Target position correction method and device based on topological relation, vehicle and medium

Abstract

The invention relates to the technical field of automatic driving environment perception, and discloses a topological relation-based target position correction method and device, a vehicle and a medium, and the method comprises the steps: obtaining an observation target and a tracking target, the observation target being at least one perception target contained in the current road perception data of a target vehicle; the tracking target refers to at least one sensing target contained in historical road sensing data of the target vehicle; respectively constructing an observation topological matrix corresponding to the observation target and a tracking topological matrix corresponding to the tracking target; respectively calculating a first topological relation corresponding to the observation topological matrix and a second topological relation corresponding to the tracking topological matrix; and calculating a target correction displacement based on the first topological relation and the second topological relation, and performing position correction on the current position of the observation target based on the target correction displacement. According to the method, the position of the target is corrected by constructing the topological relation corresponding to the observation target and the tracking target, and the authenticity and accuracy of the sensing target can be improved.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving environment perception technology, and in particular to a target position correction method, device, vehicle and medium based on topological relationship. Background Art

[0002] In recent years, with the progress of the times, all walks of life have also sprung up and flourished, and all kinds of products must have the important label of intelligence. As people become more and more wealthy, the demand for car purchases is gradually increasing. The focus of automobile intelligence is reflected in autonomous driving, which not only meets customer needs, but is also one of the important means to boost the brand. Under the general trend, autonomous driving has become an important breakthrough direction for vehicle manufacturers and has become a new round of industry craze.

[0003] At present, although autonomous driving has made breakthrough progress with the help of both hardware and software, it is still far from human perception and risk identification. Therefore, it is still difficult to achieve autonomous driving on open roads, and structured roads have become the main scenario for the implementation of automated driving. Even so, the diversity of real-life scenarios and the limitations of sensors themselves still cause various problems in environmental perception, such as the unstable visual target ranging caused by the overall longitudinal position jump of the target in consecutive frames caused by the posture of the whole vehicle, the overall lateral position drift problem caused by the turning of the radar sensor, and the multi-source heterogeneous sensor caused by the large position error of the perceived target, causing the missed association problem and then causing the fusion target position fluctuation. Therefore, accurate sensor detection has a vital impact on the single sensor / multi-source sensor association fusion, and thus plays a key role in the vehicle's motion planning and control.

[0004] However, complex road scenes in practical applications have an irresistible impact on single visual sensors, radars or multi-source heterogeneous sensors, such as target loss, target splitting and measurement anomalies. These problems can easily lead to the inability of perceived targets to accurately express real environmental information, and large errors in target positions, making it difficult to achieve better autonomous driving environment reconstruction and unable to meet the intelligent needs of vehicle autonomous driving. Summary of the invention

[0005] In view of this, the present invention provides a target position correction method, device, vehicle and medium based on topological relationships to solve the many defects of the existing environmental perception targets proposed in the above technical background, which cannot accurately express the environmental reality, thereby seriously affecting the reconstruction effect of the autonomous driving environment and making it difficult to meet the intelligent needs of autonomous driving.

[0006] In a first aspect, the present invention provides a method for correcting a target position based on a topological relationship, the method comprising:

[0007] Acquire an observation target and a tracking target, wherein the observation target refers to at least one perception target included in the current road perception data of the target vehicle; the tracking target refers to at least one perception target included in the historical road perception data of the target vehicle;

[0008] Construct the observation topology matrix corresponding to the observation target and the tracking topology matrix corresponding to the tracking target respectively;

[0009] respectively calculating a first topological relationship corresponding to the observation topological matrix and a second topological relationship corresponding to the tracking topological matrix;

[0010] The target correction displacement is calculated based on the first topological relationship and the second topological relationship, and the current position of the observed target is corrected based on the target correction displacement.

[0011] The present invention calculates the first topological relationship and the second topological relationship correspondingly by constructing an observation topological matrix corresponding to the observation target and a tracking topological matrix corresponding to the tracking target, calculates the target correction displacement according to the first topological relationship and the second topological relationship, and performs position correction on the current position of the observation target based on the target correction displacement. This can improve the authenticity of the expression of the driving environment by the observation target, make the perceived target obtained by the current actual observation more real and reliable, help to achieve better reconstruction of the autonomous driving environment, and greatly meet the intelligent needs of vehicle autonomous driving.

[0012] In an optional implementation, constructing an observation topology matrix corresponding to the observation target and a tracking topology matrix corresponding to the tracking target respectively includes:

[0013] Determine a first target number of the observed targets, and construct an observation topology matrix corresponding to the observed targets based on the first target number and current road perception data;

[0014] The number of second targets of the tracking target is determined, and a tracking topology matrix corresponding to the tracking target is constructed based on the number of second targets and historical road perception data.

[0015] The present invention constructs an observation topology matrix according to the first target number of the observed targets and the current road perception data, and constructs a tracking topology matrix according to the second target number of the tracked targets and the historical road perception data. This can ensure the construction accuracy of the corresponding topology matrix, help provide data support for the calculation of subsequent topological relationships, and further ensure the accuracy of the target correction displacement.

[0016] In an optional implementation, respectively calculating the first topological relationship corresponding to the observation topological matrix and the second topological relationship corresponding to the tracking topological matrix includes:

[0017] Determine the first target number of the observed targets according to the observation topology matrix and determine the second target number of the tracked targets according to the tracking topology matrix;

[0018] If the number of first targets is 1, the first topological relationship is determined to be a point topological relationship; if the number of first targets is 2, the first topological relationship is determined to be a line topological relationship; if the number of first targets is at least 3, the first topological relationship is determined to be a triangle topological relationship;

[0019] If the number of second targets is 1, the second topological relationship is determined to be a point topological relationship; if the number of second targets is 2, the second topological relationship is determined to be a line topological relationship; if the number of second targets is at least 3, the second topological relationship is determined to be.

[0020] The present invention determines the corresponding first topological relationship by observing the first target number of the target, and determines the corresponding second topological relationship by tracking the second target number of the target, thereby realizing a reasonable division of the topological relationships of the current and historical corresponding environmental perception targets, which can improve the authenticity, accuracy and reliability of the perceived targets, making the current perceived targets more authentic and reliable, and further realizing better reconstruction of the autonomous driving environment.

[0021] In an optional implementation, respectively calculating the first topological relationship corresponding to the observation topological matrix and the second topological relationship corresponding to the tracking topological matrix includes:

[0022] Determine a first sensor type for observing a target according to an observation topology matrix, and determine a second sensor type for tracking a target according to a tracking topology matrix;

[0023] If the first sensor type is a visual sensor, the first topological relationship is determined to be a point topological relationship; if the first sensor type is a radar sensor, the first topological relationship is determined to be a line topological relationship; if the first sensor type includes at least a visual sensor and a radar sensor, the first topological relationship is determined to be a triangle topological relationship;

[0024] If the second sensor type is a visual sensor, the second topological relationship is determined to be a point topological relationship; if the second sensor type is a radar sensor, the second topological relationship is determined to be a line topological relationship; if the second sensor type includes at least a visual sensor and a radar sensor, the second topological relationship is determined to be a triangle topological relationship.

[0025] The present invention takes into account the characteristics of different sensors in actual autonomous driving environment perception scenarios, and determines the topological relationship between current and historical corresponding environmental perception targets through the type of sensor, which can ensure the rationality of topological relationship calculation, thereby improving the accuracy of perceived targets, enhancing the true expression of perceived targets for the actual driving environment, and meeting the intelligent needs of vehicle autonomous driving.

