Tracking target trajectory optimization processing method and device

Abstract

The invention provides a tracking target trajectory optimization processing method and device. The method comprises the following steps: acquiring a target point cloud of a tracking object in a data frame collected by a laser radar; performing reconstruction processing on the target point cloud to obtain a reconstructed point cloud model; performing target frame size estimation processing based on the point cloud model to obtain a target frame of a target size; and performing tracking target trajectory optimization processing based on the target frame of the target size to obtain an optimized target trajectory. According to the tracking target trajectory optimization processing method provided by the invention, the accuracy and reliability of target tracking and labeling can be effectively improved, so that the automatic labeling effect of the tracking target trajectory is improved.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving technology, and in particular to a method and device for optimizing target trajectory tracking, and also to an electronic device and a non-transitory computer-readable storage medium. Background Art

[0002] In recent years, with the continuous development of machine vision technology, the application of autonomous driving technology has become more and more mature. The rapid development of autonomous driving technology has put forward higher requirements for high-precision target detection and tracking optimization technology. However, the existing automatic labeling system tracking target trajectory optimization solution has high limitations, resulting in poor accuracy and reliability of target tracking and labeling. Therefore, how to design a more efficient tracking target trajectory optimization processing solution has become a technical problem that needs to be solved urgently. Summary of the invention

[0003] The present invention provides a tracking target trajectory optimization processing method to solve the defect that the automatic labeling system tracking target trajectory optimization scheme in the prior art has high limitations, resulting in poor accuracy and reliability of target tracking and labeling.

[0004] The present invention provides a tracking target trajectory optimization processing method, comprising: Obtain the target point cloud of the tracked object in the data frame collected by the laser radar; Reconstructing the target point cloud to obtain a reconstructed point cloud model; Performing target frame size estimation processing based on the point cloud model to obtain a target frame of the target size; The target frame of the target size is used to optimize the tracking target trajectory to obtain an optimized target trajectory.

[0005] According to the tracking target trajectory optimization processing method of the present invention, the target point cloud is reconstructed to obtain a reconstructed point cloud model, including: Based on the different observation positions of the tracking target, construct a corresponding tracking target local coordinate system; Based on the local coordinate system of the tracked target, the target point clouds of the tracked target in continuous data frames are overlaid to obtain a reconstructed point cloud model.

[0006] According to the tracking target trajectory optimization processing method of the present invention, the corresponding tracking target local coordinate system is constructed based on the different observation positions of the tracking target, including: In the case where one edge of the tracking target is observed based on the preset laser radar, the center point of the edge is used as the origin of the local coordinate system to construct a corresponding first tracking target local coordinate system; in the case where two edges of the tracking target are observed based on the preset laser radar, the intersection formed by the two edges is used as the origin of the local coordinate system to construct a corresponding second tracking target local coordinate system; in the case where three edges of the tracking target are observed based on the preset laser radar, the center point of the longest edge among the three edges is used as the origin of the local coordinate system to construct a corresponding third tracking target local coordinate system; The step of performing superposition processing on the target point clouds of the tracked target in continuous data frames based on the local coordinate system of the tracked target to obtain a reconstructed point cloud model specifically includes: Based on the first tracking target local coordinate system, the target point clouds of the tracking target in continuous data frames are superimposed to obtain a reconstructed point cloud model; based on the second tracking target local coordinate system, the target point clouds of the tracking target in continuous data frames are superimposed to obtain a reconstructed point cloud model; based on the third tracking target local coordinate system, the target point clouds of the tracking target in continuous data frames are superimposed to obtain a reconstructed point cloud model.

[0007] According to the tracking target trajectory optimization processing method of the present invention, the target frame size estimation processing is performed based on the point cloud model to obtain the target frame of the target size, including: In the case where an edge of the tracking target is observed based on a preset laser radar, a length observation value corresponding to the edge is obtained; the length of another edge that is not observed in all target frames collected within a preset time period of the laser radar is counted to obtain a statistical result; and according to the statistical result, the median of the length of another edge that is not observed is used as a length estimation value; according to the length observation value and the length estimation value, a target frame of the target size is obtained; In the case where two edges of the tracking target are observed based on the preset laser radar, the maximum absolute value in the x-axis direction and the maximum absolute value in the y-axis direction in the local coordinate system of the second tracking target are obtained; and the longest value of the edge length of all target frames in the x-axis direction and the longest value of the other edge length of all target frames in the y-axis direction are obtained; a first distance difference between the maximum absolute value in the x-axis direction and the longest value of the edge length of all target frames in the x-axis direction in the local coordinate system of the second tracking target is determined; a second distance difference between the maximum absolute value in the y-axis direction and the longest value of the other edge length of all target frames in the y-axis direction is determined; when both the first distance difference and the second distance difference are less than the corresponding preset distance In the case of a threshold, it is determined that the point cloud model formed by the currently observed target point cloud meets the preset completeness observation condition, and the maximum absolute value in the x-axis direction and the maximum absolute value in the y-axis direction in the local coordinate system of the second tracking target are respectively used as the length estimation values ​​of the two edges of the target frame, and the target frame of the target size is obtained; in the case that the first distance difference and the second distance difference are greater than or equal to the preset distance threshold, it is determined that the point cloud model formed by the currently observed target point cloud meets the preset completeness observation condition, and the longest value of the edge length of all target frames in the x-axis direction and the longest value of the other edge length of all target frames in the y-axis direction are respectively used as the length estimation values ​​of the two edges of the target frame, and the target frame of the target size is obtained; In the case where three edges of the tracking target are observed based on a preset laser radar, a maximum absolute value and a minimum absolute value in the x-axis direction in the local coordinate system of the third tracking target are obtained, and a first difference between the maximum absolute value and the minimum absolute value in the x-axis direction is determined; and a maximum absolute value and a minimum absolute value in the y-axis direction in the local coordinate system of the third tracking target are obtained, and a second difference between the maximum absolute value and the minimum absolute value in the x-axis direction is determined; the first difference and the second difference are respectively used as length estimation values ​​of two edges of the target frame to obtain a target frame of the target size; The target frame is a rectangular frame consisting of edges of different lengths.

