Laser radar multi-target tracking system and method in complex scene

Abstract

The application discloses a laser radar multi-target tracking system and method in a complex scene, comprising a detection module, a state prediction module, a data association module, a state updating module and a trajectory management module, three-dimensional laser radar acquires original point cloud data in real time and obtains target detection results by the detection module; the position of each target at t-1 time in the subsequent time period is calculated by the state prediction module; the detection results are associated and matched with the calculation results by the association module; the detection results matched successfully in S3 are taken as new observation data to update the calculation results by the updating module; the associated results are managed by the trajectory management module; in conclusion, the application overcomes the shortcomings of the prior art, is reasonable in design, provides an efficient and accurate three-dimensional laser radar multi-target tracking method in a complex scene, and has high social use value and application prospect.

Description

Technical Field

[0001] This invention relates to the field of linear multi-target tracking technology, and in particular to a lidar multi-target tracking system and method for complex scenarios. Background Technology

[0002] In recent years, with the increasing complexity and variability of application environments, radars are required to have multi-target tracking capabilities and be able to track multiple targets simultaneously.

[0003] Existing methods are prone to misassociation or trajectory loss under occlusion conditions, making it difficult to effectively maintain the continuity of target IDs. Most methods employ a single-stage matching strategy, failing to fully utilize features at different levels for multi-level matching, resulting in decreased association accuracy in complex scenes. To address these issues, this invention proposes an efficient lidar multi-target tracking system and method for complex scenes, enhancing the robustness, occlusion recovery capability, and real-time performance of target tracking. Summary of the Invention

[0004] To address the problems mentioned in the background section, this invention provides a multi-target tracking system and method for lidar in complex scenarios.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] A multi-target tracking system for lidar in complex scenarios includes a detection module, a state prediction module, a data association module, a state update module, and a trajectory management module.

[0007] The detection module is used to detect the current data acquired by the 3D LiDAR. The raw point cloud data at each moment is used to obtain the target at... Detection results of time frames ;

[0008] The state prediction module for By predicting and calculating the position of each target at a given time in subsequent time intervals, we can obtain... The target in the time frame Prediction state of time frame ;

[0009] The data association module will Detection results of time frames and The target in the time frame Prediction state of time frame Perform correlation matching and output the detection results of successful matches. Successfully matched trajectory status and unmatched detection results Unmatched trajectory status ;

[0010] The status update module updates the detection results that were successfully matched in the data association module. As new observational data, it is used to analyze the status of each successfully matched trajectory in the data association. Updated, finally obtained All the trajectories at every moment ;

[0011] The trajectory management module will [perform a certain amount of time] Unmatched detection results where no matching trajectory exists in consecutive frames within the same frame. Initialize to a new trajectory state And incorporate it into the trajectory set; the trajectory management module will perform long-term... No test results were found inside. Matched and unmatched trajectory states The set of trajectories to be removed.

[0012] A multi-target tracking method using lidar in complex scenarios includes the following steps:

[0013] S1: The 3D LiDAR acquires raw point cloud data in real time, and the detection module obtains the target detection result;

[0014] S2: Through the state prediction module... The position of each target at a given time point is predicted and calculated for subsequent time periods.

[0015] S3: The detection results are correlated and matched with the prediction calculation results through the correlation module;

[0016] S4: The update module updates the prediction calculation results by using the detection results of successful association matching in S3 as new observation data;

[0017] S5: Perform trajectory management on the associated results.

[0018] Preferably, step S2 is specifically as follows: In order to accurately associate the candidate detection box with the existing trajectory, for For each tracked object at any given time, it is necessary to predict its current position. The state at each time step: For the target at each time step, its state information is represented using an eleven-dimensional dataset.

[0019]

[0020] in, These are the centroid coordinates of the target in a three-dimensional coordinate system. It is the target's heading angle. It refers to the three-dimensional dimensions of the target boundary. These are the velocity components of the target in different directions within the three-dimensional coordinate system; assuming that in Existing in time frames If there are several objectives, then the following set represents them. One goal:

[0021]

[0022] Furthermore, a state transformation model is constructed based on the target's coordinates and component velocities, thereby enabling the application of this model. Estimate the target pose at time:

[0023]

[0024] By targeting By predicting the position of the time interval after a given time, we can obtain the result. All targets in the current time frame Prediction status of time frame:

[0025] .

