining the first sensing result may be as follows: Figure 3 As shown, including:

[0090] Step 310: Count the number of ground sampling points belonging to the ground in the point cloud data.

[0091] The point cloud data may be first segmented into ground point clouds to determine the ground sampling points belonging to the ground in the point cloud data, and then the number of ground sampling points belonging to the ground in the point cloud data may be correspondingly counted.

[0092] Figure 4 is a flowchart of the regional division of the collection environment shown in an embodiment of the present application, Figure 4 In the above figure, the acquisition environment is divided along the height direction, where area 1, area 6 and area 11 correspond to different height ranges. Area 1-area 5 in the height range corresponding to area 1 can be regarded as the road surface area, area 6-area 10 in the height range corresponding to area 6 can be regarded as the space area where the target on the road is located, and area 11-area 15 in the height range corresponding to area 11 can be regarded as the space area where the sky is located. It is worth mentioning that the actual acquisition environment is a three-dimensional space area. Figure 4 The altitude ranges shown in the different areas are merely illustrative examples and cannot be considered as actual altitudes. For example, generally speaking, the altitude interval defined by the altitude range belonging to the ground is smaller.

[0093] Step 320: Determine whether there is water accumulation on the ground of the acquisition environment from which the point cloud data is collected, based on the number of ground sampling points.

[0094] The laser points emitted by the LiDAR device towards the ground height range will theoretically be reflected by the ground and then collected by the LiDAR device. However, in practice, if there is water on the ground, the water on the ground will cause the laser points to be abnormally reflected and thus cannot be collected by the LiDAR device. Based on this principle, the number of ground sampling points in the point cloud data can be combined to determine whether there is water on the ground in the collection environment where the point cloud data comes from. The ground height range can be Figure 4 The height range corresponding to area 1 in the figure can be set according to actual needs.

[0095] The number of laser points emitted by the laser radar device facing the ground height range can be obtained as a first number. Afterwards, if the number of ground sampling points is less than the product of the first number and a preset coefficient, it is determined that there is water accumulation on the ground of the physical environment from which the point cloud data originates. Conversely, if the number of ground sampling points is not less than the product of the first number and the preset coefficient, it is determined that there is no water accumulation on the ground of the physical environment from which the point cloud data originates. The preset coefficient can be set according to actual needs, for example, 50%, 55%, 58%, 60%, etc.

[0096] Step 330: If it is determined that there is no accumulated water on the ground of the physical environment from which the point cloud data originates, laser radar target perception is performed on the point cloud data to obtain a first perception result.

[0097] There is no accumulated water on the ground of the physical environment where the point cloud data comes from, and the number of valid sampling points in the point cloud data is relatively large. In this case, laser radar target perception is performed on the point cloud data to obtain the first perception result, which is also highly accurate.

[0098] If there is accumulated water on the ground in the collection environment where the point cloud data comes from, the number of sampling points representing the ground in the point cloud data is small. In this case, the first perception result obtained by performing target perception on the point cloud data may have a large error. Therefore, in this case, laser radar target perception may not be performed on the point cloud data, and the attributes of the fused target may be updated subsequently through at least one second perception result.

[0099] In the above embodiment, after determining the number of ground sampling points and the absence of water in the ground where the point cloud data is collected, lidar target perception is performed on the point cloud data to obtain a first perception result. This can avoid the situation where the presence of water on the ground leads to an insufficient number of effective sampling points in the point cloud data, which in turn leads to an inaccurate first perception result obtained by perception.

[0100] In some embodiments, after step 320, the method further includes: if it is determined that there is accumulated water on the ground of the physical environment from which the point cloud data originates, the attribute perception information of each reference target in at least one second perception result is used to update the attributes of the fused target in the target association result, and obtain the attribute tracking result of each fused target at the target time point.

[0101] When it is determined that there is accumulated water on the ground of the physical environment from which the point cloud data originates, there is no need to perform lidar target perception on the point cloud data. In this case, there is no need to involve the lidar object in target association and to update the attributes of the fused target. Instead, the attribute perception information of each reference target in at least one second perception result is used to update the attributes of the fused target in the target association result, thereby ensuring the accuracy of the attribute tracking result of the fused target at the target time point.

