Real-time clustering method based on 4D imaging millimeter wave radar point cloud

Abstract

The invention relates to the technical field of intelligent driving, and particularly discloses a real-time clustering method based on a 4D imaging millimeter wave radar point cloud, which comprises the following steps: acquiring all target point cloud information of a vehicle surrounding environment acquired by a 4D imaging millimeter wave radar; rasterizing and dividing all the target point clouds of the surrounding environment of the vehicle so as to divide all the target point clouds of the surrounding environment of the vehicle into corresponding grids; calculating the target point cloud density in each grid, and screening out candidate grids of which the target point cloud density reaches a standard value from all grids; combining all the screened candidate grids to obtain a plurality of combined grids, and generating a corresponding point cluster for all the target point clouds in each combined grid; and screening out reasonable point clusters from the plurality of generated point clusters, and outputting a clustering result of each reasonable point cluster. According to the invention, clustering attributes can be provided for a subsequent target tracking algorithm and a target classification algorithm, and the subsequent algorithm is simplified.

Description

Technical Field

[0001] The present invention relates to the field of intelligent driving technology, and more specifically, to a real-time clustering method based on 4D imaging millimeter wave radar point cloud. Background Art

[0002] The point cloud resolution of traditional 3D millimeter wave radar is low and it does not have the ability to measure height. The mainstream DBSCAN clustering algorithm can achieve the clustering effect in real time. Clustering of point clouds is the basis for subsequent target tracking algorithms and target classification algorithms.

[0003] The point cloud resolution of 4D imaging millimeter wave radar has been greatly improved, and it has the ability to measure height, and the point cloud has increased in magnitude. Therefore, the mainstream DBSCAN method cannot complete the real-time clustering of thousands of points within the specified time. Since 4D imaging millimeter wave radar has only been around for a short time, there is currently no real-time clustering algorithm based on 4D imaging millimeter wave radar point cloud. Summary of the invention

[0004] The purpose of the present invention is to overcome the deficiencies in the prior art and provide a real-time clustering method based on 4D imaging millimeter-wave radar point cloud to provide point cloud clustering information for subsequent target tracking algorithms and target classification algorithms.

[0005] As a first aspect of the present invention, a real-time clustering method based on 4D imaging millimeter wave radar point cloud is provided, and the real-time clustering method based on 4D imaging millimeter wave radar point cloud comprises the following steps:

[0006] Step S11: Acquire all target point cloud information of the vehicle's surrounding environment collected by the 4D imaging millimeter wave radar;

[0007] Step S12: performing rasterization division processing on all target point clouds of the vehicle surrounding environment according to all target point cloud information of the vehicle surrounding environment, so as to divide all target point clouds of the vehicle surrounding environment into corresponding grids;

[0008] Step S13: Calculate the target point cloud density in each grid, and select candidate grids whose target point cloud density reaches the standard value from all grids;

[0009] Step S14: merging all the selected candidate grids to obtain a plurality of merged grids, and generating a corresponding point cluster from all the target point clouds in each of the merged grids;

[0010] Step S15: selecting a reasonable point cluster from the generated multiple point clusters, and outputting a clustering result of each reasonable point cluster; wherein the clustering result of the reasonable point cluster includes clustering information of the reasonable point cluster and all target point cloud information of the reasonable point cluster.

[0011] Furthermore, the rasterization division process is performed on all target point clouds of the vehicle surrounding environment according to all target point cloud information of the vehicle surrounding environment so as to divide all target point clouds of the vehicle surrounding environment into corresponding grids, and further includes:

[0012] According to all target point cloud information of the vehicle surrounding environment, determining whether all target point clouds of the vehicle surrounding environment meet angle clustering conditions;

[0013] If the angle clustering condition is met, all target point clouds are rasterized and divided using the horizontal angle, pitch angle and Euclidean distance of all target point clouds as attributes, so as to divide all target point clouds of the vehicle surrounding environment into corresponding grids in the polar coordinate system;

[0014] If the angle clustering condition is not met, all target point clouds are rasterized and divided using the three-dimensional coordinates of all target point clouds as attributes, so as to divide all target point clouds of the vehicle surrounding environment into corresponding grids in a Cartesian coordinate system.

