Automatic driving tracking method, system and storage medium for long-term occlusion scenes

Abstract

The present invention relates to the field of autonomous driving tracking, and specifically to a tracking method, system, and storage medium for an autonomous driving system in a scene with long-term occlusion. A mainstream target detector is used to detect traffic participants in the scene to obtain a detection frame, and a Kalman filter based on a CV motion model is used to predict the trajectory. A virtual trajectory of the traffic participant during the occlusion period is constructed for secondary updating of the Kalman filter parameters. The IOU of the prediction frame and the detection frame are optimally aligned, and a different tracking ID is assigned to the detection frame corresponding to the traffic participant at the current moment. This solves the problem that the current detection-based tracking algorithm (DBT) algorithm is difficult to accurately predict the trajectory of a traffic participant when the traffic participant is occluded for a long time.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving tracking, and specifically to a method, system, and storage medium for autonomous driving tracking in scenes with long-term occlusion. Background Art

[0002] Multi-target tracking algorithms in the industry are divided into two categories: end-to-end joint detection and embedding (JDE) and detection-based tracking (DBT). JDE is generally a joint detection and tracking model. Although it reduces information transmission loss, it requires a large amount of data for training. In addition, the algorithm model has many problems in edge deployment due to computing power constraints. The industry mostly adopts DBT. Most of these methods rely on Kalman filtering (KF), which assumes that the object's motion state in a very short time is linear and that both prediction noise and observation noise follow a Gaussian distribution.

[0003] In the prior art, for example, Chinese patent CN113723190A discloses a multi-target tracking method for synchronously moving targets. However, the method can only accept occlusion by traffic participants for a very short period of time. When a traffic participant is occluded for a long time, the linear estimation of the Kalman filter becomes very inaccurate due to the prolonged occlusion time of the traffic participant. This is because when there is no measurement value for updating the Kalman filter parameters, the Kalman filter will use the prior state estimate as the posterior state update, and the state noise of the traffic participant will continue to accumulate, resulting in error accumulation during the period of occlusion of the traffic participant. In practical applications, the accumulated error will cause the predicted movement direction of the traffic participant to change. Summary of the Invention

[0004] In response to the shortcomings of the existing technology, the present invention provides an autonomous driving tracking method, system and storage medium for long-term occlusion scenarios, which solves the problem that the DBT algorithm is difficult to accurately predict the trajectory of a traffic participant when a traffic participant is occluded for a long time.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] An autonomous driving tracking method for a long-term occlusion scene includes the following steps:

[0007] S1. Obtain detection results of all traffic participants at any time based on a target detection algorithm, and assign a tracking ID to the detection result at the initial time; the detection result includes a detection frame of the traffic participant, the center coordinates of the detection frame, the size of the detection frame, and a confidence level;

[0008] S2. Based on the trajectory data under different tracking IDs, use the Kalman filter to predict the prediction box of different tracking IDs at the current moment;

[0009] S3. Calculate the IOU values ​​of the current detection frame and the prediction frames under different existing tracking IDs in sequence, and form an IOU cost matrix. Set the values ​​in the cost matrix that are less than the IOU threshold to infinity. Apply the Hungarian algorithm to the cost matrix to obtain the best match. Assign the tracking ID of the prediction frame corresponding to the best match to the detection frame corresponding to the best match. Assign new tracking IDs to the detection frames whose IOU values ​​are all less than the IOU threshold.

[0010] S4. Determine whether all tracking IDs except the new tracking ID at the current moment appeared at the previous moment;

[0011] If yes, return to step S1;

[0012] If not, the track corresponding to the detection box that did not appear in the previous moment is marked as a non-continuous matching track, and the tracking ID does not include the new tracking ID;

[0013] S5. Update the parameter values ​​of the Kalman filter according to the detection frames of the tracking ID corresponding to the discontinuous matching trajectory at the current moment and the last time it appeared, so as to generate a virtual trajectory between the two moments.

