A 4D millimeter-wave radar clustering method, device and electronic device

Abstract

The present invention belongs to the technical field of point cloud data processing, and relates to a 4D millimeter-wave radar clustering method, device and electronic device. The method includes inputting 4D millimeter-wave radar point cloud data; setting a first neighborhood radius and the minimum number of points in the point cloud within the first neighborhood radius; using the DBSCAN algorithm to perform point cloud density clustering to obtain the point cloud density clustering results of each target area; calculating the center point positions of each point cloud density clustering and counting the widths of each point cloud density clustering; setting a second neighborhood radius according to the widths of each point cloud density clustering to obtain a set of second neighborhood radii; and performing DBSCAN clustering on each point cloud density clustering according to the set second neighborhood radius to complete the point cloud clustering operation. The present invention can adapt to target point cloud data of different sizes, and adaptively adjust the neighborhood radius of large-size clustering according to the point cloud width, thereby improving the accuracy of point cloud clustering of 4D millimeter-wave radar.

Description

Technical Field

[0001] The present invention belongs to the technical field of point cloud data processing. Specifically, it relates to a 4D millimeter-wave radar clustering method, device, and electronic device. Background Art

[0002] Due to the sparsity of millimeter-wave radar point clouds and the insufficient continuity of the point clouds of the same target, the target clustering effect is poor, and target splitting is likely to occur.

[0003] The millimeter-wave radar can measure the distance and radial velocity of each target in a multi-target scenario simultaneously by transmitting multiple cycles of Frequency Modulation Continuous Wave (FMCW) signals. The transmitted waveform is as shown in the appendix. Figure 1 The radio wave used by the millimeter-wave radar has a relatively long wavelength. Compared with the laser beam used by lidar, the resolution is lower. This means that at the same detection distance, the millimeter-wave radar can capture fewer target details, resulting in sparse point cloud data. When millimeter waves encounter a target object, scattering occurs. Due to the relatively long wavelength, the scattering angle is relatively large, resulting in a more dispersed distribution of the reflected signals in space. It is difficult to form a dense point cloud data, and the imaging ability of the target is not strong, and the target cannot be accurately identified.

[0004] When the millimeter-wave radar detects the same target, it can obtain multiple sparse point targets. When performing target tracking, multiple point targets will cause track splitting. Therefore, before performing target tracking, target clustering is required. The basic principle of the clustering algorithm is to cluster a continuous group of point clouds into one class based on the continuity of the point clouds. If the point clouds of the same target are not continuous, they usually cannot be well clustered. Due to the sparse and uneven distribution of millimeter-wave radar point clouds, when using the DBSCAN algorithm for clustering, the fixed neighborhood radius and the minimum number of points in the neighborhood usually do not match. When the neighborhood radius is set too large, it is easy to cluster the point clouds of different targets into one class. When the neighborhood radius is set too small, it is easy to cluster the point clouds of the same target into multiple classes, and the clustering accuracy of millimeter-wave radar point clouds is low.

[0005] Since the 4D millimeter-wave radar has a high angular resolution, which can reach about 1 degree, and the point cloud density is higher than that of traditional millimeter-wave radars, the 4D millimeter-wave radar has a higher lateral detection density for targets. At the same time, since the 4D millimeter-wave radar cannot solve the problems of strong scattering characteristics and front-back occlusion of millimeter-wave radars, the longitudinal imaging effect of the 4D millimeter-wave radar on targets is not significantly different from that of traditional millimeter-wave radars. Summary of the Invention

[0006] To solve the above technical problems, the present invention provides a 4D millimeter-wave radar clustering method, device, and electronic device.