[0008] According to the tracking target trajectory optimization processing method of the present invention, the step of obtaining the target point cloud of the tracking object in the data frame collected by the laser radar specifically includes: Acquire a data frame collected by the laser radar, and extract a target point cloud in a target frame corresponding to the tracking target from the data frame; Motion compensation is performed on the target point cloud to obtain a target point cloud after motion compensation, and the target point cloud after motion compensation is used as the target point cloud of the tracking object in the data frame collected by the laser radar.

[0009] According to the tracking target trajectory optimization processing method of the present invention, the tracking target trajectory optimization processing is performed on the target frame based on the target size to obtain the optimized target trajectory, specifically including: According to the target size of the target frame, the center point of the target frame in the corresponding tracking target local coordinate system is determined, and the center point of the target frame in the corresponding tracking target local coordinate system is back-projected to the global coordinate system of the laser radar to obtain the optimized target trajectory.

[0010] The present invention also provides a tracking target trajectory optimization processing device, comprising: A target point cloud acquisition unit, used to acquire a target point cloud of a tracked object in a data frame collected by the laser radar; A point cloud model reconstruction processing unit, used to reconstruct the target point cloud to obtain a reconstructed point cloud model; A target frame size estimation processing unit, used to perform target frame size estimation processing based on the point cloud model to obtain a target frame of a target size; The tracking target trajectory optimization processing unit is used to perform tracking target trajectory optimization processing based on the target frame of the target size to obtain an optimized target trajectory.

[0011] According to the tracking target trajectory optimization processing device of the present invention, the point cloud model reconstruction processing unit is specifically used for: Based on the different observation positions of the tracking target, construct a corresponding tracking target local coordinate system; Based on the local coordinate system of the tracked target, the target point clouds of the tracked target in continuous data frames are overlaid to obtain a reconstructed point cloud model.

[0012] According to the tracking target trajectory optimization processing device of the present invention, the corresponding tracking target local coordinate system is constructed based on the different observation positions of the tracking target, including: In the case where one edge of the tracking target is observed based on the preset laser radar, the center point of the edge is used as the origin of the local coordinate system to construct a corresponding first tracking target local coordinate system; in the case where two edges of the tracking target are observed based on the preset laser radar, the intersection formed by the two edges is used as the origin of the local coordinate system to construct a corresponding second tracking target local coordinate system; in the case where three edges of the tracking target are observed based on the preset laser radar, the center point of the longest edge among the three edges is used as the origin of the local coordinate system to construct a corresponding third tracking target local coordinate system; The step of performing superposition processing on the target point clouds of the tracked target in continuous data frames based on the local coordinate system of the tracked target to obtain a reconstructed point cloud model specifically includes: Based on the first tracking target local coordinate system, the target point clouds of the tracking target in continuous data frames are superimposed to obtain a reconstructed point cloud model; based on the second tracking target local coordinate system, the target point clouds of the tracking target in continuous data frames are superimposed to obtain a reconstructed point cloud model; based on the third tracking target local coordinate system, the target point clouds of the tracking target in continuous data frames are superimposed to obtain a reconstructed point cloud model.

[0013] According to the tracking target trajectory optimization processing device of the present invention, the target frame size estimation processing unit is specifically used for: In the case where an edge of the tracking target is observed based on a preset laser radar, a length observation value corresponding to the edge is obtained; the length of another edge that is not observed in all target frames collected within a preset time period of the laser radar is counted to obtain a statistical result; and according to the statistical result, the median of the length of another edge that is not observed is used as a length estimation value; according to the length observation value and the length estimation value, a target frame of the target size is obtained; In the case where two edges of the tracking target are observed based on the preset laser radar, the maximum absolute value in the x-axis direction and the maximum absolute value in the y-axis direction in the local coordinate system of the second tracking target are obtained; and the longest value of the edge length of all target frames in the x-axis direction and the longest value of the other edge length of all target frames in the y-axis direction are obtained; a first distance difference between the maximum absolute value in the x-axis direction and the longest value of the edge length of all target frames in the x-axis direction in the local coordinate system of the second tracking target is determined; a second distance difference between the maximum absolute value in the y-axis direction and the longest value of the other edge length of all target frames in the y-axis direction is determined; when both the first distance difference and the second distance difference are less than the corresponding preset distance In the case of a threshold, it is determined that the point cloud model formed by the currently observed target point cloud meets the preset completeness observation condition, and the maximum absolute value in the x-axis direction and the maximum absolute value in the y-axis direction in the local coordinate system of the second tracking target are respectively used as the length estimation values ​​of the two edges of the target frame, and the target frame of the target size is obtained; in the case that the first distance difference and the second distance difference are greater than or equal to the preset distance threshold, it is determined that the point cloud model formed by the currently observed target point cloud meets the preset completeness observation condition, and the longest value of the edge length of all target frames in the x-axis direction and the longest value of the other edge length of all target frames in the y-axis direction are respectively used as the length estimation values ​​of the two edges of the target frame, and the target frame of the target size is obtained; In the case where three edges of the tracking target are observed based on a preset laser radar, a maximum absolute value and a minimum absolute value in the x-axis direction in the local coordinate system of the third tracking target are obtained, and a first difference between the maximum absolute value and the minimum absolute value in the x-axis direction is determined; and a maximum absolute value and a minimum absolute value in the y-axis direction in the local coordinate system of the third tracking target are obtained, and a second difference between the maximum absolute value and the minimum absolute value in the x-axis direction is determined; the first difference and the second difference are respectively used as length estimation values ​​of two edges of the target frame to obtain a target frame of the target size; The target frame is a rectangular frame composed of edges of different lengths.