[0026] In an optional implementation, calculating the target correction displacement based on the first topological relationship and the second topological relationship includes:

[0027] Calculating an optimal matching pair corresponding to the first topological relationship and the second topological relationship, wherein the optimal matching pair includes an observed target and a tracked target;

[0028] The target correction displacement is calculated based on the best matching pair.

[0029] The present invention calculates the optimal matching pair corresponding to the first topological relationship and the second topological relationship, and determines the target correction displacement according to the optimal matching pair, so that the observed target achieves the optimal expression of the real driving environment, ensures the accuracy of the current perceived target position, and contributes to the subsequent realization of better autonomous driving environment reconstruction and enhances the user's car experience.

[0030] In an optional implementation, calculating the optimal matching pair corresponding to the first topological relationship and the second topological relationship includes:

[0031] Performing similarity matching on the first topological relationship and the second topological relationship, and correspondingly obtaining a matching result including at least one matching pair;

[0032] Determine a preset matching number of the optimal matching pair according to the first topological relationship and the second topological relationship;

[0033] Based on the preset number of matches, a corresponding number of similar matching pairs are screened from the matching results to obtain the optimal matching pair.

[0034] The present invention designs three types of topological relationships corresponding to the perceived targets, including point topological relationships, line topological relationships, and triangle topological relationships. By screening the optimal matching pairs according to the corresponding set number of matches, the calculation accuracy of the optimal matching pairs can be guaranteed, thereby improving the subsequent correction accuracy of the current position of the observed target, making the observed target more real and reliable, and contributing to a better reconstruction of the autonomous driving environment.

[0035] In an optional implementation, calculating the target correction displacement based on the best matching pair includes:

[0036] Taking the target vehicle as the origin, calculate the position coordinates of each optimal matching pair;

[0037] If the number of the best matching pairs is 1, the target correction displacement is calculated based on the position coordinates of the observed target and the position coordinates of the tracked target in the best matching pair;

[0038] If the number of optimal matching pairs is 2 or 3, the corresponding displacements are calculated according to the position coordinates of the observed target and the position coordinates of the tracked target in each optimal matching pair, and the mean calculation result of all displacements is determined as the target correction displacement.

[0039] The present invention calculates the target correction displacement corresponding to the current position of the observed target according to the number of optimal matching pairs, which can improve the accuracy of the target correction displacement, make the current position of the observed target more consistent with the actual driving environment, and help improve the autonomous driving environment reconstruction effect and user experience.

[0040] In an optional implementation, obtaining an observed target and a tracked target includes:

[0041] Collecting the road environment information of the target vehicle at the current position through the preset sensor and obtaining the current collection time corresponding to the road environment information;

[0042] Performing standardized processing on the road environment information to obtain current road perception data, and determining the observation target from the perception targets contained in the current road perception data;

[0043] A target sequence corresponding to a previous acquisition time is read from a preset tracking list, and a tracking target is determined based on the target sequence, wherein the preset tracking list is a time series table constructed based on historical road perception data of the target vehicle.

[0044] The present invention takes into account the various sensor types involved in the autonomous driving environment perception technology, obtains the current road perception data after standardizing the road environment information collected by the preset sensors, and determines the observation target from the perception targets contained in the current road perception data; and obtains the tracking target from the time series table constructed by the historical road perception data of the target vehicle according to the current collection time, which can ensure the diversity, standardization and accuracy of the acquisition of observation targets and tracking targets.

[0045] In an optional implementation, the process of building the preset tracking list includes:

[0046] Collect historical road environment information of the target vehicle at different locations through preset sensors and obtain historical collection time corresponding to the historical road environment information;

[0047] All historical road environment information is processed in a standardized manner, and corresponding historical road perception data including at least one perception target is obtained;

[0048] A preset tracking list is constructed based on each historical road perception data and its corresponding historical collection time.

[0049] The present invention can ensure the standardization of the data in the preset tracking list by standardizing the historical road environment information of the target vehicle at different positions collected by the preset sensor, and constructing a preset tracking list in combination with its corresponding historical collection time, thereby providing data support for the position correction of the observed target.

[0050] In an optional implementation, before respectively constructing an observation topology matrix corresponding to the observation target and a tracking topology matrix corresponding to the tracking target, the target position correction method based on the topological relationship further includes:

[0051] The observation target and the tracking target are subjected to preset synchronization processing, and the processed observation target and tracking target are obtained accordingly, wherein the preset synchronization processing includes time compensation processing and space synchronization processing.

[0052] The present invention performs time compensation processing and space synchronization processing on the observed target and the tracked target, so that the observed target and the tracked target can be in the same time and space scale, which helps to improve the accuracy and effectiveness of subsequent target position correction.

[0053] In a second aspect, the present invention provides a target position correction device based on topological relationship, the device comprising:

[0054] An acquisition module, used to acquire an observation target and a tracking target, wherein the observation target refers to at least one perception target included in the current road perception data of the target vehicle; the tracking target refers to at least one perception target included in the historical road perception data of the target vehicle;

[0055] A construction module, used to respectively construct an observation topology matrix corresponding to the observation target and a tracking topology matrix corresponding to the tracking target;

[0056] A calculation module, used to respectively calculate a first topological relationship corresponding to the observation topological matrix and a second topological relationship corresponding to the tracking topological matrix;

[0057] The correction module is used to calculate the target correction displacement based on the first topological relationship and the second topological relationship, and perform position correction on the current position of the observed target based on the target correction displacement.

[0058] The target position correction device based on topological relationship of the present invention corrects the position of the target by constructing the topological relationship corresponding to the observation target and the tracking target, which can improve the authenticity of the expression of the driving environment by the observation target, making the observation target more realistic and reliable, which not only improves the effect of reconstructing the actual autonomous driving environment, but also meets the intelligent needs of vehicle autonomous driving.

[0059] In a third aspect, the present invention provides a vehicle, comprising a controller, the controller comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, computer instructions being stored in the memory, and the processor executing the computer instructions to execute a target position correction method based on a topological relationship according to the first aspect or any corresponding embodiment thereof.

[0060] In a fourth aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute a target position correction method based on a topological relationship according to the first aspect or any corresponding embodiment thereof.

[0061] The target position correction method and device based on topological relationship of the present invention constructs a corresponding observation topological matrix for the acquired observation target and a tracking topological matrix corresponding to the tracking target, respectively calculates the corresponding first topological relationship and second topological relationship in combination with the two constructed matrices, and calculates the target correction displacement according to the two topological relationships, which is used to correct the current position of the observation target. This can improve the authenticity of the expression of the observation target to the driving environment, make the perceived target obtained by the current actual observation more real and reliable, help to achieve better reconstruction of the autonomous driving environment, and also greatly meet the intelligent needs of vehicle autonomous driving. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0063] Figure 1 is a schematic flow chart of a target position correction method based on topological relationship according to an embodiment of the present invention;

[0064] Figure 2 is a schematic flow chart of another target position correction method based on topological relationship according to an embodiment of the present invention;

[0065] Figure 3 It is a schematic diagram of point topological relationship;

[0066] Figure 4 It is a schematic diagram of line topology relationship;

[0067] Figure 5 It is a schematic diagram of triangular topological relationship;

[0068] Figure 6 It is a flowchart of visual target position correction;

[0069] Figure 7 It is a schematic diagram of the process of radar target position correction;

[0070] Figure 8 It is a flowchart of the associated target position correction;

[0071] Fig. 9is a structural block diagram of a target position correction device based on topological relationship according to an embodiment of the present invention;

[0072] Fig.10 It is a schematic diagram of the structure of a controller of a vehicle according to an embodiment of the present invention. DETAILED DESCRIPTION

[0073] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0074] An embodiment of the present invention provides an embodiment of a target position correction method based on a topological relationship. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0075] In this embodiment, a target position correction method based on topological relationship is provided. Figure 1 FIG. 1 is a flow chart of a method for correcting a target position based on a topological relationship according to an embodiment of the present invention. Figure 1 As shown, the process includes the following steps:

[0076] Step S101, obtaining an observation target and a tracking target, wherein the observation target refers to at least one perception target included in the current road perception data of the target vehicle; the tracking target refers to at least one perception target included in the historical road perception data of the target vehicle.