[0102] In some embodiments, after step 130, the method further includes: if it is identified that there are no abnormal reflection points in the point cloud data, the attribute perception information of each lidar target in the first perception result and the attribute perception information of each reference target in at least one second perception result are used to update the attributes of the fused target in the target association result to obtain the attribute tracking results of each fused target at the target time point.

[0103] If it is identified that there are no abnormal reflection points in the point cloud data, it can be determined that the accuracy and reliability of the first perception result are relatively high. Therefore, the attribute perception information of each lidar target in the first perception result and the attribute perception information of each reference target in at least one second perception result are used to update the attributes of the fused target in the target association result to obtain the attribute tracking results of each fused target at the target time point.

[0104] Figure 5 is a flow chart of a target fusion tracking method according to a specific embodiment of the present application. Figure 5 The illustrated embodiment can be applied to a vehicle, and perception hardware can be deployed in the vehicle, and the perception hardware includes a laser radar device, a millimeter wave radar device, and a camera. The camera can be an RGB camera or an infrared camera; the millimeter wave radar device can be an ordinary 2.5D millimeter wave radar device or a 4D millimeter wave radar device; the laser radar device can be a mechanical laser radar device or a semi-solid laser radar device, or a solid-state laser radar device.

[0105] Considering that laser radar equipment usually needs to work in various outdoor climates, especially in low temperature and high humidity environments, the surface of the window of the laser radar equipment is in contact with the low temperature and high humidity environment, and the inner surface of the window is at a high temperature due to laser heating, which is easy to cause condensation on the outer surface of the window, which will affect the light emission and reception of the laser radar equipment and cause the inability to perceive the external environment normally. Based on this consideration, the window of the laser radar equipment used in this application may include a window substrate and a waterproof film; the window substrate is provided with a light-transmitting area; the area of ​​the light-transmitting film is larger than the area of ​​the light-transmitting area, and the waterproof film is attached to the outer surface of the window substrate and covers all the light-transmitting areas of the window substrate, so that condensation cannot adhere to the surface of the waterproof film, and condensation will not form on the window substrate accordingly. The laser radar window used can avoid the generation of condensation and avoid the reduction of target perception ability due to the small number of point clouds returned by the laser radar equipment.

[0106] like Figure 5 As shown, a list of millimeter-wave radar devices collected by the millimeter-wave radar device at the target time point can be obtained as the millimeter-wave radar target perception result.

[0107] The image data collected by the camera at the target time point is obtained, and visual target perception is performed on the image data to obtain the visual target perception result.

[0108] The point cloud data collected by the laser radar device at the target time point is obtained; then, whether there is water on the road surface in the acquisition environment where the point cloud data is collected is identified based on the point cloud data. If water is identified, the point cloud data is determined to be unavailable, and then the attributes are updated based on the millimeter-wave radar target in the millimeter-wave radar target perception result and the visual target in the visual target perception result. If it does not exist, the point cloud data is subjected to laser radar target perception to obtain the laser radar target perception result (i.e., the first perception result mentioned above).

[0109] Afterwards, the millimeter-wave radar target in the millimeter-wave radar target perception result, the visual target in the visual target perception result, and the lidar target in the lidar target perception result can be associated with each other to obtain the target association result at the target time point.

[0110] Afterwards, the target association results at the target time point and the lidar target perception results are combined to identify whether there are abnormal reflection points in the point cloud data.