[0014] According to the tracking target trajectory optimization processing device of the present invention, the target point cloud acquisition unit is specifically used for: Acquire a data frame collected by the laser radar, and extract a target point cloud in a target frame corresponding to the tracking target from the data frame; Motion compensation is performed on the target point cloud to obtain a motion-compensated target point cloud, and the motion-compensated target point cloud is used as the target point cloud of the tracking object in the data frame collected by the laser radar.

[0015] According to the tracking target trajectory optimization processing device of the present invention, the tracking target trajectory optimization processing unit is specifically used for: According to the target size of the target frame, the center point of the target frame in the corresponding tracking target local coordinate system is determined, and the center point of the target frame in the corresponding tracking target local coordinate system is back-projected to the global coordinate system of the laser radar to obtain the optimized target trajectory.

[0016] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for optimizing the tracking target trajectory as described in any one of the above items is implemented.

[0017] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method for optimizing the tracking target trajectory as described in any one of the above items is implemented.

[0018] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the method for optimizing the tracking target trajectory as described in any one of the above items is implemented.

[0019] The tracking target trajectory optimization processing method provided by the present invention obtains the target point cloud of the tracking object in the data frame collected by the laser radar, reconstructs the target point cloud to obtain a reconstructed point cloud model, performs target frame size estimation processing based on the point cloud model to obtain a target frame of the target size, and performs tracking target trajectory optimization processing based on the target frame of the target size to obtain an optimized target trajectory, which can effectively improve the accuracy and reliability of target tracking and labeling, thereby improving the automatic labeling effect of the tracking target trajectory. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0021] Figure 1 It is a flow chart of the tracking target trajectory optimization processing method provided by the present invention.

[0022] Figure 2 It is a schematic diagram of the relative positions of the target vehicle and the self-vehicle in the first tracking target local coordinate system provided by the present invention.

[0023] Figure 3 It is a schematic diagram of the relative positions of the target vehicle and the self-vehicle in the second tracking target local coordinate system provided by the present invention.

[0024] Figure 4 It is a schematic diagram of the relative positions of the target vehicle and the self-vehicle in the third tracking target local coordinate system provided by the present invention.

[0025] Figure 5 It is a structural schematic diagram of the tracking target trajectory optimization processing device provided by the present invention.

[0026] Figure 6 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0027] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings 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 ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0028] Combine the following Figure 1-Figure 6The target tracking trajectory optimization processing method and device of the present invention are described, and the embodiments thereof are described in detail.

[0029] The following is a detailed description of an embodiment of the tracking target trajectory optimization processing method according to the present invention. Figure 1 As shown, it is a flow chart of the tracking target trajectory optimization processing method provided by the present invention, and the specific implementation process includes the following steps: Step 101: Obtain a target point cloud of a tracked object in a data frame collected by a laser radar.

[0030] In an embodiment of the present invention, it is first necessary to obtain the data frame collected by the laser radar, and the input is the point cloud data of each collected data frame and the target frame of the tracked target in each data frame, so as to extract the target point cloud in the target frame corresponding to the tracked target from the data frame through the algorithm of judging whether the point cloud is in the polygonal bounding box in the prior art. Then, the target point cloud is motion compensated to obtain the target point cloud after motion compensation, and the target point cloud after motion compensation is used as the target point cloud of the tracked object in the data frame collected by the laser radar. Among them, the laser radar can be a laser radar installed on a vehicle, an aircraft or other equipment. In the process of performing motion compensation on the target point cloud and obtaining the target point cloud after motion compensation, the motion compensation technology can be applied to the extracted target point cloud to eliminate the point cloud distortion caused by the motion of the tracked target. Since the tracked target is still moving during the laser radar point cloud scanning, for the tracked target with a fast moving speed, the shape of the tracked target point cloud will be distorted. The tracking frame (i.e., the target frame) input by the present invention filters the speed, and can calculate the offset of the target point cloud by multiplying the target speed and the time difference between t0 and t1 during the laser radar point cloud scanning period. Then, based on the offset of the target point cloud and the timestamp of each point of the target point cloud, each target point cloud is compensated to the position corresponding to the timestamp corresponding to the current data frame, thereby obtaining the motion-compensated target point cloud.