[0077] It should be noted that the current road perception data in this embodiment is obtained by real-time collection of the target vehicle's surrounding environment data at the current moment; the historical road perception data is obtained by real-time collection of the target vehicle's surrounding environment data at different historical moments. Among them, the specific content of the perception target contained in the target vehicle's surrounding environment data can be adaptively adjusted according to the actual driving scene, such as the perception target includes lanes, lane lines, other vehicles, pedestrians, etc., which is only an example and is not limited to this.

[0078] Step S102, constructing an observation topology matrix corresponding to the observation target and a tracking topology matrix corresponding to the tracking target respectively.

[0079] It should be noted that the specific construction of the observation topology matrix and the tracking topology matrix in this embodiment can refer to the conventional construction process of the matrix, that is, the process of determining the dimension of the matrix, sorting and assigning values ​​to each element in the matrix to obtain the constructed matrix, and combining the actual project requirements and the adaptability of the obtained target vehicle surrounding environment data. For example, the dimension of the matrix (that is, the number of rows and columns of the matrix) is determined according to the type of the perceived target (such as cars, pedestrians, etc.) and the corresponding number thereof, and each element in the matrix is ​​assigned a value according to a specific rule or algorithm (such as setting the corresponding identification of cars and pedestrians in advance), and the topological matrix corresponding to the current perceived target can be obtained.

[0080] Step S103, respectively calculating a first topological relationship corresponding to the observation topological matrix and a second topological relationship corresponding to the tracking topological matrix.

[0081] It should be noted that topological relationships are used to clearly define spatial structural relationships. They describe the nature of the relationship between graphics that remains unchanged when the graphics are deformed (such as scaling, rotation, and stretching) while maintaining a continuous state. They only focus on the relative position relationship between objects, regardless of the actual size, shape, or direction of the objects. In autonomous driving environment perception, the application of topological relationships is mainly reflected in the representation of state space. The environmental perception system of autonomous vehicles needs to distinguish between free space and obstacles, obtain obstacle maps through sensor data, and store these data as structured information for use in the autonomous driving operation phase. Specifically, the topological representation models the state space as a graph, where nodes identify important locations (or features) and edges represent the topological relationships between them, such as location, direction, proximity, and connectivity; this type of topological representation helps ensure that driverless cars can navigate safely on public roads without colliding with obstacles (such as road signs and curbs).

[0082] In this embodiment, for the same perceived target perceived in the autonomous driving environment, such as car a traveling around the target vehicle, different sensors are used to obtain corresponding data related to car a, wherein there is a parallel relationship between the target data corresponding to each sensor, or the car a related data collected by the same sensor at different times has a projection relationship with the target vehicle. Therefore, in this embodiment, the first topological relationship of the observed target and the second topological relationship of the tracked target are calculated, and the position of the current perceived target is corrected based on these two types of topological relationships.

[0083] Step S104: Calculate the target correction displacement based on the first topological relationship and the second topological relationship, and perform position correction on the current position of the observed target based on the target correction displacement.

[0084] It should be noted that the target correction displacement in this embodiment is determined based on matching data pairs having similar relationships in the first topological relationship and the second topological relationship; the current position of the observed target can be obtained according to conventional target position acquisition methods in the art.

[0085] The target position correction method based on topological relationship in the embodiment of the present invention calculates the first topological relationship and the second topological relationship accordingly by constructing an observation topological matrix corresponding to the observation target and a tracking topological matrix corresponding to the tracking target, calculates the target correction displacement according to the first topological relationship and the second topological relationship, and performs position correction on the current position of the observation target based on the target correction displacement. This can improve the authenticity of the expression of the driving environment by the observation target, make the perceived target obtained by the current actual observation more real and reliable, help to achieve better reconstruction of the autonomous driving environment, and greatly meet the intelligent needs of vehicle autonomous driving.

[0086] In this embodiment, a target position correction method based on topological relationship is provided. Figure 2 FIG. 1 is a flow chart of another target position correction method based on topological relationship according to an embodiment of the present invention. Figure 2 As shown, the process includes the following steps:

[0087] Step S201, obtaining an observation target and a tracking target, wherein the observation target refers to at least one perception target included in the current road perception data of the target vehicle; the tracking target refers to at least one perception target included in the historical road perception data of the target vehicle.

[0088] Specifically, the above step S201 includes:

[0089] Step S2011, collecting the road environment information of the target vehicle at the current position through a preset sensor and obtaining the current collection time corresponding to the road environment information.

[0090] In this embodiment, the specific type of the preset sensor and the method for obtaining the current collection time are not limited here, and can be adaptively adjusted according to actual needs and conventional time acquisition methods in the field. For example, the preset sensors include cameras, lidars, etc.; the current time corresponding to the sensor collection data is recorded by a timer, which is only for illustration.

[0091] Step S2012, normalize the road environment information to obtain current road perception data, and determine the observation target from the perception targets included in the current road perception data.

[0092] In this embodiment, the purpose of the standard processing is to obtain standard structured data, and its specific content is not limited here, and it is adaptively set according to actual needs. For example, the standard processing is the preprocessing of data, including data denoising, outlier detection, data normalization and standardization, etc.; or directly outputting structured data through a trained data model, which is only an exemplary description.

[0093] Step S2013, reading the target sequence corresponding to the last acquisition time from the preset tracking list, and determining the tracking target based on the target sequence, wherein the preset tracking list is a time series table constructed based on the historical road perception data of the target vehicle.

[0094] It should be noted that, with regard to the collection time of road perception data, in theory, the road perception data at the current collection time corresponds to the current frame collection data, and the road perception data corresponding to the previous collection time corresponds to the previous frame collection data; if there is no jump between the current frame collection data and the previous frame collection data, then the road perception data corresponding to the previous collection time in this embodiment is the previous frame collection data; if there is a jump between the current frame collection data and the previous frame collection data, then the road perception data corresponding to the previous collection time in this embodiment is the road perception data (i.e., the target sequence) closest to the current collection time in the time series table constructed from the historical road perception data of the target vehicle.

[0095] Specifically, the process of constructing the preset tracking list in step S2013 includes:

[0096] Step A1, collecting historical road environment information of the target vehicle at different positions through a preset sensor and obtaining the historical collection time corresponding to the historical road environment information.

[0097] In this embodiment, the relevant contents of the preset sensor can be found in the above text and will not be repeated here.

[0098] Step A2: standardize all historical road environment information to obtain historical road perception data containing at least one perception target.

[0099] In this embodiment, the relevant contents of the standard processing can be found in the previous text and will not be repeated here.

[0100] Step A3: construct a preset tracking list based on each historical road perception data and its corresponding historical collection time.

