A track-based negative feedback method and system for enhancing point cloud density

Abstract

This application proposes a point cloud density enhancement method and system based on track negative feedback. The method includes acquiring radar current frame point tracks and historical point track sets; preprocessing the radar current frame point tracks to obtain preprocessed point tracks; filtering stable moving tracks and updating the historical point track set; extrapolating and filtering the updated historical point track set according to the radar frame period; performing point cloud clustering based on the preprocessed point tracks and the filtered historical point track set; performing track matching and tracking based on the point cloud clustering results; and performing track management based on the track matching and tracking results. This application increases point cloud density, improves track stability, and enhances track tracking accuracy by retaining historical point tracks corresponding to tracked stable moving tracks and extrapolating their positions based on their motion states.

Description

Technical Field

[0001] This application relates to the field of radar signal processing technology, and in particular to a point cloud density enhancement method and system based on track negative feedback. Background Technology

[0002] Traffic radar is a commonly used sensor used in the transportation field to achieve functions such as target detection, tracking, and track stabilization. However, in order to improve the radar's maximum detection range, the radar's range resolution is often reduced, resulting in a decrease in the point cloud density of the target. At the same time, the radar's scattering points change with the movement of the target, and the clustering results are prone to problems such as target splitting and large errors.

[0003] In practical radar operation, the radar bandwidth is a fixed parameter that determines the radar's range resolution. Range resolution represents the minimum distance between two different targets that a radar can distinguish. Generally, higher range resolution provides more detailed target distance measurements, but it also limits the radar's maximum detection range.

[0004] Due to the limitations of range resolution, the point cloud density of a target becomes very poor at long distances, failing to meet the requirements for track stability and tracking accuracy. In addition, the movement of the target causes dynamic changes in radar scattering points, which can easily lead to target splitting and large errors when performing target clustering.

[0005] Currently, there is no way to improve radar point cloud density because the radar's range resolution is determined in practice and is dictated by the radar's operating bandwidth. Therefore, a trade-off between maximum detection range and point cloud density needs to be struck when designing and optimizing radar systems, and other methods need to be sought to improve target tracking and track stability. Summary of the Invention

[0006] To address the aforementioned technical problems, this application proposes a point cloud density enhancement method and system based on track negative feedback.

[0007] Firstly, this application provides a point cloud density enhancement method based on track negative feedback, including:

[0008] S1: Obtain radar current frame traces and historical trace sets.

[0009] S2: Preprocess the radar current frame traces to obtain preprocessed traces.

[0010] Preferably, the preprocessing specifically includes: removing invalid and jittery dots, transforming the dots in the current radar frame from the radar coordinate system to the target coordinate system, and performing motion compensation and object feature extraction for the dots information;

[0011] The process involves several steps. First, based on the characteristics and preset conditions of the radar equipment, such as setting signal-to-noise ratio thresholds and dynamic range, invalid points with weak signals or those that do not meet the criteria are identified and removed. Second, filtering algorithms, such as Kalman filtering and median filtering, are used to smooth the points and remove jerky points that do not conform to the motion patterns of objects. The target coordinate system can be a Cartesian coordinate system or a latitude and longitude coordinate system. Third, motion compensation is performed on the point positions based on the velocity information carried by the points to correct their positions and improve accuracy. Finally, feature information of the object, such as position, velocity, and acceleration, is extracted from the pre-processed points for subsequent target tracking, identification, and classification.

[0012] S3: Filter stable motion tracks and update the set of historical points.

[0013] Preferably, step S3 specifically includes:

[0014] S301: Obtain the point trace corresponding to the stable motion track;

[0015] S302: Calculate the lifetime of the point trace based on its existence time or the number of frames continuously observed; wherein, the lifetime of the point trace refers to the number of frames in which the point trace exists in consecutive frames;

[0016] S303: Set a first lifespan threshold and filter tracks whose lifespan is greater than the lifespan threshold;

[0017] Furthermore, if the lifespan is greater than the first lifespan threshold, it is identified as a point corresponding to a stable motion track.