[0026] Preferably, step S3 is as follows:

[0027] S31: Adaptive Transformation Intersection-Union Ratio (AC-IoU):

[0028] First, a threshold will be set. When the calculated crossover ratio is greater than When the intersection-union ratio (IUR) is less than a certain value, the IUR is reliable and this method is used for data association; when the IUR is less than a certain value... In some cases, using only the intersection-union ratio is unreliable, so a scale parameter is introduced. and As a penalty item; among which This represents the straight-line distance between the centroids of the 3D predicted bounding box B and the ground truth bounding box A. This represents the diagonal length of the minimum bounding box between the 3D predicted bounding box B and the ground truth bounding box A. The transformed intersection-over-union (IoU) ratio directly minimizes the distance between the two target boxes, and only adds two line distance calculations, making it fast and not consuming too many computational resources. The specific formula is as follows:

[0029]

[0030] S31: Cascade Matching

[0031] The first-stage data association uses the AC-IoU calculation method as the standard for successful matching, focusing on the degree of overlap between the tracked objects. First, the detection results of the current frame are input. The trajectory prediction results of the target at the previous moment and at the current moment. The first-stage matching module calculates the cost matrix using the adaptive transformation intersection-union ratio (AC-IoU), and then feeds the cost matrix into the Hungarian algorithm for matching. Finally, it outputs the detection results of successful first-stage matching. Trajectory status of successful first-stage matching and the test results of the first stage that did not match. The first stage of the unmatched trajectory status ;

[0032] The second-stage data association process handles objects that failed to match in the first stage. To support successful association of unmatched objects again, the second-stage association introduces Euclidean distance and heading angle to calculate positional correlation. The detection results of unmatched objects from the first stage are then processed. The first stage of the unmatched trajectory status The second-stage matching module further combines geometric position and heading angle for matching, resolving target matching problems under complex conditions such as occlusion and low confidence, and outputting the detection result of successful second-stage matching. Trajectory state that is successfully matched with the second stage and the final unmatched detection results. and the final unmatched trajectory state Its cascading matching is as follows Figure 2 As shown;

[0033] Among them, the detection results of successful matches Including the detection results of successful first-stage matching Detection results that are successfully matched in the second phase Successfully matched trajectory status Including the trajectory status of a successful first-stage match Trajectory state that is successfully matched with the second stage Unmatched detection results Including the final unmatched test results Unmatched trajectory status Including the final unmatched test results .

[0034] Preferably, step S4 is as follows:

[0035] In Kalman filtering, state prediction is based solely on the motion model. However, targets in real-world environments may be affected by noise or various external factors, leading to significant deviations from the true values ​​in the predictions. This chapter addresses this by incorporating data association results into the predictions, thus reasonably correcting the previous predictions to obtain estimates closer to the true states. Considering that each target may experience abrupt state changes or shifts in the environment, a single state prediction is unstable. Therefore, the detection results successfully matched in the data association process are also incorporated into the predictions. As new observational data, it is used to analyze the status of each successfully matched trajectory in the data association. The update was performed, and the final result was obtained in All trajectories T at time T t :

[0036] .

[0037] Preferably, step S5 specifically includes the following: Trajectory management of the associated results: when there are unmatched detection results in the detection results... When the detection result is not matched Within a certain period of time When a trajectory exists in consecutive frames without a matching one, it will be initialized with a new trajectory. A unique ID is assigned to it, and its state is initialized using a Kalman filter, thereby incorporating it into the trajectory set for subsequent tracking, ensuring that new targets can be captured in real time; while for long-term... There were no unmatched trajectory states that were not matched by the detection results. The system will remove these tracks from the track set to clean up invalid tracks. This avoids interference with subsequent tracking; the synergistic effect of adding and deleting trajectories enables the system to dynamically adapt to the appearance and disappearance of targets, thereby maintaining the rationality and efficiency of the trajectory set and improving tracking performance and computational efficiency.

[0038] Compared with existing technologies, the advantages of this invention are: in complex scenarios, the state prediction module can accurately predict the current state of each target. The AC-IoU used in the association phase comprehensively considers the target's appearance and geometric features, making the association more robust. Even with target occlusion, complex motion, or large target spacing, it effectively reduces false matching and missed matching. Simultaneously, the cascaded matching strategy further optimizes the accuracy of association by processing trajectories and detection results of different priorities in stages, effectively eliminating identity switching problems caused by occlusion or target overlap, especially in complex scenarios such as long-term occlusion or the initial appearance of new targets. In the trajectory management module, the addition and deletion of trajectories work synergistically, enabling the system to dynamically adapt to the appearance and disappearance of targets, thereby maintaining the rationality and efficiency of the trajectory set and improving tracking performance and computational efficiency.