[0101] In this embodiment, by standardizing the historical road environment information of the target vehicle at different positions collected by the preset sensor, and building a preset tracking list in combination with its corresponding historical collection time, the standardization of the data in the preset tracking list can be guaranteed, and data support can be provided for the position correction of the observed target.

[0102] In the embodiment of the present invention, various sensor types involved in the autonomous driving environment perception technology are taken into account. The current road perception data is obtained after standardized processing of the road environment information collected by the preset sensor, and the observation target is determined from the perception targets contained in the current road perception data; and the tracking target is obtained from the time series table constructed by the historical road perception data of the target vehicle according to the current collection time, which can ensure the diversity, standardization and accuracy of the acquisition of observation targets and tracking targets.

[0103] Step S202: construct an observation topology matrix corresponding to the observation target and a tracking topology matrix corresponding to the tracking target respectively.

[0104] Specifically, the above step S202 includes:

[0105] Step S2021, determining a first target number of observation targets, and constructing an observation topology matrix corresponding to the observation targets based on the first target number and current road perception data.

[0106] In this embodiment, since the observed target includes at least one perceived target included in the current road perception data, it is necessary to obtain the number and type corresponding to different perceived targets respectively to construct the observation topology matrix.

[0107] Step S2022: determine the number of second targets of the tracking target, and construct a tracking topology matrix corresponding to the tracking target based on the number of second targets and historical road perception data.

[0108] In this embodiment, since the tracking target includes at least one perception target included in the historical road perception data, it is also necessary to obtain the number and type of different perception targets respectively to construct a tracking topology matrix.

[0109] In the embodiment of the present invention, an observation topology matrix is ​​constructed according to the first target number of the observed targets and the current road perception data, and a tracking topology matrix is ​​constructed according to the second target number of the tracked targets and the historical road perception data. This can ensure the construction accuracy of the corresponding topology matrix, help provide data support for the calculation of subsequent topological relationships, and further ensure the accuracy of the target correction displacement.

[0110] In practical applications, for different types of sensors, there are certain spatial differences in the sensor data collected; for the same type of sensors, the time difference between the sensor data corresponding to different collection times is relatively obvious. Therefore, in order to ensure that the sensor data is processed at the same scale, this embodiment needs to synchronize the sensor data in time and space to improve the accuracy and efficiency of sensor data fusion, optimize the accuracy of data processing and algorithms, and enhance the reliability and stability of the subsequent vehicle automatic driving system. Specifically, before constructing the observation topology matrix corresponding to the observation target and the tracking topology matrix corresponding to the tracking target, the target position correction method based on the topological relationship of this embodiment also includes: performing preset synchronization processing on the observation target and the tracking target, and correspondingly obtaining the processed observation target and tracking target, wherein the preset synchronization processing includes time compensation processing and space synchronization processing. Specifically, by performing time compensation processing and space synchronization processing on the observation target and the tracking target, the observation target and the tracking target can be at the same time and space scale, which helps to improve the accuracy and effectiveness of the subsequent target position correction.

[0111] Step S203, respectively calculating a first topological relationship corresponding to the observation topological matrix and a second topological relationship corresponding to the tracking topological matrix.

[0112] It should be noted that since the autonomous driving environment perception technology involves a single sensor (i.e., a single sensor) and multiple sensors (i.e., multi-source heterogeneous sensors), the calculation of the topological relationship in this embodiment can be determined based on the number of perceived targets and the sensor type.

[0113] In this embodiment, the topological relationship includes point topological relationship, line topological relationship and triangle topological relationship. In the VCS coordinate system constructed with the target vehicle as the origin, the point topological relationship is used to detect the target projection relationship, that is, to perceive the slope of the line connecting the target (also called the target) and the origin. Figure 3 It should be noted that target 1 and target 2 in the figure can represent the same target data collected by different sensors; or target 1 and target 2 can represent the same target data collected by the same sensor at different collection times. Line topology is used to detect the parallel relationship of target lines. For line topology, please refer to Figure 4. It should be noted that target 1 and target 3 in the figure belong to two target data collected by the same sensor, target 2 and target 4 belong to two target data collected by the same sensor, and the slope of the line connecting target 1 and target 3 is parallel to the slope of the line connecting target 2 and target 4. The line connecting target 1 and target 3 and the line connecting target 2 and target 4 can represent the same target data collected by different sensors (that is, target 1 and target 2 are the same, and target 3 and target 4 are the same); or the line connecting target 1 and target 3 and the line connecting target 2 and target 4 are the same target data collected by the same sensor at different collection times (that is, target 1 and target 2 are the same, and target 3 and target 4 are the same). Triangular topological relationships are used to detect targets and construct triangular similarity relationships, see Figure 5 That is, the observed slopes of the lines connecting target 1, target 3 and target 5 are parallel to the slopes of the lines connecting target 2, target 4 and target 6 in the historical tracking records. The triangle formed by the observed targets 1, target 3 and target 5 is similar to the triangle formed by the historical tracking records of targets 2, target 4 and target 6.

[0114] In this embodiment, if the topological relationship is determined according to the number of targets, the above step S203 includes:

[0115] Step B1, determining a first target number of observed targets according to an observation topology matrix and determining a second target number of tracked targets according to a tracking topology matrix.

[0116] Step B2, if the number of first targets is 1, determine that the first topological relationship is a point topological relationship; if the number of first targets is 2, determine that the first topological relationship is a line topological relationship; if the number of first targets is at least 3, determine that the first topological relationship is a triangle topological relationship.

[0117] Step B3, if the number of second targets is 1, determine that the second topological relationship is a point topological relationship; if the number of second targets is 2, determine that the second topological relationship is a line topological relationship; if the number of second targets is at least 3, determine that the second topological relationship is a triangle topological relationship.

[0118] In the embodiment of the present invention, the corresponding first topological relationship is determined by observing the first target number of the target, and the corresponding second topological relationship is determined by tracking the second target number of the target, thereby realizing a reasonable division of the topological relationships of the current and historical corresponding environmental perception targets, which can improve the authenticity, accuracy and reliability of the perceived targets, making the current perceived targets more real and reliable, and thus realizing better reconstruction of the autonomous driving environment.

[0119] In this embodiment, if the topological relationship is determined according to the sensor type, the above step S203 includes:

[0120] Step C1, determining a first sensor type for observing a target according to an observation topology matrix, and determining a second sensor type for tracking a target according to a tracking topology matrix.

[0121] In this embodiment, the specific type of the first or second sensor may be determined based on current or historical road perception data of the target vehicle.

[0122] Step C2: if the first sensor type is a visual sensor, determine that the first topological relationship is a point topological relationship; if the first sensor type is a radar sensor, determine that the first topological relationship is a line topological relationship; if the first sensor type includes at least a visual sensor and a radar sensor, determine that the first topological relationship is a triangle topological relationship.

[0123] It should be noted that visual sensors are visual perception methods, which mainly rely on cameras to obtain image information and then process it through computer vision algorithms. This method is very effective for identifying targets such as road signs, pedestrians, and vehicles. Radar sensors are radar perception methods and laser radar (LiDAR) perception methods. Among them, the radar perception method uses radio waves for detection, which can penetrate bad weather such as rain, snow, and fog, so it is more reliable than visual perception in these cases; radar can obtain information such as the distance, speed, and angle of an object, which is very effective for detecting and tracking obstacles around the vehicle. LiDAR perception obtains the precise distance and shape information of an object by emitting a laser beam and measuring the time it takes for it to reflect back. The perception accuracy is very high and can be used to build high-precision maps and obstacle models around the vehicle; however, the cost of LiDAR equipment is relatively high and is less affected by weather conditions, but it may still be subject to certain restrictions in bad weather.