[0018] S304: Update the point information corresponding to the track that is greater than the life cycle threshold to the historical point set; wherein, the point information includes at least position and velocity;

[0019] Furthermore, for tracks that are not selected, appropriate actions can be taken according to specific needs, such as removal, marking, or further analysis.

[0020] S305: Set the lifespan of the track points that are greater than the lifespan threshold to 1.

[0021] S4: Extrapolate and filter the updated historical point set according to the radar frame period, and perform point cloud clustering based on the preprocessed point set and the filtered historical point set.

[0022] Preferably, the step of extrapolating the updated historical point set according to the radar frame period specifically involves:

[0023] S401: Obtain the radar frame period and the current position and velocity of each point in the updated historical point set;

[0024] S402: Based on the product of the velocity and the radar frame period interval, the displacement of each point within the radar frame period interval is obtained;

[0025] S403: Add the current position and displacement of the point to get the position of the point in the next frame, and update it to the historical point set;

[0026] S404: The lifetime of each point in the extrapolated historical point set increases by 1;

[0027] Furthermore, incrementing its lifetime by 1 indicates that the lifetime of the dot is extended by one frame period.

[0028] Preferably, the step of filtering the updated historical point set according to the radar frame period specifically involves: traversing each point in the extrapolated historical point set, setting a second life cycle threshold, and determining whether the life cycle of each point is less than or equal to the second life cycle threshold. If the life cycle of a point is less than or equal to the second life cycle threshold, then the point is retained in the historical point set; otherwise, the point is removed.

[0029] Preferably, the step of clustering the point cloud based on the preprocessed point traces and the filtered historical point traces specifically involves: combining the preprocessed point traces and the filtered historical point traces into a point cloud dataset, using point cloud DBSCAN clustering to cluster the point cloud dataset, dividing points with similar densities into different clusters, and identifying unvisited points as noise points.

[0030] Preferably, DBSCAN is a density-based clustering algorithm that can group points with adjacent densities into a cluster.

[0031] Preferably, the adjustment parameters of the DBSCAN clustering algorithm include at least the minimum number of samples (MinPts) and the neighborhood radius (ε).

[0032] Preferably, the step of clustering the point cloud based on the preprocessed point traces and the filtered historical point trace set further includes: traversing each cluster in the point cloud dataset, calculating and outputting the mathematical mean center point of the cluster;

[0033] Specifically, for each cluster, the position accumulation variable and the point count variable are initialized to 0. Each point in the cluster is traversed. If a point matches the current cluster, the position information of the point is added to the initialized position accumulation variable, and the initialized point count variable is also incremented. The center point of the mathematical mean is equal to the position information of the accumulated point divided by the accumulated point count variable.

[0034] S5: Perform trajectory matching and tracking based on the results of the point cloud clustering.

[0035] Preferably, step S5 specifically includes:

[0036] S501: Initialize the track list by creating an empty track list to store track information in the current frame;

[0037] S502: Obtain the mathematical mean center point of each cluster;

[0038] S503: Traverse each track in the track list;

[0039] S504: Calculate the distance between the track and the center of the mathematical mean of the current cluster, and find the track with the shortest distance.

[0040] Among these methods, the shortest path is found by using appropriate distance metrics, such as Euclidean distance and Mahalanobis distance.

[0041] S505: By setting a maximum distance threshold, determine whether the shortest track is less than the maximum distance threshold. If the shortest track is less than the maximum distance threshold, the current cluster matches the track, and the status information of the track is updated. Otherwise, they do not match, and the unmatched cluster is regarded as a new track.

[0042] S506: Add the unmatched tracks to the track list to expand it;

[0043] Unmatched clusters are treated as new tracks and added to the track list for further tracking.

[0044] S507: Regularly clear outdated flight tracks;

[0045] Specifically, based on the track update time or other set outdated conditions, tracks that have exceeded a certain time or no longer meet the conditions are periodically cleaned up to ensure that the track list maintains the latest and most valid track information.

[0046] S6: Perform track management based on the track matching and tracking results.

[0047] Preferably, the track management includes track updates, track association, track generation and termination, track analysis and prediction, track storage and querying, etc.