[0014] Preferably, in step S1, the target detection algorithm is a CenterNet algorithm.

[0015] Preferably, in step S2, the following steps are specifically included:

[0016] S21. Define the state of the Kalman filter , detection box In the above formula, (u, v) is the center coordinate value of the traffic participant, that is, the horizontal coordinate and the vertical coordinate respectively, s is the bounding box area, r is the aspect ratio of the bounding box, assuming that the aspect ratio r is constant, the other three variables is the derivative of the corresponding variable with respect to time, w represents the width of the traffic participant, h represents the height of the traffic participant, and c represents the detection confidence;

[0017] S22. Construct the state transition equation of the CV motion model; the expression of the state transition equation is:

[0018] ,

[0019] In the above formula, and represents the center coordinate of the traffic participant at time t+1, and represents the center coordinates of the traffic participant at time t, Indicates the sensor collection interval;

[0020] S23. Build a CV motion model based on the state transition equation and obtain the prediction box of all tracking IDs under the existing trajectory at the current moment through Kalman filtering. The expression of Kalman filtering is:

[0021]

[0022]

[0023] In the above formula, Indicates that the Kalman filter uses the state transfer matrix and covariance matrix to calculate the prediction box for all current tracking IDs at the current moment. Indicates that the Kalman filter uses observations to update parameters, where To predict the state quantity at time t based on the state quantity at time t-1, is the state transfer equation, is the state quantity at time t-1, To predict the covariance matrix at time t based on the covariance matrix at time t-1, is the covariance matrix at time t, represents the process noise at time t, is the Kalman gain at time t, is the observation matrix at time t, is the observation noise at time t, is the state quantity at time t, is the observation value at the moment, is the covariance matrix at time t, is the identity matrix.

[0024] Preferably, in step S3, the following steps are specifically included:

[0025] S31, obtaining the detection frame at any moment and the prediction frame at the current moment;

[0026] S32. Calculate the IOU values ​​of the current detection frame and the prediction frames under different tracking IDs in sequence. The calculation formula of the IOU value is as follows:

[0027]

[0028]

[0029]

[0030]

[0031]

[0032]

[0033] In the above formula, A and B represent the detection box and prediction box respectively. and Respectively represent the horizontal and vertical coordinates of the center point of object A, and Represent the width and height of object A respectively, and Respectively represent the horizontal and vertical coordinates of the center point of object B, and Represent the width and height of object B respectively, and Respectively represent the horizontal and vertical overlap of object A and object B, represents the overlapping area of ​​objects A and B, and represent the areas of objects A and B respectively, Represents the IOU value of objects A and B, that is, the IOU value of the detection box and the prediction box;

[0034] S32. Construct a cost matrix based on the IOU values ​​of all predicted boxes and detection boxes, set the values ​​in the cost matrix that are less than the IOU threshold to infinity, apply the Hungarian algorithm to the cost matrix to obtain the best match, and assign the tracking ID of the predicted box corresponding to the best match to the detection box corresponding to the best match;

[0035] S33. Assign a new tracking ID to the detection boxes whose IOU values ​​are all less than the IOU threshold.

[0036] Preferably, in step S33, the tracking ID is represented by Arabic numerals, and the new tracking ID is the sum of the currently largest digital ID and a preset fixed value.

