A multi-target motion closed-loop point cloud clustering method based on millimeter wave radar
By setting a threshold for the intersection area and selecting an appropriate clustering method to process millimeter-wave radar point cloud data, the problem of overlapping trajectories of multiple targets was solved, achieving higher-precision target trajectory prediction and tracking.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHENZHEN HUAJIE ZHITONG TECH CO LTD
- Filing Date
- 2022-11-28
- Publication Date
- 2026-07-14
AI Technical Summary
In existing multi-target trajectory research based on millimeter-wave radar, moving targets are prone to cross-over and overlap, leading to missed detections and false detections in clustering methods, making it impossible to form closed loops and affecting the accuracy of target trajectory prediction and tracking.
By setting a threshold for the intersection area of targets, the distance between the motion trajectories of multiple targets is calculated. A suitable clustering method is selected to cluster the point cloud data and establish a closed-loop feedback of motion state, including primary clustering and secondary clustering methods, to distinguish between intersection and non-intersection areas for point cloud data processing.
It improves the clustering accuracy of multi-target trajectories in the intersection area, avoids target merging errors, and enhances the accuracy of multi-moving target trajectory prediction and tracking.
Smart Images
Figure CN115755016B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of millimeter-wave radar applications, and in particular to a multi-target motion closed-loop point cloud clustering method based on millimeter-wave radar. Background Technology
[0002] Millimeter-wave radar is a detection radar in the millimeter-wave band. Due to its small size, light weight and high spatial resolution, millimeter-wave radar is widely used in fields such as home network cameras and vehicle monitoring networks.
[0003] Current research on target trajectory based on millimeter-wave radar primarily utilizes the radar's multi-transmitter, multi-receiver antennas to collect distance, velocity, angle, and Doppler information of moving targets. This information is then converted into point clouds using FFT and 2DFFT transformations. Clustering methods are used to group the point clouds into features, which are then input frame-by-frame into the existing model to ultimately predict and track the target trajectory.
[0004] However, current clustering methods still have shortcomings. When the number of moving targets in the detection scene increases, any target is more likely to overlap with other targets during its movement. Existing clustering methods only cluster point clouds without forming a closed loop with the target's motion state, thus clustering multiple targets into a single target, leading to missed detections and false detections. Therefore, a multi-target motion closed-loop point cloud clustering method is urgently needed. Summary of the Invention
[0005] The purpose of this invention is to provide a multi-target motion closed-loop point cloud clustering method based on millimeter-wave radar, so as to solve the problem of missed detection and false detection in existing clustering methods when multiple target trajectories overlap and intersect.
[0006] To address the aforementioned technical problems, this invention provides a multi-target motion closed-loop point cloud clustering method based on millimeter-wave radar, comprising the following steps:
[0007] S100: Uses millimeter-wave radar to monitor moving scenes, collects and reads point cloud data frame by frame;
[0008] S200: Set the judgment conditions and detect and judge the trajectories of multiple moving targets in the current frame;
[0009] S300. Based on the judgment result, select a clustering method to cluster the point cloud data;
[0010] S400: Combine the clustered point cloud data to reconstruct the complete frame data;
[0011] S500: Update the target trajectory using the reconstructed frame data;
[0012] S600. Determine the new target trajectory, and the determination conditions are the same as those in step S100. For new target trajectories that are determined to enter the intersection area, return to step S200. For new target trajectories that are determined not to enter the intersection area, clear the intersection area of the current frame.
[0013] Furthermore, the determination condition is: setting a threshold for the target intersection area.
[0014] Furthermore, the threshold value of the intersection area is 0.1 to 1 m.
[0015] Furthermore, the specific method for detection and determination is as follows:
[0016] Calculate the distance between the motion trajectories of multiple moving targets and compare it with the threshold of the intersection area;
[0017] When the trajectory distance between targets is greater than the intersection area threshold, it is determined that multiple targets have not entered the intersection area;
[0018] When the trajectory distance between targets is less than the threshold of the intersection area, it is determined that multiple targets have entered the intersection area, and the current frame point cloud data is divided into the intersection area point cloud set and the non-intersection area point cloud set.
[0019] Furthermore, the clustering method includes a primary clustering method and a secondary clustering method, and the selection of the clustering method is specifically as follows:
[0020] When multiple targets are determined not to have entered the intersection region, the main clustering method is selected;
[0021] When multiple targets are determined to have entered the intersection region, the secondary clustering method is selected for point cloud sets classified as intersection regions, and the primary clustering method is selected for point cloud sets classified as non-intersection regions.
[0022] Furthermore, the clustering range of the primary clustering method is greater than that of the secondary clustering method.