[0039] In summary, this invention overcomes the shortcomings of existing technologies, has a reasonable design, and provides an efficient and accurate three-dimensional lidar multi-target tracking method in complex scenarios, which has high social value and application prospects. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 A flowchart illustrating the overall structure of the present invention;

[0042] Figure 2 This is a flowchart of the cascaded matching process of the present invention;

[0043] Figure 3 This is a flowchart of the trajectory management process of the present invention. Detailed Implementation

[0044] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0045] Example 1

[0046] Reference Figure 1A multi-target tracking system for lidar in complex scenarios, comprising a detection module, a state prediction module, a data association module, a state update module, and a trajectory management module;

[0047] Furthermore, the detection module is used to detect the current data acquired by the 3D LiDAR. The raw point cloud data at each moment is used to obtain the target at... Detection results of time frames .

[0048] Furthermore, the state prediction module for By predicting and calculating the position of each target at a given time in subsequent time intervals, we can obtain... The target in the time frame Prediction state of time frame The specific steps are as follows:

[0049] To accurately associate candidate detection boxes with existing trajectories, for For each tracked object at any given time, it is necessary to predict its position at the current time. The state of the target at each time step is represented by an eleven-dimensional dataset:

[0050]

[0051] in, These are the centroid coordinates of the target in a three-dimensional coordinate system. It is the target's heading angle. It refers to the three-dimensional dimensions of the target boundary. These are the velocity components of the target in different directions within the three-dimensional coordinate system. Assuming that... Existing in time frames If there are several objectives, then the following set represents them. One goal:

[0052]

[0053] Furthermore, a state transformation model is constructed based on the target's coordinates and component velocities, thereby enabling the application of this model. Estimate the target pose at time:

[0054]

[0055] By targeting By predicting the position of the time interval after a given time, we can obtain the result. All targets in the current time frame Prediction status of time frame:

[0056] .

[0057] The data association module will Detection results of time frames and The target in the time frame Prediction state of time frame Perform correlation matching and output the detection results of successful matches. Successfully matched trajectory status and unmatched detection results Unmatched trajectory status The specific principle is as follows:

[0058] 1. Adaptive Transformation Intersection over Union (AC-IoU):

[0059] First, a threshold will be set. When the calculated crossover ratio is greater than When the intersection-union ratio (IUR) is less than a certain value, the IUR is reliable and this method is used for data association; when the IUR is less than a certain value... In some cases, using only the intersection-union ratio is unreliable, so a scale parameter is introduced. and As a penalty item. This represents the straight-line distance between the centroids of the 3D predicted bounding box B and the ground truth bounding box A. This represents the diagonal length of the minimum bounding box between the 3D predicted bounding box B and the ground truth bounding box A. The transformed intersection-union ratio (IU) directly minimizes the distance between the two target boxes, and only adds two line distance calculations, making it fast and not consuming too many computational resources. The specific formula is as follows:

[0060]

[0061] 2. Cascading Matching:

[0062] The first-stage data association uses the AC-IoU calculation method as the standard for successful matching, focusing on the degree of overlap between the tracked objects. First, the detection results of the current frame are input. The trajectory prediction results of the target at the previous moment and at the current moment. The first-stage matching module calculates the cost matrix using the adaptive transformational intersection-union ratio (AC-IoU), and then feeds the cost matrix into the Hungarian algorithm for matching. Finally, it outputs the detection results of successful first-stage matching. Trajectory status of successful first-stage matching and the test results of the first stage that did not match. The first stage of the unmatched trajectory status .

[0063] The second-stage data association process handles objects that failed to match in the first stage. To support successful association of unmatched objects again, the second-stage association introduces Euclidean distance and heading angle to calculate positional correlation. (The image shows the detection results for unmatched objects from the first stage.) The first stage of the unmatched trajectory status The second-stage matching module further combines geometric position and heading angle for matching, resolving target matching problems under complex conditions such as occlusion and low confidence, and outputting the detection result of successful second-stage matching. Trajectory state that is successfully matched with the second stage and the final unmatched detection results. and the final unmatched trajectory state Its cascading matching is as follows Figure 2 As shown.

[0064] Among them, the detection results of successful matches Including the detection results of successful first-stage matching Detection results that are successfully matched in the second phase Successfully matched trajectory status Including the trajectory status of a successful first-stage match Trajectory state that is successfully matched with the second stage Unmatched detection results Including the final unmatched test results Unmatched trajectory status Including the final unmatched test results .