[0037] Preferably, in step S5, the following steps are specifically included:

[0038] S51, obtaining the detection frame at the current moment corresponding to the discontinuously matched trajectory and the detection frame where the same tracking ID appeared last time, and marking this period as an occlusion period;

[0039] S52: construct a virtual trajectory of the traffic participant corresponding to the discontinuous matching trajectory during the occlusion period; the virtual trajectory is constructed using a constant speed interpolation algorithm, and the expression of the virtual trajectory is:

[0040]

[0041] In the above formula, Indicates the detection box of the last appearance of the tracking ID of the traffic participant corresponding to the unmatched trajectory. is the current moment and the detection box The detection box with the same tracking ID;

[0042] S53. Update the Kalman filter parameters during the occlusion period according to the virtual trajectory. The expression of the Kalman filter parameters that need to be updated is:

[0043]

[0044] In the above formula, Indicates the Kalman filter parameters that need to be updated during the occlusion period;

[0045] S54. Construct a virtual trajectory based on the detection frames corresponding to the unmatched trajectories before and after the occlusion period and by using a constant speed interpolation method. The position of the traffic participant on the virtual trajectory during the occlusion period is calculated as follows:

[0046]

[0047]

[0048]

[0049]

[0050] In the above formula, and Indicates the horizontal and vertical coordinates of the center point of the detection frame at time t+1, and Indicates the horizontal and vertical coordinates of the center point of the detection frame at time t, Indicates the sensor collection interval, and The constant offsets are calculated by dividing the position difference between the vertical and horizontal coordinates of the center coordinates of the detection frame before and after the occlusion period by the occlusion interval. and Respectively and The derivative with respect to time represents the speed, and t represents the duration of the occlusion period.

[0051] The technical solution also provides a system for an autonomous driving tracking method for long-term occlusion scenes, the system comprising: a processor and a memory, the memory being used to store a computer program, and the computer program, when executed by the processor, implementing any one of the autonomous driving tracking methods for long-term occlusion scenes.

[0052] The technical solution also provides a computer storage medium, which is used to store program data. When the program data is executed by a computer, it implements an autonomous driving tracking method for long-term occlusion scenarios.

[0053] Compared with the prior art, the present invention provides an autonomous driving tracking method, system, and storage medium for long-term occlusion scenarios, which have the following beneficial effects:

[0054] 1. The present invention improves the tracking performance of the autonomous driving system by constructing a virtual trajectory of a traffic participant during the period of occlusion and using the virtual trajectory to re-update the parameters of the Kalman filter so that the tracking ID of the traffic participant does not jump after being re-detected.

[0055] 2. The present invention can also improve the tracking performance of obscured targets in multiple technical fields such as autonomous driving, intelligent logistics, intelligent monitoring, data analysis, and behavior recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0057] Figure 1 This is a flow chart of the present invention's method for autonomous driving tracking in long-term occlusion scenarios;

[0058] Figure 2 A schematic diagram showing the difference between the present invention and the existing tracking method;

[0059] Figure 3 Schematic diagram of the virtual trajectory of the present invention. DETAILED DESCRIPTION

[0060] To make the above-mentioned objectives, features, and advantages of the present invention more clearly understood, the present invention is further described below in detail with reference to the accompanying drawings and specific embodiments. This will enable a full understanding of how this application uses technical means to solve technical problems and achieve technical effects, and to implement the invention accordingly.

[0061] Those skilled in the art will appreciate that all or part of the steps in the following embodiments can be accomplished by instructing related hardware through a program. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0062] To address the problem that existing DBT algorithms have difficulty accurately predicting the trajectory of a traffic participant when the participant is obscured for a long time, the present invention provides an autonomous driving tracking method for long-term occlusion scenarios. The tracking method specifically includes the following steps:

[0063] S1. Based on the target detection algorithm, the detection results of all traffic participants at any time are obtained, and a tracking ID is assigned to the detection results at the initial time; the detection results include the detection frame of the traffic participant, the center coordinates of the detection frame, the size of the detection frame, and the confidence level. For example, when the algorithm just starts running, a tracking ID needs to be assigned to the detection frame to facilitate subsequent acquisition and further calculations; the target detection algorithm includes mainstream target detection algorithms such as CenterNet, YOLO, and DINO to better adapt to different scenarios. The present invention uses CenterNet as the target detection algorithm for comparison with other existing tracking methods. The performance comparison is shown in Table 1 below. It can be found that this shows that the performance of the present invention in HOTA, MOTA, and AssA is greatly improved. Compared with the existing tracking methods, the tracking performance can be significantly improved as a tracker in an autonomous driving system.