[0111] In some embodiments, if it is determined that there is a first laser radar target in the laser radar target perception result according to the laser radar target perception result, it is determined that there is an abnormal reflection point in the point cloud data, wherein the first laser radar target satisfies the following four points:

[0112] 1) The lateral width of the first laser radar target at the target time point is greater than 1.2 times the lateral width of the corresponding visual target in the visual target perception result, or the lateral width of the first laser radar target at the target time point is greater than 1.2 times the lateral width of the first laser radar target tracked and determined at the previous time point of the target time point; (i.e., the first condition above)

[0113] 2) The first laser radar target is located in the longitudinal area 5 to 80 meters in front of the current vehicle (i.e., the target device mentioned above) and the lateral area -20 to 20 meters; (i.e., the second condition mentioned above)

[0114] 3) The speed of the first laser radar target at the target time point is > 20m / s; (i.e. the third condition above)

[0115] 4) The sensing area of ​​the first laser radar target in the laser radar target sensing result is greater than 1.2 times the sensing area of ​​the fusion target represented by it at the previous acquisition time point of the target time point; or, the sensing area of ​​the first laser radar target in the laser radar target sensing result is less than 0.8 times the sensing area of ​​the fusion target represented by it at the previous acquisition time point of the target time point. (i.e. the fourth condition above)

[0116] If abnormal reflection points are identified in the point cloud data, it is determined that the lidar target perception results are unavailable, and the corresponding attributes are subsequently updated according to the millimeter-wave radar target in the millimeter-wave radar target perception results and the visual target in the visual target perception results. The attribute perception information in the lidar target perception results is not used for attribute update.

[0117] If it is identified that there are no abnormal reflection points in the point cloud data, the millimeter-wave radar target perception results, the visual target perception results, and the lidar target perception results are used to update the fused target attributes to obtain the attribute information of the fused target at the target time point.

[0118] After obtaining the attribute information of the fusion target at the target time point, the attribute information of the fusion target at the target time point is associated with the attribute information of the fusion target at a time point before the target time point to achieve attribute tracking of the fusion target.

[0119] In the above embodiment, before performing lidar target perception, it is determined whether there is water accumulation on the road surface based on the number of ground sampling points in the point cloud data. If there is water accumulation, there is no need to perform lidar target perception on the point cloud data, so as to avoid the situation where the number of valid sampling points in the collected point cloud data is small due to water accumulation on the road surface, which in turn leads to inaccurate attribute tracking results being used for attribute updating using inaccurate lidar target perception results.

[0120] In addition, after target association, the LiDAR target perception results are used to identify whether there are abnormal reflection points in the point cloud data. If abnormal reflection points are identified in the cloud data, the LiDAR target perception results are not used for attribute updates to avoid changes in the shape and position of the LiDAR target due to splashing water in puddles, which would cause the attributes perceived in the LiDAR target perception results to be inconsistent with the actual ones.

[0121] The following describes an apparatus embodiment of the present application, which can be used to execute the method in the above-mentioned embodiment of the present application. For details not disclosed in the apparatus embodiment of the present application, please refer to the above-mentioned method embodiment of the present application.

[0122] Figure 6 is a block diagram of a target fusion tracking device according to an embodiment of the present application. Figure 6 As shown, the target fusion tracking device includes:

[0123] The target perception module 610 is used to perform target perception based on the detection data collected at the target time point and determine the target perception result; the target perception result includes a first perception result and at least one second perception result; the first perception result is obtained by performing laser radar target perception on the point cloud data in the detection data;

[0124] A target association module 620, configured to perform target association according to the first perception result and at least one second perception result, and determine a target association result;

[0125] An identification module 630 is used to identify whether there are abnormal reflection points in the point cloud data according to the attribute perception information of each laser radar target and the target association result in the first perception result;

[0126] The attribute update module 640 is used to update the attributes of the fused targets in the target association result using the attribute perception information of each reference target in at least one second perception result if abnormal reflection points are identified in the point cloud data, so as to obtain the attribute tracking results of each fused target at the target time point.

[0127] In some embodiments, the attribute perception information of the laser radar target includes the lateral width of the laser radar target; the identification module 630 includes: a reference lateral width acquisition unit, which is used to obtain the reference lateral width of the fusion target represented by the laser radar target according to the target association result for each laser radar target in the first perception result; the reference lateral width is the lateral width of the reference target representing the same fusion target as the laser radar target in at least one second perception result, or is the lateral width of the fusion target represented by the laser radar target tracked and determined at the previous acquisition time point of the target time point; a width difference determination unit, which is used to determine the width difference between the lateral width of each laser radar target in the first perception result and the reference lateral width of the fusion target represented; an identification unit, which is used to identify whether there are abnormal reflection points in the point cloud data based on the width difference.