[0031] Step 102: reconstruct the target point cloud to obtain a reconstructed point cloud model.

[0032] In an embodiment of the present invention, a corresponding local coordinate system of the tracking target can be constructed based on the different observation positions of the tracking target, and based on the local coordinate system of the tracking target, the target point cloud of the tracking target in the continuous data frames is superimposed to obtain a reconstructed point cloud model. That is, after motion compensation, the point cloud reconstruction technology is used to integrate the target point clouds of the continuous data frames to construct a stable and continuous point cloud model (i.e., the target 3D model). In this step, how to align the target point clouds of the tracking targets in different data frames is a key issue. If aligned according to the center point, since the center points of the detection frames (i.e., the target frames or tracking frames) of different data frames have unstable corresponding point cloud positions, using the center point as the origin of the local coordinate system of the point cloud to superimpose the target trajectory point cloud will produce point cloud ghosting. Therefore, the present invention establishes different local coordinate systems of the tracking target according to the observation conditions of the tracking object.

[0033] Specifically, in the process of constructing the corresponding local coordinate system of the tracking target based on the different observation positions of the tracking target, when an edge of the tracking target is observed based on the preset laser radar, the center point of the edge is used as the origin of the local coordinate system, and the sideline where the edge is located is used as the y-axis. The x-axis passes through the origin of the local coordinate system and is perpendicular to the edge to construct the corresponding first local coordinate system of the tracking target. For example: the local vehicle (i.e., the vehicle) equipped with a laser radar only observes one edge of the tracking target (i.e., other vehicles). This situation is common in tracking targets in the distance, and the laser point cloud can only scan the position of the rear of the vehicle. In this case, the present invention will select the center point of the observable edge as the local coordinate center (i.e., the origin of the local coordinate system). Figure 2 As shown in FIG. 1 , it is the first tracking target local coordinate system when an edge of the tracking target is observed, including the ego vehicle (i.e., the local vehicle equipped with a lidar) and the tracking target (i.e., the target vehicle). The arrow represents the x-axis or axis of the local coordinate system, and t1, t2, and t3 represent different moments.

[0034] In the case where two edges of the tracking target are observed based on the preset laser radar, the intersection of the two edges is used as the origin of the local coordinate system, the sideline where one of the two edges is located is used as the y-axis, and the sideline where the other of the two edges is located is used as the x-axis to construct the corresponding second tracking target local coordinate system. For example: a local vehicle (i.e., the self-vehicle) equipped with a laser radar only observes the two edges of the tracking target through the laser radar. This situation is common in tracking targets in adjacent lanes of the self-vehicle. For this situation, the present invention selects the intersection of the two observable edges as the target local coordinate center point (i.e., the origin of the local coordinate system), such as Figure 3 As shown, it is the second tracking target local coordinate system when two edges of the tracking target are observed.

[0035] When the three edges of the tracking target are observed based on the preset laser radar, the center point of the longest edge among the three edges is used as the origin of the local coordinate system, and the x-axis is established through the origin of the local coordinate system and perpendicular to the longest edge to construct the corresponding third tracking target local coordinate system. For example: the local vehicle (i.e., the self-vehicle) equipped with a laser radar observes the three edges of the tracking target. This situation is common when the tracking target passes by the side or crosses the self-vehicle. For this situation, the present invention selects the edge with the most observations (i.e., the longest edge), as follows Figure 4 The center point of the long side in is used as the origin of the local coordinate system to overlay the target point cloud. Figure 4 As shown, it is the third tracking target local coordinate system when the three edges of the tracking target are observed.

[0036] In addition, in the process of superimposing the target point clouds of the tracking target in continuous data frames based on the local coordinate system of the tracking target to obtain a reconstructed point cloud model, it includes: based on the first local coordinate system of the tracking target, superimposing the target point clouds of the tracking target in continuous data frames to obtain a reconstructed point cloud model; based on the second local coordinate system of the tracking target, superimposing the target point clouds of the tracking target in continuous data frames to obtain a reconstructed point cloud model; based on the third local coordinate system of the tracking target, superimposing the target point clouds of the tracking target in continuous data frames to obtain a reconstructed point cloud model.

[0037] Step 103: perform target frame size estimation processing based on the point cloud model to obtain a target frame of the target size.

[0038] In an embodiment of the present invention, when an edge of the tracking target is observed based on a preset laser radar, the length observation value corresponding to the edge is obtained; the length of another edge that is not observed in all target frames collected within a preset time period of the laser radar is counted to obtain a statistical result; and the median of the length of another edge that is not observed is used as a length estimation value according to the statistical result; and the target frame of the target size is obtained according to the length observation value and the length estimation value. That is, the local vehicle (i.e., the self-vehicle) equipped with the laser radar only observes one edge of the tracking target through the laser radar, so the other edge can only be estimated. The present invention selects the median of this edge (i.e., an edge of the tracking target that is not observed) of all tracking frames in all collected data frames as the best estimate (i.e., as the length estimation value of the edge of the tracking target that is not observed), and excludes outliers that are too large or too small.