[0007] In a first aspect, the present invention provides a 4D millimeter-wave radar clustering method, including:

[0008] Inputting 4D millimeter-wave radar point cloud data;

[0009] Setting a first neighborhood radius and the minimum number of points in the point cloud within the first neighborhood radius, and taking each first neighborhood radius region as a target region;

[0010] Using the DBSCAN algorithm for point cloud density clustering to obtain the point cloud density clustering results of each target region;

[0011] Calculating the center point positions of each point cloud density clustering and statistically calculating the width of each point cloud density clustering;

[0012] Setting a second neighborhood radius according to the width of each point cloud density clustering to obtain a set of second neighborhood radii;

[0013] Performing DBSCAN clustering on each point cloud density clustering according to the set second neighborhood radius to complete the point cloud clustering operation.

[0014] In a second aspect, the present invention provides a 4D millimeter-wave radar clustering device, including an input unit, a first setting unit, a first clustering unit, a statistical unit, a second setting unit, and a second clustering unit;

[0015] The input unit is used to input 4D millimeter-wave radar point cloud data;

[0016] The first setting unit is used to set a first neighborhood radius and the minimum number of points in the point cloud within the first neighborhood radius, and take each first neighborhood radius region as a target region;

[0017] The first clustering unit is used to use the DBSCAN algorithm for point cloud density clustering to obtain the point cloud density clustering results of each target region;

[0018] The statistical unit is used to calculate the center point positions of each point cloud density clustering and statistically calculate the width of each point cloud density clustering;

[0019] The second setting unit is used to set a second neighborhood radius according to the width of each point cloud density clustering to obtain a set of second neighborhood radii;

[0020] The second clustering unit is used to perform DBSCAN clustering on each point cloud density clustering according to the set second neighborhood radius to complete the point cloud clustering operation.

[0021] In a third aspect, the present invention provides an electronic device, including:

[0022] A processor and a memory;

[0023] The memory is used to store computer operation instructions;

[0024] The processor is used to execute the described 4D millimeter-wave radar clustering method by calling the computer operation instructions.

[0025] Based on the above technical solution, the present invention can also be improved as follows.

[0026] Further, the DBSCAN algorithm is used for point cloud density clustering to obtain the point cloud density clustering results of each target area, including:

[0027] Let the first neighborhood radius be ε and the minimum number of points within the first neighborhood radius be MinPts;

[0028] Find core points: Traverse each point in the dataset and calculate the number of points within the first neighborhood radius;

[0029] If the number of points within the first neighborhood radius of a point is greater than or equal to the minimum number of points within the first neighborhood radius, then mark this point as a core point;

[0030] Construct clusters: Start from an unvisited core point, add this core point to the current cluster and mark it as visited; find all points within the first neighborhood radius of this core point. If there are core points, recursively add the points within the first neighborhood radius to the current cluster and mark them as visited; if the points within the first neighborhood radius are non-core points but are within the first neighborhood radius of a core point, then mark these points within the first neighborhood radius as border points and add them to the same cluster as the core point; repeat finding all points within the first neighborhood radius of this core point until there are no new points that can be added within the first neighborhood radius of the current cluster;

[0031] Repeat to construct new clusters: Continue to traverse the unvisited points in the dataset, repeat finding core points until all data points are visited and classified into each cluster.

[0032] Further, mark the points in the dataset that are unvisited and do not belong to any cluster as noise points.

[0033] Further, perform DBSCAN clustering on each point cloud density clustering according to the set second neighborhood radius, including:

[0034] Let the th point cloud density clustering be , the nth cloud density clustering be , the point cloud density clustering set be , the th point cloud density clustering 's second neighborhood radius be , be in the th clustering, is the center point, is the center point coordinate of the th point cloud density clustering; is the center point coordinate of the th point cloud density clustering and the th point cloud density clustering distance difference between the center coordinates;

[0035] Starting from to construct the clustering, set the second neighborhood radius of the th point cloud density clustering , set the second neighborhood radius of the th point cloud density clustering with the minimum number of points inside being 1;

[0036] Search for adjacent center points, traverse the set of point cloud density clusterings , calculate , if , then is 's neighborhood, otherwise is not 's neighborhood;

[0037] Starting from each adjacent point of , repeat to search for adjacent center points until all adjacent points of are searched; use 's neighborhood radius when repeating to search for ;

[0038] For the point cloud density clusterings not included, repeat to search for adjacent center points until all point cloud density clustering sets are traversed.