[0015] Furthermore, the judgment formula of the angle clustering condition is as follows:

[0016]

[0017] Among them, R max is the maximum detection distance of 4D imaging millimeter wave radar, R1, R2, ..., R n is the Euclidean distance from each target point cloud to the 4D imaging millimeter wave radar, Pow i is the intensity of the i-th target point cloud, and n is the number of all target point clouds collected by the 4D imaging millimeter-wave radar;

[0018] When the calculated value b is greater than a preset value, it is determined that all target point clouds of the vehicle's surrounding environment meet the angle clustering condition; otherwise, it is determined that all target point clouds of the vehicle's surrounding environment do not meet the angle clustering condition.

[0019] Furthermore, the step of merging all the selected candidate grids to obtain a plurality of merged grids further includes:

[0020] Perform a breadth-first search on each candidate grid to search for its adjacent candidate grids;

[0021] The candidate grid and its adjacent candidate grids are merged to obtain a merged grid.

[0022] Furthermore, the clustering information of the reasonable point cluster includes a maximum circumscribed rectangle, an average intensity of the target point cloud, and a standard deviation of the intensity of the target point cloud.

[0023] The real-time clustering method based on 4D imaging millimeter wave radar point cloud provided by the present invention has the following advantages:

[0024] (1) Compared with the traditional 3D millimeter-wave radar, the 4D imaging millimeter-wave radar based on the present invention has significantly improved point cloud resolution, and has height measurement capability, and greatly improved the ability to describe the target;

[0025] (2) The radar point cloud clustering algorithm based on the present invention can provide clustering attributes for subsequent target tracking algorithms and target classification algorithms, thereby simplifying subsequent algorithms;

[0026] (3) The present invention is based on the time-consuming optimization of the 4D imaging millimeter-wave radar point cloud clustering algorithm. It can provide clustering results at a speed of 30 frames per second even when the number of point clouds reaches thousands, ensuring real-time performance, and providing possibilities for subsequent real-time real-car autonomous driving solutions. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. Together with the following specific embodiments, they are used to explain the present invention, but do not constitute a limitation of the present invention.

[0028] Figure 1 A flowchart of a real-time clustering method based on 4D imaging millimeter-wave radar point cloud provided by the present invention.

[0029] Figure 2 This is a flowchart of a specific implementation of the real-time clustering method based on 4D imaging millimeter wave radar point cloud provided by the present invention. DETAILED DESCRIPTION

[0030] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation method, structure, characteristics and effects of a real-time clustering method based on 4D imaging millimeter wave radar point cloud proposed by the present invention are described in detail below in combination with the accompanying drawings and preferred embodiments. Obviously, the described embodiments are part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0031] In this embodiment, a real-time clustering method based on 4D imaging millimeter wave radar point cloud is provided. Figure 1 As shown, the real-time clustering method based on 4D imaging millimeter wave radar point cloud includes the following steps:

[0032] Step S11: Acquire all target point cloud information of the vehicle's surrounding environment collected by the 4D imaging millimeter wave radar;

[0033] It should be noted that the 4D imaging millimeter-wave radar is installed on the front bumper of the vehicle.

[0034] Step S12: performing rasterization division processing on all target point clouds of the vehicle surrounding environment according to all target point cloud information of the vehicle surrounding environment, so as to divide all target point clouds of the vehicle surrounding environment into corresponding grids;

[0035] Preferably, if Figure 2 As shown, the method of performing rasterization division processing on all target point clouds of the vehicle surrounding environment according to all target point cloud information of the vehicle surrounding environment to divide all target point clouds of the vehicle surrounding environment into corresponding grids also includes:

[0036] According to all target point cloud information of the vehicle surrounding environment, determining whether all target point clouds of the vehicle surrounding environment meet angle clustering conditions;

[0037] If the angle clustering condition is met, all target point clouds are rasterized and divided using the horizontal angle, pitch angle and Euclidean distance of all target point clouds as attributes, so as to divide all target point clouds of the vehicle surrounding environment into corresponding grids in the polar coordinate system;

[0038] It should be noted that the rasterization division process can adapt to different target point clouds and can effectively summarize point cloud information with limited data. Among them, the horizontal angle refers to the angle with the radar as the center, the vehicle forward direction as 0°, and the top angle as the positive direction clockwise. The pitch angle refers to the angle with the radar as the center, the horizontal plane as 0°, and the left side perspective as the positive direction clockwise. The Euclidean distance refers to the Euclidean distance from the target point cloud to the radar.