[0065] Furthermore, the state update module will update the detection results that are successfully matched in the data association module. As new observational data, it is used to analyze the status of each successfully matched trajectory in the data association. Updated, finally obtained All the trajectories at every moment The design steps are as follows:

[0066] In Kalman filtering, state prediction is based solely on the motion model. However, targets in real-world environments may be affected by noise or various external factors, leading to significant deviations from the true values ​​in the predictions. This chapter addresses this by incorporating data association results into the predictions, thus reasonably correcting the previous predictions to obtain estimates closer to the true states. Considering that each target may experience abrupt state changes or shifts in the environment, a single state prediction is unstable. Therefore, the detection results that were successfully matched in the data association process are also incorporated into the predictions. As new observational data, it is used to analyze the status of each successfully matched trajectory in the data association. The update was performed, and the final result was obtained in All trajectories T at time T t :

[0067] .

[0068] Furthermore, the trajectory management module will [perform a certain amount of time]. Unmatched detection results where no matching trajectory exists in consecutive frames within the same frame. Initialize to a new trajectory state And incorporate it into the trajectory set; the trajectory management module will perform long-term... No test results were found inside. Matched and unmatched trajectory states The specific design steps and principles for removing the trajectory set are as follows:

[0069] The trajectory management flowchart is as follows Figure 3 As shown.

[0070] A multi-target tracking method using lidar in complex scenarios includes the following steps:

[0071] S1: The 3D LiDAR acquires raw point cloud data in real time, and the detection module obtains the target detection results.

[0072] S2: Through the state prediction module... The position of each target at a given time point is predicted and calculated for subsequent time periods:

[0073] To accurately associate candidate detection boxes with existing trajectories, for For each tracked object at any given time, it is necessary to predict its position at the current time. The state of the target at each time step is represented by an eleven-dimensional dataset:

[0074]

[0075] in, These are the centroid coordinates of the target in a three-dimensional coordinate system. It is the target's heading angle. These are the three-dimensional dimensions of the target boundary. These are the velocity components of the target in different directions within the three-dimensional coordinate system. Assuming that... Existing in time frames If there are several objectives, then the following set represents them. One goal:

[0076]

[0077] Furthermore, a state transformation model is constructed based on the target's coordinates and component velocities, thereby enabling the application of this model. Estimate the target pose at time:

[0078]

[0079] By targeting By predicting the position of the time interval after a given time, we can obtain the result. All targets in the current time frame Prediction status of time frame:

[0080] .

[0081] S3: The detection results are correlated and matched with the prediction calculation results through the correlation module:

[0082] S31: Adaptive Transition Intersection-Union Ratio (AC-IoU):

[0083] First, a threshold will be set. When the calculated crossover ratio is greater than When the intersection-union ratio (IUR) is less than a certain value, the IUR is reliable and this method is used for data association; when the IUR is less than a certain value... In some cases, using only the intersection-union ratio is unreliable, so a scale parameter is introduced. and As a penalty item. This represents the straight-line distance between the centroids of the 3D predicted bounding box B and the ground truth bounding box A. This represents the diagonal length of the minimum bounding box between the 3D predicted bounding box B and the ground truth bounding box A. The transformed intersection-union ratio (IU) directly minimizes the distance between the two target boxes, and only adds two line distance calculations, making it fast and not consuming too many computational resources. The specific formula is as follows:

[0084]

[0085] S31: Cascade Matching

[0086] The first-stage data association uses the AC-IoU calculation method as the standard for successful matching, focusing on the degree of overlap between the tracked objects. First, the detection results of the current frame are input. The trajectory prediction results of the target at the previous moment and at the current moment. The first-stage matching module calculates the cost matrix using the adaptive transformational intersection-union ratio (AC-IoU), and then feeds the cost matrix into the Hungarian algorithm for matching. Finally, it outputs the detection results of successful first-stage matching. Trajectory status of successful first-stage matching and the test results of the first stage that did not match. The first stage of the unmatched trajectory status .

[0087] The second-stage data association process handles objects that failed to match in the first stage. To support successful association of unmatched objects again, the second-stage association introduces Euclidean distance and heading angle to calculate positional correlation. (The image shows the detection results for unmatched objects from the first stage.) The first stage of the unmatched trajectory status The second-stage matching module further combines geometric position and heading angle for matching, resolving target matching problems under complex conditions such as occlusion and low confidence, and outputting the detection result of successful second-stage matching. Trajectory state that is successfully matched with the second stage and the final unmatched detection results and the final unmatched trajectory state Its cascaded matching is as follows Figure 2 As shown.