[0023] Furthermore, the specific method for determining the new target trajectory is as follows:
[0024] If the distance between the new target trajectories is less than the intersection area threshold, that is, the new target trajectory is in the intersection area, then update the intersection area and return to step S200;
[0025] If the distance between new target trajectories is greater than the intersection area threshold, i.e., the new target trajectory is not in the intersection area, then the intersection area of the current frame is cleared.
[0026] Compared with the prior art, the present invention has at least the following beneficial effects:
[0027] The system collects the trajectories of multiple targets around millimeter-wave radar application scenarios, calculates the distance between the trajectories of multiple moving targets, compares the results with a predefined threshold for the intersection area of targets, and divides the data into different regions. Based on the regions, an appropriate clustering method is selected to cluster the point cloud data, and a closed-loop feedback between the clustering method and the motion state is established. This greatly improves the clustering accuracy of targets in the intersection area, avoids merging close targets into one target, and avoids clustering large targets into multiple targets, effectively improving the accuracy of multi-moving target trajectory prediction and tracking. Attached Figure Description
[0028] Figure 1 This is a flowchart illustrating the steps of the multi-target motion closed-loop point cloud clustering method of the present invention;
[0029] Figure 2 This is a flowchart of the multi-target motion closed-loop point cloud clustering method of the present invention. Detailed Implementation
[0030] The multi-target motion closed-loop point cloud clustering method based on millimeter-wave radar of the present invention will be described in more detail below with reference to the schematic diagrams, which illustrate preferred embodiments of the present invention. It should be understood that those skilled in the art can modify the present invention described herein while still achieving the advantageous effects of the present invention. Therefore, the following description should be understood as being of broad knowledge to those skilled in the art and is not intended to limit the present invention.
[0031] The invention is described more specifically by way of example in the following paragraphs with reference to the accompanying drawings. The advantages and features of the invention will become clearer from the following description and claims. It should be noted that the drawings are in a very simplified form and use non-precise proportions, and are only used to facilitate and clarify the illustration of the embodiments of the invention.
[0032] like Figure 1 As shown in the figure, this invention proposes a multi-target motion closed-loop point cloud clustering method based on millimeter-wave radar, which includes the following steps:
[0033] S100: Uses millimeter-wave radar to monitor moving scenes, collects and reads point cloud data frame by frame;
[0034] S200: Set the judgment conditions and detect and judge the trajectories of multiple moving targets in the current frame;
[0035] S300. Based on the judgment result, select a clustering method to cluster the point cloud data;
[0036] S400: Combine the clustered point cloud data to reconstruct the complete frame data;
[0037] S500: Update the target trajectory using the reconstructed frame data;
[0038] S600. Determine the new target trajectory, and the determination conditions are the same as those in step one. For new target trajectories that are determined to enter the intersection area, return to step two. For new target trajectories that are determined not to enter the intersection area, clear the intersection area of the current frame.
[0039] The following are preferred embodiments of the multi-target motion closed-loop point cloud clustering method based on millimeter-wave radar to clearly illustrate the content of the present invention. It should be understood that the content of the present invention is not limited to the following embodiments, and other improvements made by conventional technical means by those skilled in the art are also within the scope of the present invention.
[0040] S100 uses millimeter-wave radar to monitor moving scenes, collects and reads point cloud data frame by frame.
[0041] In this embodiment, based on the application scenario of millimeter-wave radar, multi-target trajectories are collected for surrounding moving targets, such as pedestrians, vehicles, or animals, and the distance between the trajectories of the multiple targets is calculated.
[0042] S200: Set the judgment conditions and detect and judge the trajectory of multiple moving targets in the current frame.
[0043] Specifically, the detection and determination method is as follows:
[0044] The determination condition is: setting a threshold for the target intersection area;
[0045] Calculate the distance between the motion trajectories of multiple moving targets and compare it with the threshold of the intersection area;
[0046] When the trajectory distance between targets is greater than or equal to the threshold of the intersection area, it is determined that multiple targets have not entered the intersection area;
[0047] When the trajectory distance between targets is less than the threshold of the intersection area, it is determined that multiple targets have entered the intersection area, and the current frame point cloud data is divided into the intersection area point cloud set and the non-intersection area point cloud set.
[0048] In this embodiment, the target intersection area threshold is determined based on the moving targets around the millimeter radar. For example, if the target is a car, the range of a large car is very long, so the target intersection area threshold should be increased; if the target is a person, the range of a person is relatively small, so the target intersection area threshold should be decreased.
[0049] S300. Based on the judgment result, select a clustering method to cluster the point cloud data.
[0050] Specifically, the clustering method includes a primary clustering method and a secondary clustering method, and the selection of the clustering method is as follows:
[0051] When multiple targets are determined not to have entered the intersection region, the main clustering method is selected;
[0052] When multiple targets are determined to have entered an intersection region, for point cloud sets classified as intersection regions, a secondary clustering method is selected; for point cloud sets classified as non-intersection regions, the primary clustering method is selected.