[0048] Secondly, this application also provides a point cloud density enhancement system based on track negative feedback, the system comprising:

[0049] The spot acquisition module is used to acquire the spot in the current radar frame;

[0050] The dot preprocessing module removes invalid and jittery dots, transforms the dots in the current radar frame from the radar coordinate system to the target coordinate system, performs motion compensation and object feature extraction on the dot information, and obtains accurate and reliable dot information and continuous track information.

[0051] Historical track set update module; updates the historical track set based on stable motion tracks and radar cycles;

[0052] Track matching and tracking module: can continuously track and update the target's motion status;

[0053] The track management module is used to manage and maintain track information.

[0054] Thirdly, this application also provides a computer-readable storage medium having a computer program stored thereon, including:

[0055] When the computer program is executed by the processor, it implements a point cloud density enhancement method based on track negative feedback as described in any of the preceding claims.

[0056] In summary, this application proposes a point cloud density enhancement method and system based on track negative feedback. The method includes acquiring radar current frame point tracks and historical point track sets, preprocessing the radar current frame point tracks to obtain preprocessed point tracks, filtering stable motion tracks and updating the historical point track set, extrapolating and filtering the updated historical point track set according to the radar frame period, performing point cloud clustering based on the preprocessed point tracks and the filtered historical point track set, performing track matching and tracking based on the result of the point cloud clustering, and performing track management based on the result of the track matching and tracking.

[0057] Compared with the prior art, this application has the following technical effects:

[0058] This application retains historical points corresponding to stable motion tracks and extrapolates their positions based on their motion states. The extrapolated points are then used in the clustering and tracking processing of the current frame. This increases the point cloud density, helps improve the accuracy of the clustering algorithm, and effectively suppresses changes in scattering points caused by target motion, thereby improving track stability, reducing track tracking error growth, and increasing track tracking accuracy. Attached Figure Description

[0059] Figure 1 This is a flowchart of the point cloud density enhancement method based on track negative feedback described in this application.

[0060] Figure 2 This is a flowchart illustrating the process of filtering stable motion tracks and updating the historical point set in one embodiment.

[0061] Figure 3This is a flowchart illustrating the process of extrapolating the updated historical point set according to the radar frame period in one embodiment.

[0062] Figure 4 This is a schematic diagram of the point cloud density enhancement system based on track negative feedback described in this application. Detailed Implementation

[0063] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0064] Example 1:

[0065] like Figure 1 As shown, this application provides a point cloud density enhancement method based on track negative feedback, including the following steps:

[0066] S1: Obtain radar current frame traces and historical trace sets.

[0067] S2: Preprocess the radar current frame traces to obtain preprocessed traces.

[0068] Preferably, the preprocessing specifically includes: removing invalid and jittery dots, transforming the dots in the current radar frame from the radar coordinate system to the target coordinate system, and performing motion compensation and object feature extraction for the dots information;

[0069] The process involves several steps. First, based on the characteristics and preset conditions of the radar equipment, such as setting signal-to-noise ratio thresholds and dynamic range, invalid points with weak signals or those that do not meet the criteria are identified and removed. Second, filtering algorithms, such as Kalman filtering and median filtering, are used to smooth the points and remove jerky points that do not conform to the motion patterns of objects. The target coordinate system can be a Cartesian coordinate system or a latitude and longitude coordinate system. Third, motion compensation is performed on the point positions based on the velocity information carried by the points to correct their positions and improve accuracy. Finally, feature information of the object, such as position, velocity, and acceleration, is extracted from the pre-processed points for subsequent target tracking, identification, and classification.

[0070] S3: Filter stable motion tracks and update the set of historical points.

[0071] Preferably, such as Figure 2 As shown, step S3 specifically includes:

[0072] S301: Obtain the point trace corresponding to the stable motion track;

[0073] S302: Calculate the lifetime of the point trace based on its existence time or the number of frames continuously observed; wherein, the lifetime of the point trace refers to the number of frames in which the point trace exists in consecutive frames;

[0074] S303: Set a first lifespan threshold and filter tracks whose lifespan is greater than the lifespan threshold;

[0075] Furthermore, if the lifespan is greater than the first lifespan threshold, it is identified as a point corresponding to a stable motion track.