[0128] In some embodiments, the identification unit is used to: if it is determined that the first condition is met based on the width difference, determine that there are abnormal reflection points in the point cloud data; wherein the first condition is that there is at least one first lidar target in the first perception result; the first lidar target refers to a lidar target whose width ratio exceeds a first ratio threshold, and the width ratio refers to the ratio of the lateral width in the first perception result to the corresponding reference lateral width; the first ratio threshold is greater than 1.

[0129] In other embodiments, the detection data is collected by a sensor device disposed on the target device; the identification unit is used to: determine that the point cloud data has an abnormal reflection point if the first condition is determined to be satisfied according to the width difference, and the first sensing result satisfies at least one of the second condition, the third condition, and the fourth condition;

[0130] The first condition is that there is at least one first laser radar target in the first perception result, the first laser radar target refers to a laser radar target whose width ratio exceeds a first ratio threshold, the width ratio refers to the ratio of the lateral width in the first perception result to the corresponding reference lateral width; the first ratio threshold is greater than 1;

[0131] The second condition is that the first laser radar target in the first perception result is located in front of the target device at the target time point and is located in a non-safe interval driving area relative to the target device;

[0132] The third condition is that the speed of the first laser radar target at the target time point in the first perception result exceeds the first speed threshold;

[0133] The fourth condition is that the first ratio between the first perception area and the second perception area of ​​the first lidar target at the target time point in the first perception result exceeds the second ratio threshold, or the first ratio between the first perception area and the second perception area is less than the third ratio threshold; wherein the second ratio threshold is greater than 1 and the third ratio threshold is less than 1; the second perception area refers to the perception area determined by tracking the fusion target represented by the first lidar target at the previous acquisition time point of the target time point.

[0134] In some embodiments, the detection data includes first sensor data collected by a reference sensor device and point cloud data collected by a lidar device; the target perception module 610 includes: a statistical unit, used to count the number of ground sampling points belonging to the ground in the point cloud data; a water accumulation identification unit, used to determine whether there is water accumulation on the ground of the collection environment from which the point cloud data comes from according to the number of ground sampling points; a first perception unit, used to perform lidar target perception on the point cloud data to obtain a first perception result if it is determined that there is no water accumulation on the ground of the physical environment from which the point cloud data comes; and a second perception unit, used to perform reference target perception on the first sensor data to obtain at least one second perception result.

[0135] In some embodiments, the target fusion tracking device also includes: a first attribute update module, which is used to use the attribute perception information of each reference target in at least one second perception result to update the attributes of the fused target in the target association result if it is determined that there is water accumulation on the ground of the physical environment from which the point cloud data originates, so as to obtain the attribute tracking results of each fused target at the target time point.

[0136] In some embodiments, the reference sensing device includes at least one of an image acquisition device and a millimeter wave radar device.

[0137] In some embodiments, the target fusion tracking device also includes: a second attribute update module, which is used to update the attributes of the fused target in the target association result using the attribute perception information of each lidar target in the first perception result and the attribute perception information of each reference target in at least one second perception result if it is identified that there are no abnormal reflection points in the point cloud data, so as to obtain the attribute tracking results of each fused target at the target time point.

[0138] Figure 7 1 is a schematic diagram of the structure of an electronic device according to an embodiment of the present application. The electronic device can be used to execute the target fusion tracking method provided by the present application. Figure 7As shown, the electronic device may include: a processor 1001, such as a CPU, a network interface 1004, a user interface 1003, a memory 1005, and a communication bus 1002. Among them, the communication bus 1002 is used to realize the connection and communication between these components. The user interface 1003 may include a display screen (Display), an input unit such as a keyboard (Keyboard), and optionally, the user interface 1003 may also include a standard wired interface and a wireless interface. The network interface 1004 may optionally include a standard wired interface and a wireless interface (such as a WI-FI interface). The memory 1005 may be a high-speed RAM memory, or it may be a stable memory (non-volatile memory), such as a disk memory. The memory 1005 may optionally also be a storage device independent of the aforementioned processor 1001. Those skilled in the art will understand that, Figure 7 The structure of the electronic device shown in the figure does not constitute a limitation of the electronic device, and may include more or less components than shown in the figure, or combine certain components, or arrange the components differently.