[0124] Step C3, if the second sensor type is a visual sensor, determine that the second topological relationship is a point topological relationship; if the second sensor type is a radar sensor, determine that the second topological relationship is a line topological relationship; if the second sensor type includes at least a visual sensor and a radar sensor, determine that the second topological relationship is a triangle topological relationship.

[0125] In the embodiment of the present invention, the characteristics of different sensors in the actual autonomous driving environment perception scenario are taken into account, and the topological relationship between the current and historical corresponding environmental perception targets is determined by the type of sensor, which can ensure the rationality of the topological relationship calculation, thereby improving the accuracy of the perceived target, enhancing the perception target's true expression of the actual driving environment, and meeting the intelligent needs of vehicle autonomous driving.

[0126] Step S204: Calculate the target correction displacement based on the first topological relationship and the second topological relationship, and perform position correction on the current position of the observed target based on the target correction displacement.

[0127] Specifically, in the above step S204, calculating the target correction displacement based on the first topological relationship and the second topological relationship includes:

[0128] Step D1, calculating the best matching pair corresponding to the first topological relationship and the second topological relationship, wherein the best matching pair includes an observed target and a tracked target.

[0129] Specifically, the above step D1 includes:

[0130] Step D11 , performing similarity matching on the first topological relationship and the second topological relationship, and correspondingly obtaining a matching result including at least one matching pair.

[0131] In this embodiment, the specific evaluation index of similarity matching is not limited here, and is adaptively determined according to actual needs, such as Euclidean distance, cosine similarity, Jaccard similarity coefficient, etc.

[0132] Step D12: determining a preset matching number of the optimal matching pair according to the first topological relationship and the second topological relationship.

[0133] In this embodiment, the preset matching numbers of the optimal matching pair include 1, 2 and 3, and the specific number values ​​are determined according to the point topological relationship, line topological relationship and triangle topological relationship respectively.

[0134] Step D13, based on the preset number of matches, a corresponding number of similar matching pairs are screened from the matching results to obtain the best matching pair.

[0135] The embodiments of the present invention target the three types of topological relationships designed for the perceived targets, including point topological relationships, line topological relationships, and triangle topological relationships. By screening the optimal matching pairs according to the corresponding set number of matches, the calculation accuracy of the optimal matching pairs can be guaranteed, thereby improving the subsequent correction accuracy of the current position of the observed target, making the observed target more real and reliable, and contributing to a better reconstruction of the autonomous driving environment.

[0136] Step D2, calculating the target correction displacement based on the optimal matching pair.

[0137] Specifically, the above step D2 includes:

[0138] In step D21, taking the target vehicle as the origin, the position coordinates of each optimal matching pair are calculated respectively.

[0139] In this embodiment, the position coordinates of the observed target and the position coordinates of the tracked target are calculated based on the current road perception data and the historical road perception data of the target vehicle.

[0140] Step D22: If the number of the best matching pair is 1, the target correction displacement is calculated according to the position coordinates of the observed target and the position coordinates of the tracked target in the best matching pair.

[0141] In this embodiment, the specific calculation of the target correction displacement is determined according to a conventional displacement coordinate calculation formula in the art.

[0142] Step D23, if the number of the best matching pairs is 2 or 3, the corresponding displacements are calculated according to the position coordinates of the observed target and the position coordinates of the tracked target in each best matching pair, and the average calculation result of all displacements is determined as the target correction displacement.

[0143] In this embodiment, the target correction displacement corresponding to the current position of the observed target is calculated according to the number of optimal matching pairs, which can improve the accuracy of the target correction displacement and make the current position of the observed target more consistent with the actual driving environment, which helps to improve the reconstruction effect of the autonomous driving environment and the user experience.

[0144] In the embodiment of the present invention, by calculating the optimal matching pair corresponding to the first topological relationship and the second topological relationship, and determining the target correction displacement based on the optimal matching pair, the observed target achieves the optimal expression of the real driving environment, ensuring the accuracy of the current perceived target position, which is helpful to subsequently achieve better reconstruction of the autonomous driving environment and enhance the user's car experience.

[0145] It should be noted that the actual process of performing position correction on the current position of the observed target based on the target correction displacement in this embodiment is to add the target correction displacement to the current position of the observed target to achieve correction of the target position.

[0146] In practical applications, the target vehicle collects different environmental information through various sensors carried, and uploads the collected information to the cloud to achieve different autonomous driving tasks. However, due to differences in position, angle and accuracy of different sensors, it is necessary to eliminate these differences through sensor data calibration, so that the data of different sensors can be fused in a unified coordinate system to improve the accuracy and stability of vehicle autonomous driving. Therefore, a target position correction system based on the target position topological structure is proposed in this embodiment, which can be integrated on the vehicle side to achieve position correction of the perceived target. Specifically, the system uses the similarity of the topological structure of the significant target in the scene to clarify the significant target matching pair, and estimates the horizontal and vertical correction displacement of the target position to solve the position ranging anomaly problem introduced by the sensor feature, which can not only optimize the overall longitudinal position jump of the continuous frame target caused by the whole vehicle posture of the single sensor vision, and the overall lateral position drift caused by the turning of the radar, but also solve the problem of missed association caused by the large target position error of multi-source heterogeneous sensors, enhance the sensor properties of the associated fusion target, and more accurately express the environmental authenticity.

[0147] In this embodiment, a target position correction system based on a target position topological structure includes:

[0148] A road information collection unit is used to collect road environment data and report it to the processing system, where the road environment data includes basic information such as lanes, lane lines, vehicle targets, pedestrian targets, road traffic signs, traffic flows, traffic lights and intersection instructions;

[0149] A road information receiving unit, used to receive road environment information reported by each sensor;

[0150] A road information cache unit, used to cache historical road environment data of each sensor;

[0151] The target topology structure building unit is used to calculate the respective topological matrices between the observed target and the tracked target.

[0152] A salient target selection unit is used to select salient targets based on a topological matrix using point, line and triangle topological relationships, wherein target information of the salient targets includes target tracking ID, position, speed, acceleration, heading angle, motion state and other basic information;

[0153] The target correction displacement calculation unit is used to calculate the target correction displacement corresponding to the significant target, wherein the target correction displacement includes the lateral and longitudinal offset displacements of the single sensor tracking target and the fusion target.

[0154] In a specific embodiment, for the target position correction system based on the target position topological structure in the above-mentioned embodiment, a target position correction scheme is designed accordingly, and the specific process includes the following steps:

[0155] Step 1: Collect images / reflected waves of the area of ​​interest through general autonomous driving sensors such as forward / surrounding cameras, millimeter wave / lidar, and quantify them into structured information.

[0156] Step 2: Receive quantified structured information reported by multi-source heterogeneous sensors, such as dynamic and static target positions / speeds, lane lengths / curvatures, turn / speed limit road signs, etc., and cache them to form perception history data with time series.

[0157] Step 3: Estimate three types of topological relationships of the observed target position and construct the observed target topological relationship matrix. Specifically, estimate three types of topological relationships of the tracking or fusion target position and construct the tracking or fusion target topological relationship matrix.

[0158] Step 4: Estimate the similarity probabilities of the triangle relationships, line relationships, and point relationships in the observed target topological relationship matrix and the tracking or fusion target topological relationship matrix, and traverse the matching pairs with the greatest similarity.

[0159] Step 5: Combine the best significant matching pair position to compensate for the horizontal and vertical correction distances of the observed target.