[0039] Furthermore, the first neighborhood radius is less than the second neighborhood radius.

[0040] Furthermore, when setting the second neighborhood radius according to the width of each point cloud density clustering, set the second neighborhood radius according to the relationship between the length and width of the target.

[0041] The beneficial effects of the present invention are as follows: After the traditional clustering is completed, the width of each high-density horizontal clustering result is statistically analyzed, and a large-size clustering operation is performed on each high-density clustering result. The second neighborhood radius ε during the large-size clustering is determined according to the size of the horizontal clustering width, that is, the larger the current horizontal clustering width, the larger the set second neighborhood radius ε, until all point cloud clustering is completed. Therefore, the present invention can adapt to target point cloud data of different sizes, adaptively adjust the neighborhood radius of the large-size clustering according to the point cloud width, thereby improving the accuracy of point cloud clustering of the 4D millimeter-wave radar. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 FIG. is a schematic diagram of a 4D millimeter-wave radar clustering method provided in Embodiment 1 of the present invention;

[0043] Figure 2 FIG. is a schematic diagram of a single-target point cloud provided in Embodiment 1 of the present invention;

[0044] Figure 3 FIG. is a schematic diagram of the processing result of the horizontal clustering result provided in Embodiment 1 of the present invention;

[0045] Figure 4 FIG. is a schematic diagram of a 4D millimeter-wave radar clustering device provided in Embodiment 2 of the present invention;

[0046] Figure 5 FIG. is a schematic diagram of an electronic device provided in Embodiment 3 of the present invention.

[0047] Reference numerals: 30 - electronic device; 310 - processor; 320 - bus; 330 - memory; 340 - transceiver. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0048] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and illustrated herein can generally be arranged and designed in a variety of different configurations.

[0049] Embodiment 1

[0050] As an embodiment, as shown in the accompanying Figure 1 drawings, to solve the above technical problems, this embodiment provides a 4D millimeter-wave radar clustering method, including:

[0051] Input 4D millimeter-wave radar point cloud data;

[0052] Set the first neighborhood radius and the minimum number of points in the point cloud within the first neighborhood radius, and use each first neighborhood radius region as the target region;

[0053] Use the DBSCAN algorithm to perform point cloud density clustering to obtain the point cloud density clustering results of each target area;

[0054] Calculate the center point position of each point cloud density clustering and count the width of each point cloud density clustering;

[0055] Set the second neighborhood radius according to the width of each point cloud density clustering to obtain a set of second neighborhood radii;

[0056] Perform DBSCAN (Density-Based Spatial Clustering of Applications with Noise) clustering on each point cloud density clustering according to the set second neighborhood radius to complete the point cloud clustering operation.

[0057] When clustering the target, first, use the DBSCAN algorithm to count the horizontal width of the target point cloud, set the neighborhood radius ε to a smaller value (generally less than 0.2 meters), and set the minimum number of points MinPts to a larger value (generally greater than 50 points); after clustering, obtain the horizontal clustering results of the high-density areas of each target ; then, count the width of each high-density horizontal clustering result; start performing large-size clustering operations from each high-density clustering result ; the first neighborhood radius ε during large-size clustering is determined by the size of the horizontal clustering width, that is, the larger the current horizontal clustering width, the larger the set first neighborhood radius ε, until all point cloud clustering is completed.

[0058] Input a frame of 4D millimeter-wave radar point cloud data. The schematic diagram of a single-target point cloud is as attached Figure 2 shown. The radar point cloud has the characteristics of non-uniformity and sparsity.

[0059] Optionally, mark the points in the dataset that have not been visited and do not belong to any clustering as noise points.