[0088] Among them, the detection results of successful matches Including the detection results of successful first-stage matching Detection results that are successfully matched in the second phase Successfully matched trajectory status Including the trajectory status of a successful first-stage match Trajectory state that is successfully matched with the second stage Unmatched detection results Including the final unmatched test results Unmatched trajectory status Including the final unmatched test results .

[0089] S4: The update module updates the prediction calculation results by using the detection results of successful association matching in S3 as new observation data.

[0090] In Kalman filtering, state prediction is based solely on the motion model. However, targets in real-world environments may be affected by noise or various external factors, leading to significant deviations from the true values ​​in the predictions. This chapter addresses this by incorporating data association results into the predictions, thus reasonably correcting the previous predictions to obtain estimates closer to the true states. Considering that each target may experience abrupt state changes or shifts in the environment, a single state prediction is unstable. Therefore, the detection results that were successfully matched in the data association process are also incorporated into the predictions. As new observational data, it is used to analyze the status of each successfully matched trajectory in the data association. The update was performed, and the final result was obtained in All trajectories T at time T t :

[0091] .

[0092] S5: Track management of the associated results: When there are unmatched detection results in the detection results. When the detection result is not matched Within a certain period of time When a trajectory exists in consecutive frames without a matching one, it will be initialized with a new trajectory. A unique ID is assigned to it, and its state is initialized using a Kalman filter, thereby incorporating it into the trajectory set for subsequent tracking, ensuring that new targets can be captured in real time. However, for long-term... There were no unmatched trajectory states that were not matched by the detection results. The system will remove these tracks from the track set to clean up invalid tracks. This avoids interfering with subsequent tracking. The synergistic effect of adding and deleting trajectories allows the system to dynamically adapt to the appearance and disappearance of targets, thereby maintaining the rationality and efficiency of the trajectory set and improving tracking performance and computational efficiency.

[0093] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," "counterclockwise," "axial," "radial," and "circumferential" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0094] In this invention, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," "link," and "fix" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection, a direct connection, or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0095] The control method of this invention is automatic control through a controller. The control circuit of the controller can be implemented by simple programming by those skilled in the art. The power supply is also common knowledge in the art. Furthermore, since this invention is mainly used to protect mechanical devices, the control method and circuit connection will not be explained in detail here.