[0160] It should be noted that in this embodiment, the topological relationship of points, lines and triangles is determined according to the sensor type and the number of targets. For a single type of sensor, such as a visual sensor or a radar sensor, the corresponding targets can be called visual targets or radar targets; for multiple types of sensors, the corresponding targets can be called associated targets.

[0161] In a specific embodiment, the target position correction method based on topological relationship as in this embodiment is used to correct the visual target position (that is, the visual target obtained by the visual sensor is corrected for the target longitudinal position based on the topological relationship of the target position points). Figure 6 It is a flowchart of visual target position correction.

[0162] Depend on Figure 6 It can be seen that the specific correction includes the following steps:

[0163] Step 1: Camera target data collection.

[0164] In this embodiment, road image data is acquired through a visual camera, and structured road information is output from a trained neural network model, wherein the structured road target information includes target position, speed, tracking ID, target type, and heading angle, etc.

[0165] Step 2: Construct the topological relationship matrix of the observed target points.

[0166] In this embodiment, for the number of observed targets N, the point topological relationship is calculated as (k0, k1, k2, ..., kn-1). It should be noted that due to the visual occlusion phenomenon, the point projection relationship has obvious characteristics.

[0167] Step 3: Track target data acquisition.

[0168] In this embodiment, the target set closest to the current time in the tracking list is obtained. Specifically, the latest tracking target set in the tracking list is obtained by using the observed target time.

[0169] Step 4: Time / space synchronization.

[0170] In this embodiment, the tracked target is subjected to time-space synchronization processing. Specifically, due to the time difference between the observed target and the tracked target, the observed time and the vehicle information are used to perform time-space synchronization processing on the tracked target.

[0171] Step 5: Construct the topological relationship matrix of the tracking target points.

[0172] In this embodiment, for the number of tracking targets M, the calculation point topological relationship is (p0, p1, p2, ..., pm-1).

[0173] Step 6: Select the optimal matching pair of target points.

[0174] In this embodiment, the best matching pair of target point relationship is the closest matching pair of target point relationship. Specifically, the Euclidean distance between the corresponding elements of the observed target point relationship and the tracked target point relationship is calculated, and the matching pair with the smallest Euclidean distance is selected as the best matching pair of target point relationship.

[0175] Step 7: Calculate the visual target correction displacement.

[0176] In this embodiment, the visual target time sequence matching is combined with the target position to calculate the visual target correction displacement including the horizontal and vertical directions.

[0177] In a specific embodiment, the radar target position is corrected by using the target position correction method based on the topological relationship of the present embodiment (i.e., the radar target obtained by the radar sensor is corrected for the target lateral position based on the topological relationship of the target position line). Figure 7 It is a flowchart of radar target position correction.

[0178] Depend on Figure 7 It can be seen that the specific correction includes the following steps:

[0179] Step 1: Radar target data collection.

[0180] In this embodiment, a road reflection point cloud is acquired by radar, and structured road information is generated through post-processing, wherein the structured road information includes target position, speed, acceleration, tracking ID, heading angle, etc.

[0181] Step 2: Construct the topological relationship matrix of the observation target line.

[0182] In this embodiment, for the number of observed targets N, the calculated line topological relationship is (k01, k02, k03, ..., kn1, kn2, ..., knn-1). Specifically, due to radar splitting, the line parallel relationship feature is obvious.

[0183] Step 3: Track target data acquisition.

[0184] In this embodiment, the target set closest to the current time in the tracking list is obtained. Specifically, the latest tracking target set in the tracking list is obtained by using the observed target time.

[0185] Step 4: Time / space synchronization.

[0186] In this embodiment, the step 4 may refer to the relevant content above and will not be repeated here.

[0187] Step 5: Construct the topological relationship matrix of the tracking target line.

[0188] In this embodiment, for the number of tracking targets M, the calculation line topology relationship is (p01, p02, p03, ..., pm1, pm2, ..., pmm-1).

[0189] Step 6: Select the best matching pair of line topology relationships.

[0190] In this embodiment, the optimal matching pair of line topology relationship is to select the matching pair with the closest line topology relationship. Specifically, the Euclidean distance of the corresponding elements in the line topology relationship matrix of the observed target and the line topology relationship matrix of the tracked target is calculated, and the matching pair with the smallest Euclidean distance is selected as the optimal matching pair of line topology relationship.

[0191] Step 7: Calculate the radar target correction displacement.

[0192] In this embodiment, the radar target correction displacement including the horizontal and vertical directions is calculated in combination with the radar target time sequence matching to the target position.

[0193] In a specific embodiment, the target position correction method based on topological relationship as in this embodiment is used to correct the associated target position (that is, the associated target obtained by the heterogeneous sensor is corrected based on the target position point, line or triangle topological relationship). Figure 8 is a flowchart of the associated target position correction. Figure 8It can be seen that the specific correction includes the following steps:

[0194] Step 1: Sensor target data collection.

[0195] In this embodiment, structured road information is acquired through various sensors, wherein the structured road target information includes target position, speed, tracking ID, target type, heading angle, etc.

[0196] Step 2: Time / space synchronization.

[0197] In this embodiment, the step 2 may refer to the relevant content above and will not be repeated here.

[0198] Step 3: Construct the topological relationship matrix of observation target points, lines and triangles.

[0199] In this embodiment, for the number of observed targets N, the point topological relationship is calculated as (k0, k1, k2, …, kn-1); the line topological relationship is calculated as (k01, k02, k03, …, kn1, kn2, …, knn-1); and the triangle topological relationship is calculated as (1[ka, kb, kc], 2[ka, kb, kc], …(n-1)[ka, kb, kc], n[ka, kb, kc]).

[0200] Step 4: Acquire fusion tracking target data.

[0201] In this embodiment, the target sequence closest to the current time in the tracking list is obtained. Specifically, the latest tracking target set in the tracking list is obtained by associating and fusing the target time points.

[0202] Step 5: Time / space synchronization.

[0203] In this embodiment, since there is a time difference in the fusion triggering of the tracking target, it is necessary to use the fusion time point and the vehicle information to perform spatiotemporal synchronization processing on the tracking target. The relevant content of the spatiotemporal synchronization processing can be found in the previous text and will not be repeated here.

[0204] Step 6: Construct the tracking target point, line and triangle topological relationship matrix.

[0205] In this embodiment, for the number of tracking targets M, the point topological relationship is calculated as (k0, k1, k2, …, km-1); the line topological relationship is calculated as (k01, k02, k03, …, km1, km2, … kmm-1); and the triangle topological relationship is calculated as (1[ka, kb, kc], 2[ka, kb, kc], …(m-1)[ka, kb, kc], m[ka, kb, kc]).

[0206] Step 7: Select the best matching pair of points, lines or triangle topological relationships.

[0207] In this embodiment, the Euclidean distances of the corresponding elements in the observed target point, line and triangle topological relationship matrix and the tracked target point, line and triangle topological relationship matrix are calculated respectively, and the matching pairs with the smallest Euclidean distance are selected as the optimal matching pairs of the corresponding topological relationship. It should be noted that the specific type of topological relationship in this embodiment is determined according to the number of targets.

[0208] Step 8: Calculate the sensor target correction displacement.

[0209] In this embodiment, the sensor target correction displacement is the associated target correction displacement, and the target correction displacement of the associated target is calculated by combining the observed target and the fused tracking target matching pair target positions.