[0076] S304: Update the point information corresponding to the track that is greater than the life cycle threshold to the historical point set; wherein, the point information includes at least position and velocity;

[0077] Furthermore, for tracks that are not selected, appropriate actions can be taken according to specific needs, such as removal, marking, or further analysis.

[0078] S305: Set the lifespan of the track points that are greater than the lifespan threshold to 1.

[0079] S4: Extrapolate and filter the updated historical point set according to the radar frame period, and perform point cloud clustering based on the preprocessed point set and the filtered historical point set.

[0080] Preferably, such as Figure 3 As shown, the step of extrapolating the updated historical point set according to the radar frame period specifically involves:

[0081] S401: Obtain the radar frame period and the current position and velocity of each point in the updated historical point set;

[0082] S402: Based on the product of the velocity and the radar frame period interval, the displacement of each point within the radar frame period interval is obtained;

[0083] S403: Add the current position and displacement of the point to get the position of the point in the next frame, and update it to the historical point set;

[0084] S404: The lifetime of each point in the extrapolated historical point set increases by 1;

[0085] Furthermore, incrementing its lifetime by 1 indicates that the lifetime of the dot is extended by one frame period.

[0086] In one embodiment, for each trace in the historical trace set, the current position (x, y), velocity (Vx, Vy), and time interval Δt of the previous frame are obtained. The velocity data of the trace is scaled using the product of velocity and time interval to obtain the displacement moved within the time interval Δt (Δx = Vx * Δt, Δy = Vy * Δt). The current position and displacement are added together to obtain the position of the trace in the next frame (x' = x + Δx, y' = y + Δy). The extrapolated trace (x', y') represents the predicted position of the historical trace in the next frame. The lifespan of the extrapolated trace is increased by one frame period.

[0087] Preferably, the step of filtering the updated historical point set according to the radar frame period specifically involves: traversing each point in the extrapolated historical point set, setting a second life cycle threshold, and determining whether the life cycle of each point is less than or equal to the second life cycle threshold. If the life cycle of a point is less than or equal to the second life cycle threshold, then the point is retained in the historical point set; otherwise, the point is removed.

[0088] Preferably, the step of clustering the point cloud based on the preprocessed point traces and the filtered historical point traces specifically involves: combining the preprocessed point traces and the filtered historical point traces into a point cloud dataset, using point cloud DBSCAN clustering to cluster the point cloud dataset, dividing points with similar densities into different clusters, and identifying unvisited points as noise points.

[0089] Preferably, DBSCAN is a density-based clustering algorithm that can group points with adjacent densities into a cluster.

[0090] Preferably, the adjustment parameters of the DBSCAN clustering algorithm include at least the minimum number of samples (MinPts) and the neighborhood radius (ε).

[0091] In one embodiment, when performing DBSCAN clustering on the point cloud, the parameters are first initialized, specifying the neighborhood radius ε and the minimum number of samples MinPts as parameters of the DBSCAN algorithm, and randomly selecting an unvisited point P as the current point. The number of points in the neighborhood of the current point P is checked. If the number of points in the neighborhood is greater than or equal to the minimum number of points MinPts, the current point P is marked as a core point; otherwise, the current point P is marked as a noise point. If the current point P is a core point, an unvisited point is selected from the neighborhood of the current point P as the next point Q and added to a new cluster. The neighborhood of point Q is recursively explored. For points in the neighborhood, if they have not been visited, they are marked as visited and added to the current cluster. The above steps are repeated by continuing to select unvisited points as new current points until all points have been visited. For unmarked points, they are marked as noise points.

[0092] Preferably, the step of clustering the point cloud based on the preprocessed point traces and the filtered historical point trace set further includes: traversing each cluster in the point cloud dataset, calculating and outputting the mathematical mean center point of the cluster;

[0093] Specifically, for each cluster, the position accumulation variable and the point count variable are initialized to 0. Each point in the cluster is traversed. If a point matches the current cluster, the position information of the point is added to the initialized position accumulation variable, and the initialized point count variable is also incremented. The center point of the mathematical mean is equal to the position information of the accumulated point divided by the accumulated point count variable.