[0039] In addition, when two edges of the tracking target are observed based on the preset laser radar, the maximum absolute value in the x-axis direction and the maximum absolute value in the y-axis direction in the local coordinate system of the second tracking target are obtained; and the longest value of the edge length of all target frames in the x-axis direction and the longest value of the other edge length of all target frames in the y-axis direction are obtained; a first distance difference between the maximum absolute value in the x-axis direction and the longest value of the edge length of all target frames in the x-axis direction in the local coordinate system of the second tracking target is determined; a second distance difference between the maximum absolute value in the y-axis direction and the longest value of the other edge length of all target frames in the y-axis direction is determined; when both the first distance difference and the second distance difference are less than the corresponding preset In the case of a distance threshold, it is determined that the point cloud model formed by the currently observed target point cloud meets the preset completeness observation conditions, and the maximum absolute value in the x-axis direction and the maximum absolute value in the y-axis direction in the local coordinate system of the second tracking target are respectively used as the estimated lengths of the two edges of the target frame to obtain the target frame of the target size; in the case that the first distance difference and the second distance difference are greater than or equal to the preset distance threshold, it is determined that the point cloud model formed by the currently observed target point cloud meets the preset completeness observation conditions, and the longest value of the edge lengths of all target frames in the x-axis direction and the longest value of the other edge lengths of all target frames in the y-axis direction are respectively used as the estimated lengths of the two edges of the target frame to obtain the target frame of the target size.That is, although the two sides of the tracked target are observed, there is still an incomplete observation. Therefore, the present invention calculates the maximum absolute value coordinates max_x (i.e., the maximum absolute value in the x-axis direction in the second tracking target local coordinate system) and max_y (i.e., the maximum absolute value in the y-axis direction) in the x-axis direction and the y-axis direction in the second tracking target local coordinate system for the reconstructed point cloud model, as well as the longest value max_l (i.e., obtaining the longest value of the edge length of all target frames in the x-axis direction) and max_w (i.e., the longest value of the other edge length of all target frames in the y-axis direction) in the two directions corresponding to the x-axis direction and the y-axis direction. When the point cloud model is in the x-axis direction and the y-axis direction, the maximum coordinates and the length distance difference of the two sides are calculated. When dx (i.e., the first distance difference) and dy (i.e., the second distance difference) are less than the preset distance threshold, the point cloud model observation is considered to be complete, and max_x and max_y are used as the optimal estimated values ​​of the target frame (i.e., the maximum absolute value in the x-axis direction and the maximum absolute value in the y-axis direction in the local coordinate system of the second tracking target are used as the estimated lengths of the two edges of the target frame, respectively, to obtain the target frame of the target size); otherwise, it is considered that the reconstructed point cloud model is still incompletely observed, and max_l and max_w are used as the estimated lengths of the target frame (i.e., the longest value of the edge lengths of all target frames in the x-axis direction and the longest value of the other edge lengths of all target frames in the y-axis direction are used as the estimated lengths of the two edges of the target frame, respectively, to obtain the target frame of the target size).

[0040] In addition, when the three edges of the tracking target are observed based on the preset laser radar, the maximum absolute value and the minimum absolute value in the x-axis direction of the third tracking target local coordinate system are obtained, and the first difference between the maximum absolute value and the minimum absolute value in the x-axis direction is determined; and the maximum absolute value and the minimum absolute value in the y-axis direction of the third tracking target local coordinate system are obtained, and the second difference between the maximum absolute value and the minimum absolute value in the x-axis direction is determined; the first difference and the second difference are respectively used as the length estimates of the two edges of the target frame to obtain the target frame of the target size. The target frame is as follows: Figure 2The rectangular frame shown includes edges of different lengths. That is, the present invention observes three edges of the tracked target, and it can be considered that the point cloud model observation is relatively complete. The reconstructed point cloud model is used to calculate the maximum and minimum values ​​of the x-axis and y-axis directions in the corresponding local coordinate system of the tracked target to obtain the estimated length of the side length, the first difference max_l = max_x - min_x, the second difference max_w = max_y- min_y (that is, the first difference and the second difference are respectively used as the estimated length values ​​of the two edges of the target frame to obtain the target frame of the target size).

[0041] Step 104: Optimize the target trajectory based on the target frame of the target size to obtain an optimized target trajectory.

[0042] In an embodiment of the present invention, the center point of the target frame in the corresponding local coordinate system of the tracking target can be determined according to the target size of the target frame, and the center point of the target frame in the corresponding local coordinate system of the tracking target can be projected back to the global coordinate system of the laser radar to obtain the optimized target trajectory. That is, in the process of optimizing the target frame trajectory, the trajectory of the target frame is adjusted to ensure that the position and size of the target frame maintain the best consistency and accuracy during the entire tracking process. In the above size optimization step, the center point of the target frame in the local coordinate system of the corresponding tracking target is obtained according to the target frame of the optimized size (i.e., the target frame of the target size). By tracking the global pose of the target at different times, the center point of the tracking target in the corresponding local coordinate system of the tracking target can be projected back to the global coordinate system to obtain the optimized target trajectory. Among them, the local coordinate system of the tracking target can be the first local coordinate system of the tracking target, the second local coordinate system of the tracking target, or the third local coordinate system of the tracking target. The global coordinate system is a coordinate system with the center of the vehicle on which the laser radar is installed as the origin.