[0096] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A multi-target tracking system for lidar in complex scenarios, characterized in that: It includes a detection module, a state prediction module, a data association module, a state update module, and a trajectory management module; The detection module is used to detect the current data acquired by the 3D LiDAR. The raw point cloud data at each moment is used to obtain the target at... Detection results of time frames ; The state prediction module for By predicting and calculating the position of each target at a given time in subsequent time intervals, we can obtain... The target in the time frame Prediction state of time frame ; The data association module will Detection results of time frames and The target in the time frame Prediction state of time frame Perform correlation matching and output the detection results of successful matches. Successfully matched trajectory status and unmatched detection results Unmatched trajectory status ; The status update module updates the detection results that were successfully matched in the data association module. As new observational data, it is used to analyze the status of each successfully matched trajectory in the data association. Updated, finally obtained All the trajectories at every moment ; The trajectory management module will [perform a certain amount of time] Unmatched detection results where no matching trajectory exists in consecutive frames within the same frame. Initialize to a new trajectory state And incorporate it into the trajectory set; the trajectory management module will perform long-term... No test results were found inside. Matched and unmatched trajectory states Remove from the set of trajectories; A multi-target tracking method using lidar in complex scenarios includes the following steps: S1: The 3D LiDAR acquires raw point cloud data in real time, and the detection module obtains the target detection result; S2: Through the state prediction module... The position of each target at a given time point is predicted and calculated for subsequent time periods. S3: The detection results are correlated and matched with the prediction calculation results through the correlation module; S4: The update module updates the prediction calculation results by using the detection results of successful association matching in S3 as new observation data; S5: Perform trajectory management on the associated results; Step S3 is as follows: S31: Adaptive Transformation Intersection-Union Ratio (AC-IoU): First, a threshold will be set. When the calculated crossover ratio is greater than When the intersection-union ratio (IUR) is less than a certain value, the IUR is reliable and this method is used for data association; when the IUR is less than a certain value... In some cases, using only the intersection-union ratio is unreliable, so a scale parameter is introduced. and As a penalty item; among which This represents the straight-line distance between the centroids of the 3D predicted bounding box B and the ground truth bounding box A. This represents the diagonal length of the minimum bounding box between the 3D predicted bounding box B and the ground truth bounding box A. The transformed intersection-union ratio (IU) directly minimizes the distance between the two target boxes, and only adds two line distance calculations, making it fast and not consuming too many computational resources. The specific formula is as follows: S32: Cascaded Matching The first-stage data association uses the AC-IoU calculation method as the standard for successful matching, focusing on the degree of overlap between the tracked objects. First, the detection results of the current frame are input. The trajectory prediction results of the target at the previous moment and at the current moment. The first-stage matching module calculates the cost matrix using the adaptive transformation intersection-union ratio (AC-IoU), and then feeds the cost matrix into the Hungarian algorithm for matching. Finally, it outputs the detection results of successful first-stage matching. Trajectory status of successful first-stage matching and the test results of the first stage that did not match. The first stage of the unmatched trajectory status ; The second-stage data association process handles objects that failed to match in the first stage. To support successful association of unmatched objects again, the second-stage association introduces Euclidean distance and heading angle to calculate positional correlation. The detection results of unmatched objects from the first stage are then processed. The first stage of the unmatched trajectory status The second-stage matching module further combines geometric position and heading angle for matching, resolving target matching problems under complex conditions such as occlusion and low confidence, and outputting the detection result of successful second-stage matching. Trajectory state that is successfully matched with the second stage and the final unmatched detection results. and the final unmatched trajectory state ; Among them, the detection results of successful matches Including the detection results of successful first-stage matching Detection results that are successfully matched in the second phase Successfully matched trajectory status Including the trajectory status of a successful first-stage match Trajectory state that is successfully matched with the second stage Unmatched detection results Including the final unmatched test results Unmatched trajectory status Including the final unmatched test results .

2. The multi-target tracking system for lidar in complex scenarios according to claim 1, characterized in that: Step S2 is as follows: To accurately associate candidate detection boxes with existing trajectories, for For each tracked object at any given time, it is necessary to predict its current position. The state at each time step: For the target at each time step, its state information is represented using an eleven-dimensional dataset. in, These are the centroid coordinates of the target in a three-dimensional coordinate system. It is the target's heading angle. It refers to the three-dimensional dimensions of the target boundary. These are the velocity components of the target in different directions within the three-dimensional coordinate system; assuming that in Existing in time frames If there are several objectives, then the following set represents them. One goal: Furthermore, a state transformation model is constructed based on the target's coordinates and component velocities, thereby enabling the application of this model. Estimate the target pose at time: By targeting By predicting the position of the time interval after a given time, we can obtain the result. All targets in the current time frame Prediction status of time frame: 。 3. The multi-target tracking system for lidar in complex scenarios according to claim 1, characterized in that: Step S4 is as follows: In Kalman filtering, state prediction is based solely on the motion model. However, targets in real-world environments may be affected by noise or various external factors, leading to significant deviations from the true values ​​in the predictions. This chapter addresses this by incorporating data association results into the predictions, thus reasonably correcting the previous predictions to obtain estimates closer to the true states. Considering that each target may experience abrupt state changes or shifts in the environment, a single state prediction is unstable. Therefore, the detection results successfully matched in the data association process are also incorporated into the predictions. As new observational data, it is used to analyze the status of each successfully matched trajectory in the data association. The update was performed, and the final result was obtained in All trajectories T at time T t : 。 4. The multi-target tracking system for lidar in complex scenarios according to claim 1, characterized in that: Step S5 is as follows: Trajectory management is performed on the associated results: when there are unmatched detection results in the detection results... When the detection result is not matched Within a certain period of time When a trajectory exists in consecutive frames without a matching one, it will be initialized with a new trajectory. A unique ID is assigned to it, and its state is initialized using a Kalman filter, thereby incorporating it into the trajectory set for subsequent tracking, ensuring that new targets can be captured in real time; while for long-term... There were no unmatched trajectory states that were not matched by the detection results. The system will remove these tracks from the track set to clean up invalid tracks. This avoids interference with subsequent tracking; the synergistic effect of adding and deleting trajectories enables the system to dynamically adapt to the appearance and disappearance of targets, thereby maintaining the rationality and efficiency of the trajectory set and improving tracking performance and computational efficiency.

Images