[0094] In one embodiment, for each category: initialize the position accumulation variables sum_x and sum_y to 0, and initialize the point count variable count to 0; for each point in the dataset: if the point's category matches the current category, accumulate the point's position information into sum_x and sum_y, and increment the point count count; calculate the position of the mathematical mean center point: if count is greater than 0, the x-coordinate of the mathematical mean center point is sum_x divided by count, and the y-coordinate of the mathematical mean center point is sum_y divided by count.

[0095] In another embodiment, for each category, the cluster center point coordinates center_x and center_y are initialized to 0, the current cluster center point coordinates new_center_x and new_center_y are initialized to the position of the random point trace of the current category, the convergence judgment variable converged is initialized to False, and the iteration counter iteration is initialized to 0.

[0096] The iterative update of cluster center positions is implemented as follows: When the convergence condition is not met and the number of iterations is less than the maximum number of iterations, the position accumulation variables sum_x and sum_y are initialized to 0, and the point count variable count is initialized to 0; for each point in the dataset, if the point's class matches the current class, the distance from the point to the current cluster center is calculated. If the distance between the point and the current cluster center is the shortest, the point's position information is accumulated in sum_x and sum_y, and the point count is incremented; if count is greater than 0, the new cluster center position new_center_x and new_center_y are calculated as sum_x divided by count; it is determined whether the distance between the new cluster center and the current cluster center is less than the convergence threshold. If it is, converged is set to True, and the iteration stops; otherwise, the position of the current cluster center is updated to the new cluster center position; the iteration counter is incremented.

[0097] The above calculation process uses an iterative approach to update the position of the cluster center point until the convergence condition is met or the maximum number of iterations is reached. By continuously updating the position of the cluster center point, the optimal mathematical mean center point can be found.

[0098] S5: Perform trajectory matching and tracking based on the results of the point cloud clustering.

[0099] Preferably, step S5 specifically includes:

[0100] S501: Initialize the track list by creating an empty track list to store track information in the current frame;

[0101] S502: Obtain the mathematical mean center point of each cluster;

[0102] S503: Traverse each track in the track list;

[0103] S504: Calculate the distance between the track and the center of the mathematical mean of the current cluster, and find the track with the shortest distance.

[0104] Among these methods, the shortest path is found by using appropriate distance metrics, such as Euclidean distance and Mahalanobis distance.

[0105] S505: By setting a maximum distance threshold, determine whether the shortest track is less than the maximum distance threshold. If the shortest track is less than the maximum distance threshold, the current cluster matches the track, and the status information of the track is updated. Otherwise, they do not match, and the unmatched cluster is regarded as a new track.

[0106] S506: Add the unmatched tracks to the track list to expand it;

[0107] Unmatched clusters are treated as new tracks and added to the track list for further tracking.

[0108] S507: Regularly clear outdated flight tracks;

[0109] Specifically, based on the track update time or other set outdated conditions, tracks that have exceeded a certain time or no longer meet the conditions are periodically cleaned up to ensure that the track list maintains the latest and most valid track information.

[0110] S6: Perform track management based on the track matching and tracking results.

[0111] Preferably, the track management includes track updates, track association, track generation and termination, track analysis and prediction, track storage and querying, etc.

[0112] Furthermore, the trajectory association is based on the trajectory matching results to establish the association relationship between trajectories, forming a trajectory set or trajectory group, and associating points belonging to the same target to form a complete motion trajectory.

[0113] Furthermore, the track generation and termination are based on the track matching and tracking results, generating new tracks or terminating invalid tracks. If a new unmatched point appears, it can be generated as a new track.

[0114] Furthermore, the trajectory analysis and prediction, by analyzing the trajectory's motion characteristics and behavioral patterns, can infer the target's intentions and predict its future trajectory.

[0115] Furthermore, the track storage and query function stores the updated track information in the track database and provides query and retrieval functions.