[0060] The schematic diagram of the processing result of the horizontal clustering result is as attached Figure 3 shown. The x-axis is the horizontal distance, unit: m, and the y-axis is the longitudinal distance, unit: m.

[0061] The th point cloud density clustering is , and the width of the point cloud density clustering is , the second neighborhood radius is , and according to the width of each point cloud density clustering set the neighborhood radius , and obtain , when setting the neighborhood radius, based on the comparison of the common target width and length. For example, the length and width of a human target do not exceed 1 * 0.5m, the length and width of a motorcycle do not exceed 3 * 1m, the length and width of a car do not exceed 6 * 2m, and the length and width of a large vehicle do not exceed 10 * 3m. Use the length and width relationship of typical targets to set the second neighborhood radius for clustering, then:

[0062] .

[0063] Optionally, perform DBSCAN clustering on each point cloud density clustering according to the set second neighborhood radius, including:

[0064] Let the th point cloud density clustering be , the nth cloud density clustering be , the point cloud density clustering set be , the th point cloud density clustering 's second neighborhood radius be , be 's th clustering, be 's center point, be the th point cloud density clustering 's center point coordinates, be the th point cloud density clustering 's center point coordinates and the th point cloud density clustering 's distance difference between the center coordinates;

[0065] Start building clusters from , set the second neighborhood radius of the th point cloud density clustering , set the second neighborhood radius of the th point cloud density clustering with the minimum number of points inside being 1;

[0066] Find the 's neighboring center points, traverse the point cloud density clustering set , calculate , if , then be 's neighborhood, otherwise is not 's neighborhood;

[0067] Start from each neighboring point of , repeat finding to the adjacent center points until all are searched all adjacent points; repeat the search when looking for the adjacent center points of use the neighborhood radius of ;

[0068] Cluster the point cloud density of the unincluded points and repeat the search for the adjacent center points until all point cloud density clustering sets are traversed .

[0069] As shown in the appendix Figure 4 all the point clouds of the same target are clustered into one class

[0070] Optionally, the first neighborhood radius is less than the second neighborhood radius

[0071] Optionally, when setting the second neighborhood radius according to the width of each point cloud density clustering, set the second neighborhood radius according to the relationship between the length and width of the target

[0072] After the traditional clustering is completed in the present invention, the widths of the high-density horizontal clustering results are statistically calculated; the large-size clustering operation is performed starting from each high-density clustering result; the second neighborhood radius ε during the large-size clustering is determined according to the size of the horizontal clustering width, that is, the larger the current horizontal clustering width, the larger the set second neighborhood radius ε, until all point cloud clustering is completed. Therefore, the present invention can adapt to target point cloud data of different sizes, adaptively adjust the neighborhood radius of the large-size clustering according to the point cloud width, thereby improving the accuracy of point cloud clustering of the 4D millimeter-wave radar

[0073] Embodiment 2

[0074] Based on the same principle as the method shown in Embodiment 1 of the present invention, a 4D millimeter-wave radar clustering device is further provided in an embodiment of the present invention, including an input unit, a first setting unit, a first clustering unit, a statistical unit, a second setting unit and a second clustering unit;

[0075] The input unit is used to input 4D millimeter-wave radar point cloud data;

[0076] The first setting unit is used to set the first neighborhood radius and the minimum number of points in the point cloud within the first neighborhood radius, and regard each first neighborhood radius region as a target region;

[0077] The first clustering unit is used to perform point cloud density clustering using the DBSCAN algorithm to obtain the point cloud density clustering results of each target region;

[0078] The statistical unit is used to calculate the center point positions of each point cloud density clustering and statistically calculate the widths of each point cloud density clustering;

[0079] A second setting unit, configured to set a second neighborhood radius according to the width of each point cloud density cluster, so as to obtain a set of second neighborhood radii;

[0080] A second clustering unit, configured to perform DBSCAN clustering on each point cloud density cluster according to the set second neighborhood radius, so as to complete the point cloud clustering operation.