[0139] like Figure 7 As shown, the memory 1005 as a computer-readable storage medium may include an operating system, a network communication module, a user interface module, and a program for implementing the target fusion tracking method.

[0140] exist Figure 7 In the electronic device shown, the network interface 1004 is mainly used to communicate with other devices. The user interface 1003 is mainly used to connect to the client and perform data communication with the client; and the processor 1001 can be used to call the program for implementing the target fusion tracking method stored in the memory 1005, and execute the steps of the target fusion tracking method in any of the above method embodiments.

[0141] According to one aspect of an embodiment of the present application, a computer-readable storage medium is provided, on which computer-readable instructions are stored. When the computer-readable instructions are executed by a processor, the target fusion tracking method in any of the above method embodiments is implemented.

[0142] According to one aspect of an embodiment of the present application, a computer program product is provided, which includes computer instructions. When the computer instructions are executed by a processor, the target fusion tracking method in any of the above method embodiments is implemented.

[0143] The units involved in the embodiments of the present application may be implemented by software or hardware, and the units described may also be set in a processor. The names of these units do not, in some cases, constitute limitations on the units themselves.

[0144] Through the description of the above implementation methods, it is easy for those skilled in the art to understand that the example implementation methods described here can be implemented by software or by combining software with necessary hardware. Therefore, the technical solution according to the implementation methods of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a touch terminal, or a network device, etc.) to execute the method according to the implementation methods of the present application.

[0145] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the embodiments disclosed herein. The present application is intended to cover any variations, uses or adaptations of the present application, which follow the general principles of the present application and include common knowledge or customary technical means in the art that are not disclosed in the present application.

[0146] It should be understood that the present application is not limited to the precise structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

Claims

1. A target fusion tracking method, characterized in that: include: Perform target perception based on the detection data collected at the target time point and determine the target perception result; The target perception result includes a first perception result and at least one second perception result; The first perception result is obtained by performing laser radar target perception on the point cloud data in the detection data; Performing target association according to the first perception result and the at least one second perception result to determine a target association result; According to the attribute perception information of each laser radar target in the first perception result and the target association result, identifying whether there is an abnormal reflection point in the point cloud data; If abnormal reflection points are identified in the point cloud data, the attribute perception information of each reference target in the at least one second perception result is used to update the attributes of the fused target in the target association result to obtain the attribute tracking result of each fused target at the target time point.

2. The method according to claim 1, characterized in that The attribute perception information of the laser radar target includes the lateral width of the laser radar target; The step of identifying whether the point cloud data has an abnormal reflection point according to the attribute perception information of each laser radar target in the first perception result and the target association result includes: For each laser radar target in the first perception result, according to the target association result, a reference lateral width of the fused target represented by the laser radar target is obtained; the reference lateral width is the lateral width of a reference target representing the same fused target as the laser radar target in the at least one second perception result, or is the lateral width of the fused target represented by the laser radar target tracked and determined at a previous acquisition time point of the target time point; Determine the width difference between the lateral width of each of the laser radar targets in the first perception result and the reference lateral width of the represented fusion target; According to the width difference, it is identified whether there are abnormal reflection points in the point cloud data.

3. The method according to claim 2, characterized in that The step of identifying whether the point cloud data has an abnormal reflection point according to the width difference includes: If the first condition is determined to be satisfied based on the width difference, it is determined that there are abnormal reflection points in the point cloud data; wherein, the first condition is that there is at least one first lidar target in the first perception result; the first lidar target refers to a lidar target whose width ratio exceeds a first ratio threshold, and the width ratio refers to the ratio of the lateral width in the first perception result to the corresponding reference lateral width; the first ratio threshold is greater than 1.