[0210] In summary, the target position correction method based on topological relationship in the embodiment of the present invention has the following advantages:

[0211] 1. For the visual targets of the visual sensor, the longitudinal position anomaly caused by the vehicle posture or sloped road can be suppressed based on the target position topology structure, which not only increases the tracking stability of the visual target, but also improves the continuity of target detection;

[0212] 2. For radar targets of radar sensors, the radar target can be suppressed from drifting in the lateral position of the curve based on the target position topology, which not only enhances the accuracy of the radar target, but also increases the probability of its association with other sensors.

[0213] 3. For associated targets, it can make the observed targets of heterogeneous sensors more likely to be associated with the tracked targets, which not only enhances the optimal attributes of the associated targets to express the authenticity of the environment, but also better serves various vehicle control needs.

[0214] In the present embodiment, a target position correction device based on a topological relationship is also provided, and the device is used to implement the above-mentioned embodiments and preferred implementation modes, and the descriptions that have been made are not repeated here. As used below, the term "module" refers to a combination of software and / or hardware that can implement a predetermined function. Although the device described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0215] The present invention provides a target position correction device based on topological relationship, such as Fig. 9 As shown, the device comprises:

[0216] The acquisition module 901 is used to acquire an observation target and a tracking target, wherein the observation target refers to at least one perception target included in the current road perception data of the target vehicle; the tracking target refers to at least one perception target included in the historical road perception data of the target vehicle.

[0217] The construction module 902 is used to respectively construct an observation topology matrix corresponding to the observation target and a tracking topology matrix corresponding to the tracking target.

[0218] The calculation module 903 is used to calculate the first topological relationship corresponding to the observation topological matrix and the second topological relationship corresponding to the tracking topological matrix respectively.

[0219] The correction module 904 is used to calculate the target correction displacement based on the first topological relationship and the second topological relationship, and perform position correction on the current position of the observed target based on the target correction displacement.

[0220] In some optional embodiments, the acquisition module 901 includes: a first acquisition submodule, a second acquisition submodule and a third acquisition submodule; wherein the first acquisition submodule is used to collect road environment information of the target vehicle at the current position through a preset sensor and obtain the current acquisition time corresponding to the road environment information; the second acquisition submodule is used to standardize the road environment information, obtain the current road perception data, and determine the observation target from the perception targets contained in the current road perception data; the third acquisition submodule is used to read the target sequence corresponding to the previous acquisition time from the preset tracking list, and determine the tracking target based on the target sequence, wherein the preset tracking list is a time series table constructed based on the historical road perception data of the target vehicle.

[0221] In some optional embodiments, the third acquisition submodule includes: a first acquisition unit, a second acquisition unit and a third acquisition unit; wherein the first acquisition unit is used to collect historical road environment information of the target vehicle at different positions through a preset sensor and obtain the historical collection time corresponding to the historical road environment information; the second acquisition unit is used to perform standardized processing on all historical road environment information to obtain historical road perception data containing at least one perception target; the third acquisition unit is used to construct a preset tracking list based on each historical road perception data and its corresponding historical collection time.

[0222] In some optional embodiments, the construction module 902 includes: a first construction submodule and a second construction submodule; wherein the first construction submodule is used to determine the first target number of the observed targets, and construct an observation topology matrix corresponding to the observed targets based on the first target number and the current road perception data; the second construction submodule is used to determine the second target number of the tracked targets, and construct a tracking topology matrix corresponding to the tracked targets based on the second target number and the historical road perception data.

[0223] In some optional embodiments, the calculation module 903 includes: a first calculation submodule, a second calculation submodule and a third calculation submodule; wherein the first calculation submodule is used to determine the first target number of the observed target according to the observation topology matrix and the second target number of the tracked target according to the tracking topology matrix; the second calculation submodule is used to determine that the first topological relationship is a point topological relationship if the first target number is 1; if the first target number is 2, the first topological relationship is determined to be a line topological relationship; if the first target number is at least 3, the first topological relationship is determined to be a triangle topological relationship; the third calculation submodule is used to determine that the second topological relationship is a point topological relationship if the second target number is 1; if the second target number is 2, the second topological relationship is determined to be a line topological relationship; if the second target number is at least 3, the second topological relationship is determined to be a triangle topological relationship.

[0224] In some optional embodiments, the calculation module 903 includes: a first solution submodule, a second solution submodule and a third solution submodule; wherein the first solution submodule is used to determine the first sensor type of the observed target according to the observation topology matrix, and to determine the second sensor type of the tracked target according to the tracking topology matrix; the second solution submodule is used to determine that the first topological relationship is a point topological relationship if the first sensor type is a visual sensor; if the first sensor type is a radar sensor, the first topological relationship is determined to be a line topological relationship; if the first sensor type includes at least a visual sensor and a radar sensor, the first topological relationship is determined to be a triangle topological relationship; the third solution submodule is used to determine that the second topological relationship is a point topological relationship if the second sensor type is a visual sensor; if the second sensor type is a radar sensor, the second topological relationship is determined to be a line topological relationship; if the second sensor type includes at least a visual sensor and a radar sensor, the second topological relationship is determined to be a triangle topological relationship.

[0225] In some optional embodiments, the correction module 904 includes: a first correction submodule and a second correction submodule; wherein the first correction submodule is used to calculate the optimal matching pair corresponding to the first topological relationship and the second topological relationship, wherein the optimal matching pair includes the observed target and the tracked target; and the second correction submodule is used to calculate the target correction displacement based on the optimal matching pair.

[0226] In some optional embodiments, the first correction submodule includes: a first correction unit, a second correction unit and a third correction unit; wherein the first correction unit is used to perform similarity matching on the first topological relationship and the second topological relationship, and obtain a corresponding matching result containing at least one matching pair; the second correction unit is used to determine a preset number of matches for the optimal matching pair based on the first topological relationship and the second topological relationship; and the third correction unit is used to screen a corresponding number of similar matching pairs from the matching results based on the preset number of matches to obtain the optimal matching pair.

[0227] In some optional embodiments, the second correction submodule includes: a first correction unit, a second correction unit and a third correction unit; wherein the first correction unit is used to calculate the position coordinates of each optimal matching pair with the target vehicle as the origin; the second correction unit is used to calculate the target correction displacement according to the position coordinates of the observed target and the position coordinates of the tracked target in the optimal matching pair if the number of the optimal matching pairs is 1; the third correction unit is used to calculate the corresponding displacement according to the position coordinates of the observed target and the position coordinates of the tracked target in each optimal matching pair if the number of the optimal matching pairs is 2 or 3, and determine the average calculation result of all displacements as the target correction displacement.

[0228] In some optional embodiments, the device further includes: a synchronization module, which is used to perform preset synchronization processing on the observation target and the tracking target, and obtain the processed observation target and tracking target accordingly, wherein the preset synchronization processing includes time compensation processing and space synchronization processing.

[0229] The further functional description of each of the above modules is the same as that of the above corresponding embodiments and will not be repeated here.

[0230] The target position correction device based on topological relationship in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0231] The target position correction device based on topological relationship in the embodiment of the present invention corrects the position of the target by constructing a topological relationship corresponding to the observed target and the tracked target, which can improve the realistic expression of the observed target on the driving environment, making the observed target more realistic and reliable, which not only improves the effect of reconstructing the actual autonomous driving environment, but also meets the intelligent needs of vehicle autonomous driving.

[0232] A vehicle is also provided in an embodiment of the present invention, and the vehicle includes a controller. The controller in this embodiment is a vehicle controller, which is used to perform operations such as power supply / power off, sleep and wake up of the sub-controllers and network nodes under it, and each power supply interface thereof can collect the real-time current output. Other controllers with the above functions are applicable.