[0043] Compared with the method of performing Kalman filtering and geometric averaging on the target frame, the present invention adopts a method of joint optimization of tracking target and point cloud distribution, which can accurately estimate the size of the tracking target, and by using the target motion information to perform point cloud compensation, the ghosting phenomenon when the target point cloud is superimposed is reduced, and the accuracy of the tracking frame is improved. The target frame size is estimated based on comprehensive observations to ensure the accuracy of the tracking frame size. This method can accurately estimate the size of the tracking frame, providing a more reliable basis for subsequent target recognition. In addition, the tracking frame size consistency is maintained. After determining the target frame size, the system will dynamically adjust the center position of other tracking frames according to the degree of fit between the target frame and the point cloud model to ensure that the size of the tracking frame remains consistent throughout the tracking process. This method ensures the consistency of the tracking frame size, thereby improving the accuracy and reliability of target tracking and annotation of the entire system. Through the optimization of these two aspects, the fit between the target frame and the point cloud can be maximized while maintaining the consistency of the target frame size, thereby achieving a more accurate and reliable 3D tracking target automatic annotation effect. In the present invention, different local coordinate systems are established for 3D tracking targets under different observation conditions to align and reconstruct 3D target point clouds. For 3D targets in different observation situations, the reconstructed point cloud model is used to automatically annotate the target box predicted by the model, and the size of the target box is jointly optimized to obtain the optimal size of the target. The tracking box size is adjusted through adaptive technology to enhance the accuracy of tracking and annotation by utilizing point cloud data and target motion information. The consistency and fit of single target tracking are optimized by adjusting the position and size of the tracking box.

[0044] The tracking target trajectory optimization processing device provided by the present invention is described below. The tracking target trajectory optimization processing device described below and the tracking target trajectory optimization processing method described above can be referred to each other. Figure 5 As shown, it is a schematic diagram of the structure of the tracking target trajectory optimization processing device provided by the present invention. The tracking target trajectory optimization processing device of the present invention specifically includes the following parts: The target point cloud acquisition unit 501 is used to acquire the target point cloud of the tracked object in the data frame collected by the laser radar.

[0045] The point cloud model reconstruction processing unit 502 is used to reconstruct the target point cloud to obtain a reconstructed point cloud model.

[0046] The target frame size estimation processing unit 503 is used to perform target frame size estimation processing based on the point cloud model to obtain a target frame of a target size.

[0047] The tracking target trajectory optimization processing unit 504 is used to perform tracking target trajectory optimization processing based on the target frame of the target size to obtain an optimized target trajectory.

[0048] According to the tracking target trajectory optimization processing device of the present invention, the point cloud model reconstruction processing unit is specifically used for: Based on the different observation positions of the tracking target, construct a corresponding tracking target local coordinate system; Based on the local coordinate system of the tracked target, the target point clouds of the tracked target in continuous data frames are overlaid to obtain a reconstructed point cloud model.

[0049] According to the tracking target trajectory optimization processing device of the present invention, the corresponding tracking target local coordinate system is constructed based on the different observation positions of the tracking target, including: In the case where one edge of the tracking target is observed based on the preset laser radar, the center point of the edge is used as the origin of the local coordinate system to construct a corresponding first tracking target local coordinate system; in the case where two edges of the tracking target are observed based on the preset laser radar, the intersection formed by the two edges is used as the origin of the local coordinate system to construct a corresponding second tracking target local coordinate system; in the case where three edges of the tracking target are observed based on the preset laser radar, the center point of the longest edge among the three edges is used as the origin of the local coordinate system to construct a corresponding third tracking target local coordinate system; The step of performing superposition processing on the target point clouds of the tracked target in continuous data frames based on the local coordinate system of the tracked target to obtain a reconstructed point cloud model specifically includes: Based on the first tracking target local coordinate system, the target point clouds of the tracking target in continuous data frames are superimposed to obtain a reconstructed point cloud model; based on the second tracking target local coordinate system, the target point clouds of the tracking target in continuous data frames are superimposed to obtain a reconstructed point cloud model; based on the third tracking target local coordinate system, the target point clouds of the tracking target in continuous data frames are superimposed to obtain a reconstructed point cloud model.

[0050] According to the tracking target trajectory optimization processing device of the present invention, the target frame size estimation processing unit is specifically used for: In the case where an edge of the tracking target is observed based on a preset laser radar, a length observation value corresponding to the edge is obtained; the length of another edge that is not observed in all target frames collected within a preset time period of the laser radar is counted to obtain a statistical result; and according to the statistical result, the median of the length of another edge that is not observed is used as a length estimation value; according to the length observation value and the length estimation value, a target frame of the target size is obtained; In the case where two edges of the tracking target are observed based on the preset laser radar, the maximum absolute value in the x-axis direction and the maximum absolute value in the y-axis direction in the local coordinate system of the second tracking target are obtained; and the longest value of the edge length of all target frames in the x-axis direction and the longest value of the other edge length of all target frames in the y-axis direction are obtained; a first distance difference between the maximum absolute value in the x-axis direction and the longest value of the edge length of all target frames in the x-axis direction in the local coordinate system of the second tracking target is determined; a second distance difference between the maximum absolute value in the y-axis direction and the longest value of the other edge length of all target frames in the y-axis direction is determined; when both the first distance difference and the second distance difference are less than the corresponding preset distance In the case of a threshold, it is determined that the point cloud model formed by the currently observed target point cloud meets the preset completeness observation condition, and the maximum absolute value in the x-axis direction and the maximum absolute value in the y-axis direction in the local coordinate system of the second tracking target are respectively used as the length estimation values ​​of the two edges of the target frame, and the target frame of the target size is obtained; in the case that the first distance difference and the second distance difference are greater than or equal to the preset distance threshold, it is determined that the point cloud model formed by the currently observed target point cloud meets the preset completeness observation condition, and the longest value of the edge length of all target frames in the x-axis direction and the longest value of the other edge length of all target frames in the y-axis direction are respectively used as the length estimation values ​​of the two edges of the target frame, and the target frame of the target size is obtained; In the case where three edges of the tracking target are observed based on a preset laser radar, a maximum absolute value and a minimum absolute value in the x-axis direction in the local coordinate system of the third tracking target are obtained, and a first difference between the maximum absolute value and the minimum absolute value in the x-axis direction is determined; and a maximum absolute value and a minimum absolute value in the y-axis direction in the local coordinate system of the third tracking target are obtained, and a second difference between the maximum absolute value and the minimum absolute value in the x-axis direction is determined; the first difference and the second difference are respectively used as length estimation values ​​of two edges of the target frame to obtain a target frame of the target size; The target frame is a rectangular frame consisting of edges of different lengths.