[0116] Example 2:

[0117] As another preferred embodiment, this application also provides a point cloud density enhancement system based on track negative feedback, such as... Figure 4 As shown, the system includes:

[0118] The spot acquisition module is used to acquire the spot in the current radar frame;

[0119] The dot preprocessing module removes invalid and jittery dots, transforms the dots in the current radar frame from the radar coordinate system to the target coordinate system, performs motion compensation and object feature extraction on the dot information, and obtains accurate and reliable dot information and continuous track information.

[0120] The point cloud clustering module is used to divide point cloud data into different clusters;

[0121] Historical track set update module; updates the historical track set based on stable motion tracks and radar cycles;

[0122] Track matching and tracking module: can continuously track and update the target's motion status;

[0123] The track management module is used to manage and maintain track information.

[0124] Example 3:

[0125] As another preferred embodiment, this application also provides a computer-readable storage medium, characterized in that it stores a computer program, which, when executed by a processor, implements a point cloud density enhancement method based on track negative feedback as described in any of the preceding claims.

[0126] In summary, this application proposes a point cloud density enhancement method and system based on track negative feedback. The method includes acquiring radar current frame point tracks and historical point track sets; preprocessing the radar current frame point tracks to obtain preprocessed point tracks; filtering stable moving tracks and updating the historical point track set; extrapolating and filtering the updated historical point track set according to the radar frame period; performing point cloud clustering based on the preprocessed point tracks and the filtered historical point track set; performing track matching and tracking based on the point cloud clustering results; and performing track management based on the track matching and tracking results. This application increases point cloud density, improves track stability, and enhances track tracking accuracy by retaining historical point tracks corresponding to tracked stable moving tracks and extrapolating their positions based on their motion states.

[0127] Although exemplary embodiments have been described herein with reference to the accompanying drawings, it should be understood that the above exemplary embodiments are merely illustrative and are not intended to limit the scope of this application. Various changes and modifications can be made therein by those skilled in the art without departing from the scope and spirit of this application. All such changes and modifications are intended to be included within the scope of this application as claimed in the appended claims.

[0128] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0129] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed.

[0130] The various component embodiments of this application can be implemented in hardware, or as software modules running on one or more processors, or a combination thereof. Those skilled in the art will understand that microprocessors or digital signal processors (DSPs) can be used in practice to implement some or all of the functions of some modules according to the embodiments of this application. This application can also be implemented as an apparatus program (e.g., a computer program and computer program product) for performing part or all of the methods described herein. Such an implementation of this application can be stored on a computer-readable medium, or can be in the form of one or more signals. Such signals can be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.

Claims

1. A point cloud density enhancement method based on track negative feedback, characterized in that, include: S1: Obtain radar current frame traces and historical trace sets; S2: Preprocess the radar current frame traces to obtain preprocessed traces; S3: Filter stable motion tracks and update the historical point set; S4: Extrapolate and filter the updated historical point set according to the radar frame period, and perform point cloud clustering based on the preprocessed point set and the filtered historical point set. S5: Perform trajectory matching and tracking based on the results of the point cloud clustering; S6: Perform track management based on the track matching and tracking results; The point cloud clustering based on the preprocessed point traces and the filtered historical point trace set specifically involves: The preprocessed point traces and the filtered historical point traces are combined into a point cloud dataset. The point cloud dataset is clustered using point cloud DBSCAN clustering. Points with similar density are divided into different clusters, and unvisited points are identified as noise points. In the DBSCAN clustering process for point clouds, the parameters are initialized, specifying the neighborhood radius ε and the minimum number of samples MinPts as parameters for the DBSCAN algorithm. An unvisited point P is randomly selected as the current point. The number of points in the neighborhood of the current point P is checked. If the number of points in the neighborhood is greater than or equal to the minimum number of points MinPts, the current point P is marked as a core point; otherwise, the current point P is marked as a noise point. If the current point P is a core point, an unvisited point is selected from the neighborhood of the current point P as the next point Q and added to a new cluster. The neighborhood of point Q is recursively explored. For points in the neighborhood that have not been visited, they are marked as visited and added to the current cluster. The process continues by selecting unvisited points as the new current point and repeating the above steps until all points have been visited. Unmarked points are marked as noise points. Specifically, S5 is: S501: Initialize the track list by creating an empty track list to store aircraft information in the current frame; S502: Obtain the mathematical mean center point of each cluster; S503: Traverse each track in the track list; S504: Calculate the distance between the track and the center of the mathematical mean of the current cluster, and find the track with the shortest distance. S505: By setting a maximum distance threshold, determine whether the shortest track is less than the maximum distance threshold. If the shortest track is less than the maximum distance threshold, the current cluster matches the track, and the status information of the track is updated. Otherwise, they do not match, and the unmatched cluster is regarded as a new track. S506: Add the unmatched tracks to the track list to expand it; S507: Regularly clear outdated flight tracks.