[0081] Optionally, the points in the dataset that are not accessed and do not belong to any cluster are marked as noise points.

[0082] Optionally, performing DBSCAN clustering on each point cloud density cluster according to the set second neighborhood radius includes:

[0083] Let the th point cloud density cluster be , the nth cloud density cluster be , the set of point cloud density clusters be , the th point cloud density cluster 's second neighborhood radius be , be 's th cluster, be 's center point, be the th point cloud density cluster 's center point coordinates, be the th point cloud density cluster 's center point coordinates and the th point cloud density cluster 's distance difference between the center coordinates;

[0084] Start building a cluster from , set the second neighborhood radius of the th point cloud density cluster, set the second neighborhood radius of the th point cloud density cluster, and the minimum number of points inside it is 1;

[0085] Find 's neighboring center points, traverse the set of point cloud density clusters , calculate , if , then is 's neighborhood, otherwise is not 's neighborhood;

[0086] Starting from each adjacent point of repeatedly search for the adjacent central point of until all adjacent points of are searched; when repeatedly searching for the adjacent central point of use the neighborhood radius of ; ;

[0087] Perform density clustering on the unincluded point clouds and repeatedly search for the adjacent central point until all point cloud density clustering sets are traversed ; .

[0088] Optionally, the first neighborhood radius is less than the second neighborhood radius.

[0089] Optionally, when setting the second neighborhood radius according to the width of each point cloud density clustering, set the second neighborhood radius according to the relationship between the length and width of the target.

[0090] Embodiment 3

[0091] Based on the same principle as the method shown in the embodiments of the present invention, an electronic device is also provided in the embodiments of the present invention. As shown in the appendix Figure 5 , the electronic device may include but is not limited to: a processor and a memory; the memory is used to store a computer program; the processor is used to execute a 4D millimeter-wave radar clustering method shown in the embodiments of the present invention by calling the computer program.

[0092] In an optional embodiment, an electronic device is provided. Figure 5 The electronic device 30 shown includes: a processor 310 and a memory 330. Among them, the processor 310 and the memory 330 are connected, such as connected through a bus 320.

[0093] Optionally, the electronic device 30 may further include a transceiver 340. The transceiver 340 may be used for data interaction between the electronic device and other electronic devices, such as data sending and / or data receiving, etc. It should be noted that in actual applications, the transceiver 340 is not limited to one, and the structure of the electronic device 30 does not constitute a limitation to the embodiments of the present invention.

[0094] The processor 310 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute various exemplary logical blocks, modules, and circuits described in connection with the disclosure of the present invention. The processor 310 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0095] The bus 320 may include a path for transmitting information between the above components. The bus 320 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The bus 320 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 5 only a thick line is shown in the figure, but it does not mean that there is only one bus or one type of bus.

[0096] The memory 330 may be a ROM (Read Only Memory) or other type of static storage device that can store static information and instructions, a RAM (Random Access Memory) or other type of dynamic storage device that can store information and instructions, or it may also be an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory), or other optical disc storage, optical disc storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), magnetic disk storage media, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto.

[0097] The memory 330 is used to store the application program code (computer program) for implementing the solution of the present invention, and is controlled by the processor 310 for execution. The processor 310 is used to execute the application program code stored in the memory 330 to implement the content shown in the foregoing method embodiments.

[0098] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various changes and modifications. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A 4D millimeter wave radar clustering method, characterized in that: include: Input 4D millimeter wave radar point cloud data; Set the first neighborhood radius and the minimum number of points in the point cloud within the first neighborhood radius, and use each first neighborhood radius area as the target area; Use the DBSCAN algorithm to perform point cloud density clustering and obtain the point cloud density clustering results of each target area; Calculate the center point position of each point cloud density cluster and count the width of each point cloud density cluster; The second neighborhood radius is set according to the width of each point cloud density cluster to obtain a set of second neighborhood radii; when the second neighborhood radius is set according to the width of each point cloud density cluster, the second neighborhood radius is set according to the relationship between the length and width of the target; Perform DBSCAN clustering on each point cloud density cluster according to the set second neighborhood radius to complete the point cloud clustering operation.