[0064] S2. Based on the existing trajectory data under different tracking IDs, a Kalman filter is used to predict the prediction frame of the different tracking IDs at the current moment. The trajectory data is the tracking frame of the different tracking IDs at any moment. To further describe the process of calculating the prediction frame by the Kalman filter, step S2 specifically includes the following steps:

[0065] S21. Define the state of the Kalman filter , detection box In the above formula, (u, v) is the center coordinate value of the traffic participant, s is the bounding box area, r is the bounding box aspect ratio, assuming that the aspect ratio r is constant, the other three variables is the derivative of the corresponding variable with respect to time, w represents the width of the traffic participant, h represents the height of the traffic participant, and c represents the detection confidence;

[0066] S22. Construct the state transition equation of the CV motion model; the expression of the state transition equation is:

[0067] ,

[0068] In the above formula, and represents the center coordinate of the traffic participant at time t+1, and represents the center coordinates of the traffic participant at time t, Indicates the sensor collection interval;

[0069] S23. Build a CV motion model based on the state transition equation and obtain the prediction box of all tracking IDs under the existing trajectory at the current moment through Kalman filtering. The expression of Kalman filtering is:

[0070]

[0071]

[0072] In the above formula, it means that the Kalman filter uses the state transfer matrix and covariance matrix to calculate the prediction box under all current tracking IDs at the current moment, and it means that the Kalman filter uses the observation value to update the parameters, where To predict the state quantity at time t based on the state quantity at time t-1, is the state transfer equation, is the state quantity at time t-1, To predict the covariance matrix at time t based on the covariance matrix at time t-1, is the covariance matrix at time t, represents the process noise at time t, is the Kalman gain at time t, is the observation matrix at time t, is the observation noise at time t, is the state quantity at time t, is the moment observation value, i.e. the detection box The calculation formulas of the parameters can be used to bring the calculation results into In the various formulas of the parameters, the parameters of the prediction box are obtained, namely , is the covariance matrix at time t, is the identity matrix,

[0073] S3, calculate the IOU value of the detection frame at the current moment and the prediction frame under different tracking IDs in turn, construct a cost matrix with the IOU values ​​of all prediction frames and detection frames, set the values ​​less than the IOU threshold in the cost matrix to infinity, use the Hungarian algorithm to obtain the best match for the cost matrix, assign the tracking ID of the prediction frame corresponding to the best match to the detection frame corresponding to the best match, and assign a new tracking ID to the detection frame whose IOU values ​​are all less than the IOU threshold. In order to further illustrate the tracking ID assignment process, step S3 specifically includes the following steps:

[0074] S31. Obtain the detection frame at any moment and the prediction frame under all tracking IDs at the current moment;

[0075] S32. Calculate the IOU values ​​of the current detection frame and the prediction frames under different tracking IDs in sequence. The calculation formula of the IOU value is as follows:

[0076]

[0077]

[0078]

[0079]

[0080]

[0081]

[0082] In the above formula, A and B represent the detection box and prediction box respectively. and Respectively represent the horizontal and vertical coordinates of the center point of object A, and Represent the width and height of object A respectively, and Respectively represent the horizontal and vertical coordinates of the center point of object B, and Represent the width and height of object B respectively, and Respectively represent the horizontal and vertical overlap of object A and object B, represents the overlapping area of ​​objects A and B, and represent the areas of objects A and B respectively, Represents the IOU value of objects A and B, that is, the IOU value of the detection box and the prediction box;

[0083] S32. According to setting the values ​​in the cost matrix that are less than the IOU threshold to infinity, the Hungarian algorithm is used on the cost matrix to obtain the best match, and the tracking ID of the prediction box corresponding to the best match is assigned to the detection box corresponding to the best match; in actual use, the IOU threshold is generally set to 0.3.