2. The point cloud density enhancement method based on track negative feedback according to claim 1, characterized in that, The preprocessing specifically includes: Invalid and jittery dots are removed, and the dots in the current radar frame are transformed from the radar coordinate system to the target coordinate system. Motion compensation and object feature extraction are then performed on the dots.

3. The point cloud density enhancement method based on track negative feedback according to claim 2, characterized in that, Specifically, S3 is: S301: Obtain the point trace corresponding to the stable motion track; S302: Calculate the lifetime of the point trace based on its existence time or the number of frames continuously observed; wherein, the lifetime of the point trace refers to the number of frames in which the point trace exists in consecutive frames; S303: Set a first lifecycle threshold and filter tracks whose lifecycle is greater than the lifecycle threshold; S304: Update the point information corresponding to the track that is greater than the life cycle threshold to the historical point set; wherein, the point information includes at least position and velocity; S305: Set the lifetime of the track points that are greater than the lifetime threshold to 1.

4. The point cloud density enhancement method based on track negative feedback according to claim 3, characterized in that, The step of extrapolating the updated historical point set according to the radar frame period is specifically as follows: S401: Obtain the radar frame period and the current position and velocity of each point in the updated historical point set; S402: Based on the product of the velocity and the radar frame period interval, the displacement of each point within the radar frame period interval is obtained; S403: Add the current position and displacement of the point to get the position of the point in the next frame, and update it to the historical point set; S404: The lifetime of each point in the extrapolated historical point set is increased by 1.

5. The point cloud density enhancement method based on track negative feedback according to claim 4, characterized in that, The step of filtering the updated historical point set according to the radar frame period is as follows: Iterate through each point in the extrapolated historical point set, set a second lifespan threshold, and determine whether the lifespan of each point is less than or equal to the second lifespan threshold. If the lifespan of a point is less than or equal to the second lifespan threshold, then retain the point in the historical point set; otherwise, remove the point.

6. The point cloud density enhancement method based on track negative feedback according to claim 5, characterized in that, The step of clustering point clouds based on the preprocessed point traces and the filtered historical point trace set further includes: Iterate through each cluster in the point cloud dataset, calculate and output the mathematical mean center point of the cluster; Specifically, for each cluster, the position accumulation variable and the point count variable are initialized to 0. Each point in the cluster is traversed. If a point matches the current cluster, the position information of the point is added to the initialized position accumulation variable, and the initialized point count variable is also incremented. The center point of the mathematical mean is equal to the position information of the accumulated point divided by the accumulated point count variable.

7. A system for point cloud density enhancement based on track negative feedback according to any one of claims 1-6, characterized in that, include: The spot acquisition module is used to acquire the spot in the current radar frame; The dot preprocessing module removes invalid and jittery dots, transforms the dots in the current radar frame from the radar coordinate system to the target coordinate system, performs motion compensation and object feature extraction on the dot information, and obtains accurate and reliable dot information and continuous track information. The point cloud clustering module is used to divide point cloud data into different clusters; Historical track set update module; updates the historical track set based on stable motion tracks and radar cycles; Track matching and tracking module: can continuously track and update the target's motion status; The track management module is used to manage and maintain track information.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements a point cloud density enhancement method based on track negative feedback as described in any one of claims 1 to 6.

Images