[0053] S400 combines the clustered point cloud data to reconstruct the complete frame data.
[0054] S500 updates the target trajectory using the reconstructed frame data.
[0055] Specifically, the main clustering method used by S500 for clustered point clouds is the same as that used in S300. Therefore, only clustered point cloud data using secondary clustering methods can update the target trajectory.
[0056] S600. Determine the new target trajectory, and the determination conditions are the same as those in step one. For new target trajectories that are determined to enter the intersection area, return to step S200. For new target trajectories that are determined not to enter the intersection area, clear the intersection area of the current frame.
[0057] Specifically, if the distance between the new target trajectories is less than the intersection area threshold, that is, the new target trajectory is in the intersection area, then the intersection area is updated and the process returns to step S200.
[0058] If the distance between new target trajectories is greater than the intersection area threshold, i.e., the new target trajectory is not in the intersection area, then the intersection area of the current frame is cleared.
[0059] In this embodiment, when step S600 determines that the new target trajectory is not in the intersection area, it means that the new target trajectory will not have an intersection event, so such target trajectory does not need to enter the closed loop, and thus such target trajectory is cleared; when step S600 determines that the new target trajectory is in the intersection area, the new target is updated and the process returns to step two to enter the closed loop feedback.
[0060] Furthermore, the clustering range of the primary clustering method is larger than that of the secondary clustering method. The primary clustering method focuses on avoiding clustering large targets into multiple targets, while the secondary clustering method has a smaller range, focusing on avoiding clustering small, closely spaced targets into a single target. By selecting different clustering methods based on the different ranges of moving targets, the clustering accuracy is improved.
[0061] In summary, the beneficial effects of this invention are as follows: it collects the trajectories of multiple targets around the application scenario of millimeter-wave radar, calculates the distance between the trajectories of multiple moving targets, compares the results with a pre-set threshold for the intersection area of targets, divides the data into different regions, selects an appropriate clustering method based on the region to cluster the point cloud data, and establishes a closed-loop feedback between the clustering method and the motion state, thereby greatly improving the clustering accuracy of targets in the intersection area, avoiding merging close targets into one target, and avoiding clustering large targets into multiple targets, effectively improving the accuracy of multi-moving target trajectory prediction and tracking.
[0062] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.
Claims
1. A multi-target motion closed-loop point cloud clustering method based on millimeter-wave radar, characterized in that, Includes the following steps: S100: Uses millimeter-wave radar to monitor moving scenes, collects and reads point cloud data frame by frame; S200. Set judgment conditions and detect and judge the trajectories of multiple moving targets in the current frame. The judgment conditions are: setting a threshold for the target intersection area; the specific method for detection and judgment is as follows: Calculate the distance between the motion trajectories of multiple moving targets and compare it with the threshold of the intersection area; When the trajectory distance between targets is greater than the intersection area threshold, it is determined that multiple targets have not entered the intersection area; When the trajectory distance between targets is less than the threshold of the intersection area, it is determined that multiple targets have entered the intersection area, and the current frame point cloud data is divided into the intersection area point cloud set and the non-intersection area point cloud set. S300. Based on the judgment result, select a clustering method to cluster the point cloud data; S400: Combine the clustered point cloud data to reconstruct the complete frame data; S500: Update the target trajectory using the reconstructed frame data; S600. Determine the new target trajectory, and the determination conditions are the same as those in step S200. For new target trajectories that are determined to enter the intersection area, return to step S200. For new target trajectories that are determined not to enter the intersection area, clear the intersection area of the current frame.
2. The multi-target motion closed-loop point cloud clustering method based on millimeter-wave radar as described in claim 1, characterized in that, The threshold for the intersection area is 0.1~1m.
3. The multi-target motion closed-loop point cloud clustering method based on millimeter-wave radar as described in claim 1, characterized in that, The clustering method includes a primary clustering method and a secondary clustering method, and the selection of the clustering method is specifically as follows: When multiple targets are determined not to have entered the intersection region, the main clustering method is selected; When multiple targets are determined to have entered the intersection region, the secondary clustering method is selected for point cloud sets classified as intersection regions, and the primary clustering method is selected for point cloud sets classified as non-intersection regions.
4. The multi-target motion closed-loop point cloud clustering method based on millimeter-wave radar as described in claim 3, characterized in that, The clustering range of the primary clustering method is greater than that of the secondary clustering method.
5. The multi-target motion closed-loop point cloud clustering method based on millimeter-wave radar as described in claim 1, characterized in that, The specific method for determining the new target trajectory is as follows: If the distance between the new target trajectories is less than the intersection area threshold, that is, the new target trajectory is in the intersection area, then update the intersection area and return to step S200; If the distance between new target trajectories is greater than the intersection area threshold, i.e., the new target trajectory is not in the intersection area, then the intersection area of the current frame is cleared.