[0084] S33. Assign new tracking IDs to the detection boxes whose IOU values ​​are all less than the IOU threshold. In step S33, the tracking IDs are represented by Arabic numerals, and the new tracking ID is the sum of the current largest digital ID and a preset fixed value. The preset fixed value can be set to 1. For example, if the current largest digital ID is 4, the new tracking ID will be assigned 5.

[0085] S4. Determine whether all tracking IDs except the new tracking ID at the current moment appeared at the previous moment;

[0086] If yes, return to step S1;

[0087] If not, the trajectory of the detection frame corresponding to the tracking ID that did not appear in the previous moment is marked as a non-continuous matching trajectory, that is, the traffic participant is blocked, and the tracking ID does not contain the new tracking ID;

[0088] S5. Based on the detection frames before and after the blocked traffic participant ID is blocked, a detection frame at each moment during the blocking period is constructed using a constant speed interpolation method, thereby generating a virtual trajectory of the traffic participant during the blocking period. The parameter values ​​of the Kalman filter belonging to the traffic participant during the blocking period are then updated to ensure that the traffic participant can be continuously tracked. In step S5, the following steps are specifically included:

[0089] S51, obtaining the detection frame at the current moment corresponding to the discontinuously matched trajectory and the detection frame where the same tracking ID appeared last time, and marking this period as an occlusion period;

[0090] S52: construct a virtual trajectory of the traffic participant corresponding to the discontinuous matching trajectory during the occlusion period; the virtual trajectory is constructed using a constant speed interpolation algorithm, and the expression of the virtual trajectory is:

[0091]

[0092] In the above formula, Indicates the detection box of the last appearance of the tracking ID of the traffic participant corresponding to the unmatched trajectory. is the current moment and the detection box The detection box with the same tracking ID;

[0093] S53. Update the Kalman filter parameters during the occlusion period according to the virtual trajectory. The expression of the Kalman filter parameters that need to be updated is:

[0094]

[0095] In the above formula, Indicates the Kalman filter parameters that need to be updated during the occlusion period; is the observed value of the virtual trajectory of the traffic participant during the occlusion period, and the predicted value at any moment on the virtual trajectory is calculated.

[0096] S54. Construct a virtual trajectory based on the detection frames corresponding to the unmatched trajectories before and after the occlusion period and by using a constant speed interpolation method. The position of the traffic participant on the virtual trajectory during the occlusion period is calculated as follows:

[0097]

[0098]

[0099]

[0100]

[0101] In the above formula, and Indicates the horizontal and vertical coordinates of the center point of the detection frame at time t+1, and Indicates the horizontal and vertical coordinates of the center point of the detection frame at time t, Indicates the sensor collection interval, and The constant offsets are calculated by dividing the position difference between the vertical and horizontal coordinates of the center coordinates of the detection frame before and after the occlusion period by the occlusion interval. and Respectively and The derivative with respect to time represents the velocity, and t represents the duration of the occlusion period.

[0102] like Figure 2 As shown, the red box is the detection box, the orange box is the participant's trajectory, the blue box is the trajectory where tracking was lost, and the blue dashed box is the tracking trajectory predicted by the Kalman filter. At time t+1, participant 1 is occluded. At time t+2, participant 1 is tracked. Existing methods have a poor match between the tracking box and the actual participant, resulting in subsequent tracking losses. The method in this paper recalculates the KF parameters from time t to t+2, ensuring that tracking of participant 1 is not lost.

[0103] like Figure 3 As shown in Figure 1, (a) at time t1, the traffic participant is lost due to occlusion, and the KF predicted traffic participant trajectory tracking frame is re-associated with the detection frame of the same tracking ID at time t2. (b) in the existing method, at the next moment of re-association, even if the KF state is updated at time t2, there is still a direction difference between the real traffic participant trajectory and the KF predicted trajectory, so the tracking will be lost again at the next moment t2 (blue). (c) is the virtual trajectory (red) during the recalculation of the traffic participant occlusion proposed by the present invention. It can be seen from the figure that the virtual trajectory is closer to the real state observation of the object, and can still be associated with the detection frame of the tracking ID that has appeared before at the moment after t2.