[0039] It should be noted that the angle rasterization is different from the conventional 3D rasterization, and the grid size in Euclidean space is not uniform. However, the grid size in polar coordinates is uniform, so this method is more reliable for processing radar information that outputs point clouds in polar coordinates.

[0040] If the angle clustering condition is not met, all target point clouds are rasterized and divided with the three-dimensional coordinates of all target point clouds as attributes, so as to divide all target point clouds around the vehicle into corresponding grids in the Cartesian coordinate system, wherein the grid at this time is uniform in the Euclidean space. Among them, the three-dimensional coordinates of all target point clouds are X, Y, and Z coordinates, wherein the X coordinate refers to the lateral distance centered on the radar and the right side of the vehicle is positive, the Y coordinate refers to the forward distance centered on the radar and the forward direction of the vehicle is positive, and the Z coordinate refers to the height distance centered on the radar and the horizontal plane is positive upward.

[0041] Specifically, the judgment formula of the angle clustering condition is as follows:

[0042]

[0043] Among them, R max is the maximum detection distance of 4D imaging millimeter wave radar, R1, R2, ..., R n is the Euclidean distance from each target point cloud to the 4D imaging millimeter wave radar, Pow i is the intensity of the i-th target point cloud, and n is the number of all target point clouds collected by the 4D imaging millimeter-wave radar;

[0044] When the calculated value b is greater than a preset value, it is determined that all target point clouds of the vehicle's surrounding environment meet the angle clustering condition; otherwise, it is determined that all target point clouds of the vehicle's surrounding environment do not meet the angle clustering condition.

[0045] It should be noted that the intensity used here is real intensity, which is different from the logarithmic intensity used by some millimeter-wave radars. It is common to use angle clustering when b>0.5.

[0046] Step S13: Calculate the target point cloud density in each grid, and select candidate grids whose target point cloud density reaches the standard value from all grids;

[0047] It should be noted that when using angle gridding, the calculated grid point cloud density is the target point cloud density of the grid in the polar coordinate system. When using Cartesian coordinate gridding, the calculated grid point cloud density is the target point cloud density of the grid in the Cartesian coordinate system. The target point cloud density can be calculated using the weighted statistical method of the intensity of each target point cloud pair.

[0048] Step S14: merging all the selected candidate grids to obtain a plurality of merged grids, and generating a corresponding point cluster from all the target point clouds in each of the merged grids;

[0049] Preferably, the step of merging all the selected candidate grids to obtain a plurality of merged grids further includes:

[0050] Perform a breadth-first search on each candidate grid to search for its adjacent candidate grids; wherein the adjacent judgment condition may be the Euclidean distance between two candidate grids;

[0051] The candidate grid and its adjacent candidate grids are merged to obtain a merged grid.

[0052] It should be noted that a breadth-first search is performed on the selected candidate grids, and then it is determined whether the merging condition is met. The judgment condition used in the present invention is whether two candidate grids are connected, the candidate grids that meet the judgment condition are merged, and the target point cloud in the candidate grids that meet the condition is obtained, and the target point cloud information is recorded, and these target point clouds are remapped to point clusters.

[0053] Step S15: Filter out reasonable point clusters from the generated multiple point clusters, and output the clustering results of each reasonable point cluster; wherein the clustering results of the reasonable point clusters include the clustering information of the reasonable point cluster and all target point cloud information of the reasonable point cluster. wherein the clustering information of the reasonable point cluster includes the maximum circumscribed rectangle, the average intensity of the target point cloud, and the standard deviation of the intensity of the target point cloud.