2. A 4D millimeter wave radar clustering method according to claim 1, characterized in that: Use the DBSCAN algorithm to perform point cloud density clustering and obtain the point cloud density clustering results of each target area, including: Let the radius of the first neighborhood be ε, and the minimum number of points within the radius of the first neighborhood be MinPts; Find core points: traverse each point in the data set and calculate the number of points within the first neighborhood radius; If the number of points within the first neighborhood radius of a point is greater than or equal to the minimum number of points within the first neighborhood radius, the point is marked as a core point; Construct clusters: Start from an unvisited core point, add the core point to the current cluster, and mark it as visited; find all points within the first neighborhood radius of the core point, if there is a core point, recursively add the points within the first neighborhood radius to the current cluster, and mark them as visited; if the point within the first neighborhood radius is a non-core point but is within the first neighborhood radius of a core point, mark the point within the first neighborhood radius as a boundary point and add it to the same cluster as the core point; repeatedly find all points within the first neighborhood radius of the core point until there are no new points that can be added within the first neighborhood radius of the current cluster; Repeatedly build new clusters: Continue to traverse the unvisited points in the data set and repeatedly search for core points until all data points have been visited and classified into various clusters.

3. A 4D millimeter wave radar clustering method according to claim 2, characterized in that: The points in the dataset that have not been visited and do not belong to any cluster are marked as noise points.

4. The 4D millimeter wave radar clustering method according to claim 1, characterized in that: Perform DBSCAN clustering on each point cloud density cluster according to the set second neighborhood radius, including: Set up The point cloud density is clustered as , the nth point cloud density cluster is , the point cloud density clustering set is , No. Point cloud density clustering The second neighborhood radius of , for The clusters, for The center point of For the Point cloud density clustering The center point coordinates, For the Point cloud density clustering The center point coordinates and Point cloud density clustering The distance difference between the center coordinates of from Start building clusters and set Point cloud density clustering The second neighborhood radius , set the Point cloud density clustering The second neighborhood radius The minimum number of points is 1; Search Nearby center points, traverse the point cloud density clustering set ,calculate ,if ,but for Neighborhood of no Neighborhood of; from Starting from each neighboring point of Until the nearest center point is found All neighboring points of Use when the center point is close to Neighborhood radius ; Clustering of point cloud density not included , repeat the search Nearby center points until all point cloud density clustering sets are traversed .

5. The 4D millimeter wave radar clustering method according to claim 1, characterized in that: The first neighborhood radius is smaller than the second neighborhood radius.

6. A 4D millimeter wave radar clustering device, characterized in that: It includes an input unit, a first setting unit, a first clustering unit, a statistical unit, a second setting unit and a second clustering unit; Input unit, used to input 4D millimeter wave radar point cloud data; A first setting unit is used to set a first neighborhood radius and a minimum number of points in a point cloud within the first neighborhood radius, and to use each first neighborhood radius area as a target area; The first clustering unit is used to perform point cloud density clustering using the DBSCAN algorithm to obtain point cloud density clustering results for each target area; The statistical unit is used to calculate the center point position of each point cloud density cluster and count the width of each point cloud density cluster; The second setting unit is used to set the second neighborhood radius according to the width of each point cloud density cluster to obtain a set of second neighborhood radii; when setting the second neighborhood radius according to the width of each point cloud density cluster, the second neighborhood radius is set according to the relationship between the length and width of the target; The second clustering unit is used to perform DBSCAN clustering on each point cloud density cluster according to the set second neighborhood radius to complete the point cloud clustering operation.

7. An electronic device, characterized in that: include: Processor and memory; The memory is used to store computer operation instructions; The processor is used to execute a 4D millimeter-wave radar clustering method according to any one of claims 1 to 5 by calling the computer operation instruction.

Images