[0104] The technical solution also provides a system for an autonomous driving tracking method for long-term occlusion scenarios, the system comprising: a processor and a memory, the memory being used to store a computer program, which, when executed by the processor, implements an autonomous driving tracking method for long-term occlusion scenarios.

[0105] The technical solution also provides a computer storage medium, which is used to store program data. When the program data is executed by a computer, it is used to implement an automatic driving tracking method for long-term occlusion scenarios.

[0106] The above embodiments provide a detailed introduction to the present invention. Specific examples are used herein to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core ideas. At the same time, for those skilled in the art, according to the ideas of the present invention, there may be changes in the specific implementation methods and application scopes. In summary, the contents of this specification should not be understood as limiting the present invention.

Claims

1. An autonomous driving tracking method for long-term occlusion scenarios, characterized by: The tracking method specifically includes the following steps: S1. Obtain detection results of all traffic participants at any time based on a target detection algorithm, and assign a tracking ID to the detection result at the initial time; the detection result includes a detection frame of the traffic participant, the center coordinates of the detection frame, the size of the detection frame, and a confidence level; S2. Based on the trajectory data under different tracking IDs, use the Kalman filter to predict the prediction box of different tracking IDs at the current moment; In step S2, the following steps are specifically included: S21. Define the state of the Kalman filter , detection box In the above formula, (u, v) is the center coordinate value of the traffic participant, that is, the horizontal coordinate and the vertical coordinate respectively, s is the bounding box area, r is the aspect ratio of the bounding box, assuming that the aspect ratio r is constant, the other three variables is the derivative of the corresponding variable with respect to time, w represents the width of the traffic participant, h represents the height of the traffic participant, and c represents the detection confidence; S22. Construct the state transition equation of the CV motion model; the expression of the state transition equation is: , ； In the above formula, and represents the center coordinate of the traffic participant at time t+1, and represents the center coordinates of the traffic participant at time t, Indicates the sensor collection interval; S23. Build a CV motion model based on the state transition equation and obtain the prediction box of all tracking IDs under the existing trajectory at the current moment through Kalman filtering. The expression of Kalman filtering is: ； ； In the above formula, Indicates that the Kalman filter uses the state transfer matrix and covariance matrix to calculate the prediction box for all current tracking IDs at the current moment. Indicates that the Kalman filter uses observations to update parameters, where To predict the state quantity at time t based on the state quantity at time t-1, is the state transfer equation, is the state quantity at time t-1, To predict the covariance matrix at time t based on the covariance matrix at time t-1, is the covariance matrix at time t, represents the process noise at time t, is the Kalman gain at time t, is the observation matrix at time t, is the observation noise at time t, is the state quantity at time t, is the observation value at the moment, is the covariance matrix at time t, is the identity matrix; S3. Calculate the IOU values ​​of the current detection frame and the prediction frames under different existing tracking IDs in sequence, and form an IOU cost matrix. Set the values ​​in the cost matrix that are less than the IOU threshold to infinity. Apply the Hungarian algorithm to the cost matrix to obtain the best match. Assign the tracking ID of the prediction frame corresponding to the best match to the detection frame corresponding to the best match. Assign new tracking IDs to the detection frames whose IOU values ​​are all less than the IOU threshold. S4. Determine whether all tracking IDs except the new tracking ID at the current moment appeared at the previous moment; If yes, return to step S1; If not, the track corresponding to the detection box that did not appear in the previous moment is marked as a non-continuous matching track, and the tracking ID does not include the new tracking ID; S5. Generate a virtual trajectory during the occlusion period using a fixed-speed interpolation algorithm based on the detection frames of the tracking ID corresponding to the discontinuously matched trajectory at the current moment and the last time it appeared, so as to update the parameter value of the Kalman filter during the occlusion period.