[0051] According to the tracking target trajectory optimization processing device of the present invention, the target point cloud acquisition unit is specifically used for: Acquire a data frame collected by the laser radar, and extract a target point cloud in a target frame corresponding to the tracking target from the data frame; Motion compensation is performed on the target point cloud to obtain a target point cloud after motion compensation, and the target point cloud after motion compensation is used as the target point cloud of the tracking object in the data frame collected by the laser radar.

[0052] According to the tracking target trajectory optimization processing device of the present invention, the tracking target trajectory optimization processing unit is specifically used for: According to the target size of the target frame, the center point of the target frame in the corresponding tracking target local coordinate system is determined, and the center point of the target frame in the corresponding tracking target local coordinate system is back-projected to the global coordinate system of the laser radar to obtain the optimized target trajectory.

[0053] The tracking target trajectory optimization processing device provided by the present invention obtains the target point cloud of the tracking object in the data frame collected by the laser radar, reconstructs the target point cloud to obtain a reconstructed point cloud model, performs target frame size estimation processing based on the point cloud model to obtain a target frame of the target size, and performs tracking target trajectory optimization processing based on the target frame of the target size to obtain an optimized target trajectory, which can effectively improve the accuracy and reliability of target tracking and labeling, thereby improving the automatic labeling effect of the tracking target trajectory.

[0054] Figure 6 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 6 As shown, the electronic device may include: a processor 601, a communication interface 604, a memory 602 and a communication bus 603, wherein the processor 601, the communication interface 604 and the memory 602 communicate with each other through the communication bus 603. The processor 601 may call the logic instructions in the memory 602 to execute a tracking target trajectory optimization processing method, the method comprising: obtaining a target point cloud of a tracking object in a data frame collected by a laser radar; reconstructing the target point cloud to obtain a reconstructed point cloud model; estimating the size of a target frame based on the point cloud model to obtain a target frame of the target size; optimizing the tracking target trajectory based on the target frame of the target size to obtain an optimized target trajectory.

[0055] In addition, the logic instructions in the above-mentioned memory 602 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0056] On the other hand, the present application also provides a computer program product, which includes a computer program. The computer program can be stored on a computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the tracking target trajectory optimization processing method provided by the above methods, and the method includes: obtaining a target point cloud of the tracking object in a data frame collected by a laser radar; reconstructing the target point cloud to obtain a reconstructed point cloud model; estimating the size of a target box based on the point cloud model to obtain a target box of the target size; and optimizing the tracking target trajectory based on the target box of the target size to obtain an optimized target trajectory.

[0057] On the other hand, the present application also provides a computer-readable storage medium, which includes a stored program, wherein the program, when running, executes the tracking target trajectory optimization processing method provided by the above-mentioned methods, the method including: obtaining a target point cloud of the tracking object in a data frame collected by a laser radar; reconstructing the target point cloud to obtain a reconstructed point cloud model; estimating the size of a target frame based on the point cloud model to obtain a target frame of the target size; optimizing the tracking target trajectory based on the target frame of the target size to obtain an optimized target trajectory.

[0058] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0059] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0060] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A tracking target trajectory optimization processing method, characterized in that: include: Obtain the target point cloud of the tracked object in the data frame collected by the laser radar; Reconstructing the target point cloud to obtain a reconstructed point cloud model; Performing target frame size estimation processing based on the point cloud model to obtain a target frame of the target size; The target frame of the target size is used to optimize the tracking target trajectory to obtain an optimized target trajectory.

2. The tracking target trajectory optimization processing method according to claim 1 is characterized in that: The step of reconstructing the target point cloud to obtain a reconstructed point cloud model includes: Based on the different observation positions of the tracking target, construct a corresponding tracking target local coordinate system; Based on the local coordinate system of the tracked target, the target point clouds of the tracked target in continuous data frames are overlaid to obtain a reconstructed point cloud model.