4. The method according to claim 2, characterized in that: The detection data is collected by a sensor device disposed on the target device; The step of identifying whether the point cloud data has an abnormal reflection point according to the width difference includes: If it is determined that the first condition is met according to the width difference, and the first sensing result satisfies at least one of the second condition, the third condition and the fourth condition, it is determined that there is an abnormal reflection point in the point cloud data; The first condition is that there is at least one first laser radar target in the first perception result, the first laser radar target refers to a laser radar target whose width ratio exceeds a first ratio threshold, the width ratio refers to the ratio of the lateral width in the first perception result to the corresponding reference lateral width; the first ratio threshold is greater than 1; The second condition is that the first laser radar target in the first perception result is located in front of the target device at the target time point and is located in a non-safe interval driving area relative to the target device; The third condition is that the speed of the first laser radar target at the target time point in the first perception result exceeds a first speed threshold; The fourth condition is that in the first perception result, the first ratio between the first perception area and the second perception area of ​​the first lidar target at the target time point exceeds the second ratio threshold, or the first ratio between the first perception area and the second perception area is less than the third ratio threshold; wherein the second ratio threshold is greater than 1, and the third ratio threshold is less than 1; the second perception area refers to the perception area tracked and determined by the fusion target represented by the first lidar target at the previous acquisition time point of the target time point.

5. The method according to any one of claims 1 to 4, characterized in that The detection data includes the first sensing data collected by the reference sensing device and the point cloud data collected by the laser radar device; The performing target perception according to the detection data collected at the target time point and determining the target perception result includes: Counting the number of ground sampling points belonging to the ground in the point cloud data; Determine, based on the number of the ground sampling points, whether there is water accumulation on the ground of the collection environment from which the point cloud data is collected; If it is determined that there is no accumulated water on the ground of the physical environment from which the point cloud data originates, performing laser radar target perception on the point cloud data to obtain the first perception result; A reference target is sensed on the first sensing data to obtain the at least one second sensing result.

6. The method according to claim 5, characterized in that After determining whether there is water accumulation on the ground of the acquisition environment from which the point cloud data is collected based on the number of the ground sampling points, the method further includes: If it is determined that there is accumulated water on the ground of the physical environment from which the point cloud data originates, the attribute perception information of each reference target in the second perception result is used to update the attributes of the fused target in the target association result to obtain the attribute tracking result of each fused target at the target time point.

7. The method according to claim 5, characterized in that The reference sensing device includes at least one of an image acquisition device and a millimeter wave radar device.

8. The method according to any one of claims 1 to 4, characterized in that: After identifying whether there are abnormal reflection points in the point cloud data according to the attribute perception information of each laser radar target in the first perception result and the target association result, the method further includes: If it is identified that there are no abnormal reflection points in the point cloud data, the attribute perception information of each lidar target in the first perception result and the attribute perception information of each reference target in the second perception result are used to update the attributes of the fused target in the target association result to obtain the attribute tracking result of each fused target at the target time point.

9. A target fusion tracking device, characterized in that: include: The target perception module is used to perceive the target based on the detection data collected at the target time point and determine the target perception result; The target perception result includes a first perception result and at least one second perception result; The first perception result is obtained by performing laser radar target perception on the point cloud data in the detection data; a target association module, configured to perform target association according to the first perception result and the at least one second perception result, and determine a target association result; an identification module, configured to identify whether the point cloud data has an abnormal reflection point according to the attribute perception information of each laser radar target in the first perception result and the target association result; The attribute update module is used to use the attribute perception information of each reference target in the at least one second perception result to update the attributes of the fused target in the target association result if abnormal reflection points are identified in the point cloud data, so as to obtain the attribute tracking results of each fused target at the target time point.

10. An electronic device, characterized in that: include: processor; A memory, wherein computer instructions are stored in the memory, and when the computer instructions are executed by the processor, the method according to any one of claims 1 to 8 is implemented.

11. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and when the computer instructions are executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images