2. The tracking method according to claim 1, characterized in that In step S1, the target detection algorithm is the CenterNet algorithm.

3. The tracking method according to claim 1, wherein: In step S3, the following steps are specifically included: S31, obtaining the detection frame at any moment and the prediction frame at the current moment; S32. Calculate the IOU values ​​of the current detection frame and the prediction frames under different tracking IDs in sequence. The calculation formula of the IOU value is as follows: ； ； ； ； ； ； In the above formula, A and B represent the detection box and prediction box respectively. and Respectively represent the horizontal and vertical coordinates of the center point of object A, and Represent the width and height of object A respectively, and Respectively represent the horizontal and vertical coordinates of the center point of object B, and Represent the width and height of object B respectively, and Respectively represent the horizontal and vertical overlap of object A and object B, represents the overlapping area of ​​objects A and B, and represent the areas of objects A and B respectively, Represents the IOU value of objects A and B, that is, the IOU value of the detection box and the prediction box; S32. Construct a cost matrix based on the IOU values ​​of all predicted boxes and detection boxes, set the values ​​in the cost matrix that are less than the IOU threshold to infinity, apply the Hungarian algorithm to the cost matrix to obtain the best match, and assign the tracking ID of the predicted box corresponding to the best match to the detection box corresponding to the best match; S33. Assign a new tracking ID to the detection boxes whose IOU values ​​are all less than the IOU threshold.

4. The tracking method according to claim 1, wherein: In step S33, the tracking ID is represented by Arabic numerals, and the new tracking ID is the sum of the current largest digital ID and a preset fixed value.

5. The tracking method according to claim 1, characterized in that: In step S5, the following steps are specifically included: S51, obtaining the detection frame at the current moment corresponding to the discontinuously matched trajectory and the detection frame where the same tracking ID appeared last time, and marking this period as an occlusion period; S52: construct a virtual trajectory of the traffic participant corresponding to the discontinuous matching trajectory during the occlusion period; the virtual trajectory is constructed using a constant speed interpolation algorithm, and the expression of the virtual trajectory is: ； In the above formula, Indicates the detection box of the last appearance of the tracking ID of the traffic participant corresponding to the unmatched trajectory. is the current moment and the detection box The detection box with the same tracking ID; S53. Update the Kalman filter parameters during the occlusion period according to the virtual trajectory. The expression of the Kalman filter parameters that need to be updated is: ； In the above formula, Indicates the Kalman filter parameters that need to be updated during the occlusion period; S54. Construct a virtual trajectory based on the detection frames corresponding to the unmatched trajectories before and after the occlusion period and by using a constant speed interpolation method. The position of the traffic participant on the virtual trajectory during the occlusion period is calculated as follows: ； ； ； ； In the above formula, and Indicates the horizontal and vertical coordinates of the center point of the detection frame at time t+1, and Indicates the horizontal and vertical coordinates of the center point of the detection frame at time t, Indicates the sensor collection interval, and The constant offsets are calculated by dividing the position difference between the vertical and horizontal coordinates of the center coordinates of the detection frame before and after the occlusion period by the occlusion interval. and Respectively and The derivative with respect to time represents the speed, and t represents the duration of the occlusion period.

6. A system for implementing the tracking method according to any one of claims 1 to 5, characterized in that: The system comprises a processor and a memory, wherein the memory is used to store a computer program, and when the computer program is executed by the processor, the tracking method according to any one of claims 1 to 5 is implemented.

7. A computer storage medium, wherein the computer storage medium is used to store program data, and when the program data is executed by a computer, the program data implements the automatic driving tracking method for long-term occlusion scenes according to any one of claims 1 to 5.

Images