3. The tracking target trajectory optimization processing method according to claim 2 is characterized in that: The constructing a corresponding local coordinate system of the tracking target based on the different observation positions of the tracking target includes: In the case where one edge of the tracking target is observed based on the preset laser radar, the center point of the edge is used as the origin of the local coordinate system to construct a corresponding first tracking target local coordinate system; in the case where two edges of the tracking target are observed based on the preset laser radar, the intersection formed by the two edges is used as the origin of the local coordinate system to construct a corresponding second tracking target local coordinate system; in the case where three edges of the tracking target are observed based on the preset laser radar, the center point of the longest edge among the three edges is used as the origin of the local coordinate system to construct a corresponding third tracking target local coordinate system; The step of performing superposition processing on the target point clouds of the tracked target in continuous data frames based on the local coordinate system of the tracked target to obtain a reconstructed point cloud model specifically includes: Based on the first tracking target local coordinate system, the target point clouds of the tracking target in continuous data frames are superimposed to obtain a reconstructed point cloud model; based on the second tracking target local coordinate system, the target point clouds of the tracking target in continuous data frames are superimposed to obtain a reconstructed point cloud model; based on the third tracking target local coordinate system, the target point clouds of the tracking target in continuous data frames are superimposed to obtain a reconstructed point cloud model.

4. The tracking target trajectory optimization processing method according to claim 3 is characterized in that: The performing target frame size estimation processing based on the point cloud model to obtain a target frame of a target size includes: In the case where an edge of the tracking target is observed based on a preset laser radar, a length observation value corresponding to the edge is obtained; the length of another edge that is not observed in all target frames collected within a preset time period of the laser radar is counted to obtain a statistical result; and according to the statistical result, the median of the length of another edge that is not observed is used as a length estimation value; according to the length observation value and the length estimation value, a target frame of the target size is obtained; In the case where two edges of the tracking target are observed based on the preset laser radar, the maximum absolute value in the x-axis direction and the maximum absolute value in the y-axis direction in the local coordinate system of the second tracking target are obtained; and the longest value of the edge length of all target frames in the x-axis direction and the longest value of the other edge length of all target frames in the y-axis direction are obtained; a first distance difference between the maximum absolute value in the x-axis direction and the longest value of the edge length of all target frames in the x-axis direction in the local coordinate system of the second tracking target is determined; a second distance difference between the maximum absolute value in the y-axis direction and the longest value of the other edge length of all target frames in the y-axis direction is determined; when both the first distance difference and the second distance difference are less than the corresponding preset distance In the case of a threshold, it is determined that the point cloud model formed by the currently observed target point cloud meets the preset completeness observation condition, and the maximum absolute value in the x-axis direction and the maximum absolute value in the y-axis direction in the local coordinate system of the second tracking target are respectively used as the length estimation values ​​of the two edges of the target frame, and the target frame of the target size is obtained; in the case that the first distance difference and the second distance difference are greater than or equal to the preset distance threshold, it is determined that the point cloud model formed by the currently observed target point cloud meets the preset completeness observation condition, and the longest value of the edge length of all target frames in the x-axis direction and the longest value of the other edge length of all target frames in the y-axis direction are respectively used as the length estimation values ​​of the two edges of the target frame, and the target frame of the target size is obtained; In the case where three edges of the tracking target are observed based on a preset laser radar, a maximum absolute value and a minimum absolute value in the x-axis direction in the local coordinate system of the third tracking target are obtained, and a first difference between the maximum absolute value and the minimum absolute value in the x-axis direction is determined; and a maximum absolute value and a minimum absolute value in the y-axis direction in the local coordinate system of the third tracking target are obtained, and a second difference between the maximum absolute value and the minimum absolute value in the x-axis direction is determined; the first difference and the second difference are respectively used as length estimation values ​​of two edges of the target frame to obtain a target frame of the target size; The target frame is a rectangular frame consisting of edges of different lengths.

5. The tracking target trajectory optimization processing method according to claim 1 is characterized in that: The step of obtaining a target point cloud of a tracked object in a data frame collected by a laser radar specifically includes: Acquire a data frame collected by the laser radar, and extract a target point cloud in a target frame corresponding to the tracking target from the data frame; Motion compensation is performed on the target point cloud to obtain a target point cloud after motion compensation, and the target point cloud after motion compensation is used as the target point cloud of the tracking object in the data frame collected by the laser radar.

6. The tracking target trajectory optimization processing method according to claim 1, characterized in that: The target frame based on the target size performs tracking target trajectory optimization processing to obtain an optimized target trajectory, specifically including: According to the target size of the target frame, the center point of the target frame in the corresponding tracking target local coordinate system is determined, and the center point of the target frame in the corresponding tracking target local coordinate system is back-projected to the global coordinate system of the laser radar to obtain the optimized target trajectory.

7. A tracking target trajectory optimization processing device, characterized in that: include: A target point cloud acquisition unit, used to acquire a target point cloud of a tracked object in a data frame collected by the laser radar; A point cloud model reconstruction processing unit, used to reconstruct the target point cloud to obtain a reconstructed point cloud model; A target frame size estimation processing unit, used to perform target frame size estimation processing based on the point cloud model to obtain a target frame of a target size; The tracking target trajectory optimization processing unit is used to perform tracking target trajectory optimization processing based on the target frame of the target size to obtain an optimized target trajectory.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the tracking target trajectory optimization processing method as described in any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the tracking target trajectory optimization processing method as claimed in any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the tracking target trajectory optimization processing method as claimed in any one of claims 1 to 6 is implemented.

Images