[0054] The present invention provides a real-time clustering method based on 4D imaging millimeter wave radar point cloud, which accesses millimeter wave radar point cloud information, vehicle body signals, radar installation information, etc., provides point cloud input for clustering algorithm, and finally realizes clustering function. (1) Compared with traditional 3D millimeter wave radar, the 4D imaging millimeter wave radar based on the present invention has significantly improved point cloud resolution, and has height measurement capability, which greatly improves the ability to describe targets; (2) The radar point cloud clustering algorithm based on the present invention can provide clustering attributes for subsequent target tracking algorithms and target classification algorithms, simplifying subsequent algorithms; (3) The present invention is based on the time-consuming optimization of the 4D imaging millimeter wave radar point cloud clustering algorithm, and can provide clustering results at a speed of 30 frames per second when the number of point clouds reaches thousands, ensuring real-time performance, and providing possibilities for subsequent real-time real-car autonomous driving solutions.

[0055] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technician familiar with this profession can make some changes or modifications to equivalent embodiments of equivalent changes using the technical contents disclosed above without departing from the scope of the technical solution of the present invention. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention still falls within the scope of the technical solution of the present invention.

Claims

1. A real-time clustering method based on 4D imaging millimeter wave radar point cloud, characterized in that: The real-time clustering method based on 4D imaging millimeter wave radar point cloud comprises the following steps: Step S11: Acquire all target point cloud information of the vehicle's surrounding environment collected by the 4D imaging millimeter wave radar; Step S12: performing rasterization division processing on all target point clouds of the vehicle surrounding environment according to all target point cloud information of the vehicle surrounding environment, so as to divide all target point clouds of the vehicle surrounding environment into corresponding grids; Step S13: Calculate the target point cloud density in each grid, and select candidate grids whose target point cloud density reaches the standard value from all grids; Step S14: merging all the selected candidate grids to obtain a plurality of merged grids, and generating a corresponding point cluster from all the target point clouds in each of the merged grids; Step S15: selecting a reasonable point cluster from the generated multiple point clusters, and outputting a clustering result of each reasonable point cluster; wherein the clustering result of the reasonable point cluster includes clustering information of the reasonable point cluster and all target point cloud information of the reasonable point cluster.

2. The real-time clustering method based on 4D imaging millimeter wave radar point cloud according to claim 1, characterized in that: The method of performing rasterization division processing on all target point clouds of the vehicle surrounding environment according to all target point cloud information of the vehicle surrounding environment so as to divide all target point clouds of the vehicle surrounding environment into corresponding grids further includes: According to all target point cloud information of the vehicle surrounding environment, determining whether all target point clouds of the vehicle surrounding environment meet angle clustering conditions; If the angle clustering condition is met, all target point clouds are rasterized and divided using the horizontal angle, pitch angle and Euclidean distance of all target point clouds as attributes, so as to divide all target point clouds of the vehicle surrounding environment into corresponding grids in the polar coordinate system; If the angle clustering condition is not met, all target point clouds are rasterized and divided using the three-dimensional coordinates of all target point clouds as attributes, so as to divide all target point clouds of the vehicle surrounding environment into corresponding grids in a Cartesian coordinate system.

3. The real-time clustering method based on 4D imaging millimeter wave radar point cloud according to claim 2, characterized in that: The judgment formula of the angle clustering condition is as follows: Among them, R max is the maximum detection distance of 4D imaging millimeter wave radar, R1, R2, ..., R n is the Euclidean distance from each target point cloud to the 4D imaging millimeter wave radar, Pow i is the intensity of the i-th target point cloud, and n is the number of all target point clouds collected by the 4D imaging millimeter-wave radar; When the calculated value b is greater than a preset value, it is determined that all target point clouds of the vehicle's surrounding environment meet the angle clustering condition; otherwise, it is determined that all target point clouds of the vehicle's surrounding environment do not meet the angle clustering condition.

4. The real-time clustering method based on 4D imaging millimeter wave radar point cloud according to claim 1, characterized in that: The step of merging all the selected candidate grids to obtain a plurality of merged grids further includes: Perform a breadth-first search on each candidate grid to search for its adjacent candidate grids; The candidate grid and its adjacent candidate grids are merged to obtain a merged grid.

5. The real-time clustering method based on 4D imaging millimeter wave radar point cloud according to claim 1, characterized in that: The clustering information of the reasonable point cluster includes the maximum circumscribed rectangle, the average intensity of the target point cloud, and the standard deviation of the intensity of the target point cloud.

Images