[0233] Fig.10 is a schematic diagram of the structure of the controller provided in an optional embodiment of the present invention, such as Fig.10 As shown, the controller includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. The various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process instructions executed in the controller, including instructions stored in or on the memory to display graphical information of the GUI on an external input / output device (such as a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple controllers can be connected, and each controller provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Fig.10 A processor 10 is taken as an example.

[0234] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.

[0235] The memory 20 stores instructions executable by at least one processor 10, so that at least one processor 10 executes the method shown in the above embodiment.

[0236] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required by at least one function; the data storage area may store data created according to the use of the controller, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the controller via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0237] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.

[0238] The controller also includes a communication interface 30 for the main control chip to communicate with other devices or a communication network.

[0239] A computer-readable storage medium is also provided in an embodiment of the present invention. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium and downloaded through a network, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor main control chip or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor or hardware, the method shown in the above embodiment is implemented.

[0240] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A target position correction method based on topological relationship, characterized in that: The method comprises: Acquire an observation target and a tracking target, wherein the observation target refers to at least one perception target included in the current road perception data of the target vehicle; the tracking target refers to at least one perception target included in the historical road perception data of the target vehicle; Constructing an observation topology matrix corresponding to the observation target and a tracking topology matrix corresponding to the tracking target respectively; Respectively calculating a first topological relationship corresponding to the observation topological matrix and a second topological relationship corresponding to the tracking topological matrix; A target correction displacement is calculated based on the first topological relationship and the second topological relationship, and a position correction is performed on a current position of the observed target based on the target correction displacement.

2. The target position correction method based on topological relationship according to claim 1 is characterized in that: The respectively constructing the observation topology matrix corresponding to the observation target and the tracking topology matrix corresponding to the tracking target comprises: Determine a first target number of observation targets, and construct an observation topology matrix corresponding to the observation targets based on the first target number and current road perception data; A second target number of the tracking target is determined, and a tracking topology matrix corresponding to the tracking target is constructed based on the second target number and historical road perception data.

3. The target position correction method based on topological relationship according to claim 1, characterized in that: The respectively calculating the first topological relationship corresponding to the observation topological matrix and the second topological relationship corresponding to the tracking topological matrix includes: Determine the first target number of the observed targets according to the observation topology matrix and determine the second target number of the tracked targets according to the tracking topology matrix; If the first target number is 1, the first topological relationship is determined to be a point topological relationship; if the first target number is 2, the first topological relationship is determined to be a line topological relationship; if the first target number is at least 3, the first topological relationship is determined to be a triangle topological relationship; If the number of the second targets is 1, the second topological relationship is determined to be a point topological relationship; if the number of the second targets is 2, the second topological relationship is determined to be a line topological relationship; if the number of the second targets is at least 3, the second topological relationship is determined to be a triangle topological relationship.

4. The target position correction method based on topological relationship according to claim 1, characterized in that: The respectively calculating the first topological relationship corresponding to the observation topological matrix and the second topological relationship corresponding to the tracking topological matrix includes: Determine a first sensor type for observing a target according to an observation topology matrix, and determine a second sensor type for tracking a target according to a tracking topology matrix; If the first sensor type is a visual sensor, the first topological relationship is determined to be a point topological relationship; if the first sensor type is a radar sensor, the first topological relationship is determined to be a line topological relationship; if the first sensor type includes at least a visual sensor and a radar sensor, the first topological relationship is determined to be a triangle topological relationship; If the second sensor type is a visual sensor, the second topological relationship is determined to be a point topological relationship; if the second sensor type is a radar sensor, the second topological relationship is determined to be a line topological relationship; if the second sensor type includes at least a visual sensor and a radar sensor, the second topological relationship is determined to be a triangle topological relationship.

5. The target position correction method based on topological relationship according to any one of claims 1 to 4, characterized in that: The calculating the target correction displacement based on the first topological relationship and the second topological relationship includes: Calculating an optimal matching pair corresponding to the first topological relationship and the second topological relationship, wherein the optimal matching pair includes an observed target and a tracked target; A target correction displacement is calculated based on the best matching pair.

6. The target position correction method based on topological relationship according to claim 5, characterized in that: The calculating the optimal matching pair corresponding to the first topological relationship and the second topological relationship includes: Performing similarity matching on the first topological relationship and the second topological relationship, and correspondingly obtaining a matching result including at least one matching pair; Determine a preset matching number of an optimal matching pair according to the first topological relationship and the second topological relationship; Based on the preset number of matches, a corresponding number of similar matching pairs are screened from the matching results to obtain the best matching pair.

7. The target position correction method based on topological relationship according to claim 6 is characterized in that: The calculating the target correction displacement based on the optimal matching pair comprises: Taking the target vehicle as the origin, calculate the position coordinates of each optimal matching pair; If the number of the optimal matching pairs is 1, the target correction displacement is calculated according to the position coordinates of the observed target and the position coordinates of the tracked target in the optimal matching pairs; If the number of the optimal matching pairs is 2 or 3, the corresponding displacements are calculated according to the position coordinates of the observed target and the position coordinates of the tracked target in each optimal matching pair, and the average calculation result of all the displacements is determined as the target correction displacement.

8. The target position correction method based on topological relationship according to claim 1, characterized in that: The obtaining of the observation target and the tracking target comprises: Collecting the road environment information of the target vehicle at the current position through a preset sensor and obtaining the current collection time corresponding to the road environment information; Performing standard processing on the road environment information to obtain current road perception data, and determining an observation target from the perception targets included in the current road perception data; A target sequence corresponding to a previous acquisition time is read from a preset tracking list, and a tracking target is determined based on the target sequence, wherein the preset tracking list is a time series table constructed based on historical road perception data of the target vehicle.

9. The target position correction method based on topological relationship according to claim 8, characterized in that: The process of constructing the preset tracking list includes: Collect historical road environment information of the target vehicle at different locations through a preset sensor and obtain the historical collection time corresponding to the historical road environment information; All historical road environment information is processed in a standardized manner, and corresponding historical road perception data including at least one perception target is obtained; A preset tracking list is constructed based on each historical road perception data and its corresponding historical collection time.

10. The target position correction method based on topological relationship according to claim 1, characterized in that: Before respectively constructing the observation topology matrix corresponding to the observation target and the tracking topology matrix corresponding to the tracking target, the method further includes: Preset synchronization processing is performed on the observation target and the tracking target to obtain the processed observation target and the tracking target, wherein the preset synchronization processing includes time compensation processing and space synchronization processing.

11. A target position correction device based on topological relationship, characterized in that: The device comprises: An acquisition module, used to acquire an observation target and a tracking target, wherein the observation target refers to at least one perception target included in the current road perception data of the target vehicle; the tracking target refers to at least one perception target included in the historical road perception data of the target vehicle; A construction module, used to respectively construct an observation topology matrix corresponding to the observation target and a tracking topology matrix corresponding to the tracking target; A calculation module, used to respectively calculate a first topological relationship corresponding to the observation topological matrix and a second topological relationship corresponding to the tracking topological matrix; A correction module is used to calculate a target correction displacement based on the first topological relationship and the second topological relationship, and to perform position correction on a current position of the observed target based on the target correction displacement.

12. A vehicle, characterized in that: The vehicle includes a controller, which includes: a memory and a processor, the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the target position correction method based on topological relationship described in any one of claims 1 to 10 by executing the computer instructions.

13. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the target position correction method based on topological relationship according to any one of claims 1 to 10.

Images