An adaptive vehicle-mounted millimeter-wave radar DBSCAN clustering method and device

Abstract

The present invention discloses an adaptive DBSCAN clustering method and device for vehicle-mounted millimeter-wave radars, comprising: obtaining vehicle point cloud data and processing the point cloud data to obtain the vehicle's length factor, width factor, phantom radius, and speed deviation; using an elliptical search area as the shape, and calculating the major and minor axes of the elliptical search area based on the vehicle length factor, width factor, phantom radius, and speed deviation; constructing a dataset of L-dimensional vectors containing all point cloud data based on the major and minor axes of the elliptical search area, where L is a positive integer; calculating a domain radius and a minimum number of samples based on the dataset; and performing DBSCAN clustering on the dataset based on the domain radius and the minimum number of samples to traverse all data points in the dataset. The present invention can obtain accurate vehicle point cloud data and can adaptively adjust the domain radius and minimum number of samples, thereby improving the accuracy of system clustering.

Description

Technical Field

[0001] The present invention relates to the field of automobile intelligent driving technology, and in particular to an adaptive vehicle-mounted millimeter-wave radar DBSCAN clustering method and device. Background Art

[0002] In recent years, with the increasing complexity of traffic environments, automotive millimeter-wave radar has played an increasingly important role in identifying the motion of obstacles on the road. Compared to camera sensors, millimeter-wave radar offers superior robustness in harsh environments, and as a result, has attracted widespread attention from researchers. Millimeter-wave radar can be used to acquire and process point cloud data of obstacles in road scenes. Clustering algorithms can be used to remove noise points from the data. The accuracy of point cloud data clustering directly impacts the quality of target tracking. Clustering algorithms ensure that similar point cloud data are clustered together while maintaining the ability to distinguish between clusters.

[0003] Currently common clustering algorithms include: the K-means algorithm requires experience to pre-select the K value, and it is difficult to determine the appropriate K value. It is difficult to converge for non-convex data and is sensitive to noise points. The hierarchical AGNES clustering algorithm has high computational complexity and may cluster into chains. The density-based spatial clustering of applications with noise (DBSCAN) algorithm can cluster data clusters of any shape and is insensitive to noise points, which can be identified during clustering.

[0004] However, the traditional DBSCAN algorithm can lead to inaccurate obstacle recognition in situations with uneven density and large differences in cluster spacing. The traditional DBSCAN algorithm manually determines parameters such as the radius of the domain circle and the minimum number of samples. In practice, millimeter-wave radars must adjust the radius and minimum number of samples based on factors such as distance, speed, radar cross-section (RCS), and emission intensity. Because the recognized data points don't take into account the length and width of the vehicle, using the traditional DBSCAN algorithm can cause two parallel obstacles to merge into a single cluster, necessitating parameter refinement. Summary of the Invention

[0005] The purpose of the present invention is to address the deficiencies of the above-mentioned prior art and propose an adaptive vehicle-mounted millimeter-wave radar DBSCAN clustering method and device, aiming to solve the problem that the prior art cannot accurately identify vehicles.

[0006] In a first aspect, the present invention provides an adaptive vehicle-mounted millimeter-wave radar DBSCAN clustering method, the method comprising:

[0007] Obtaining point cloud data of the vehicle, and performing data processing on the point cloud data to obtain a vehicle length factor, a vehicle width factor, a phantom radius, and a speed deviation of the vehicle;

[0008] Taking an ellipse as the shape of the search area, and calculating the major axis and minor axis of the ellipse search area according to the vehicle length factor, vehicle width factor, imaginary radius, and speed deviation;

[0009] Constructing a data set of L-dimensional vectors containing all point cloud data according to the major axis and minor axis of the elliptical search area; wherein L is a positive integer;

[0010] Based on the data set, calculate the domain radius and the minimum number of samples;

[0011] According to the domain radius and the minimum number of samples, DBSCAN clustering is performed on the data set to traverse all data points in the data set.

[0012] The present invention processes point cloud data to obtain the vehicle's length factor, vehicle width factor, phantom radius, and speed deviation. The vehicle's length factor, vehicle width factor, phantom radius, and speed deviation are comprehensively considered to calculate an elliptical search area, thereby obtaining accurate vehicle point cloud data without causing data confusion. The obtained field radius and minimum number of samples can be adaptively adjusted according to the distribution density of the data set without the need for manual adjustment, thereby improving the accuracy of system clustering.

[0013] Furthermore, the point cloud data of the vehicle is obtained and the point cloud data is processed to obtain the vehicle length factor, vehicle width factor, imaginary radius, and speed deviation of the vehicle. The calculation of the vehicle length factor and vehicle width factor includes:

[0014] The point cloud data is processed to obtain the number of data points of the point cloud cluster in terms of length and width, and the vehicle length factor and vehicle width factor of the vehicle are calculated based on the length and width of the rectangular point cloud data.

[0015] Furthermore, the speed deviation is calculated based on the speed information of the point cloud data; wherein the speed deviation is proportional to the speed information contained in the point cloud data.

[0016] Furthermore, the calculating of the major axis and minor axis of the elliptical search area according to the vehicle length factor, vehicle width factor, imaginary radius, and speed deviation includes:

[0017] The major axis is linearly represented by the vehicle length factor, the imaginary radius, and the speed deviation, and the minor axis is linearly represented by the vehicle width factor, the imaginary radius, and the speed deviation.

[0018] Furthermore, the data set of L-dimensional vectors containing all point cloud data is constructed, and the data set is constructed as follows: each data point in the data set is represented by a two-dimensional vector of the major axis and the minor axis of the elliptical search area of ​​the data point.

[0019] The present invention extracts the major axis and minor axis of an elliptical search area as a two-dimensional vector data set. The major axis and minor axis of the elliptical search area simultaneously include the vehicle length factor, vehicle width factor, phantom radius, and speed deviation influencing factors. The major axis and minor axis of the elliptical search area can adapt to the density of point cloud data. The elliptical search area can accurately characterize the data characteristics of the vehicle, thereby improving the accuracy of system clustering.

[0020] Furthermore, the domain radius and the minimum number of samples are calculated based on the data set, including: the minimum number of samples is in direct proportion to the value of the L-dimensional vector of the data set.

[0021] Furthermore, performing DBSCAN clustering on the data set according to the domain radius and the minimum number of samples to traverse all data points in the data set includes:

[0022] Mark all data points in the data set as unread, and randomly select an unread data point as the first data point;

[0023] determining, based on a domain radius of the first data point, whether the number of data points within the domain radius is not less than the minimum sample size; if the number of data points is not less than the minimum sample size, determining that the first data point is a first core point, and marking the first core point as read;

[0024] Sequentially determine whether the remaining unread data points within the domain of the first core point are second core points. If an unread data point within the domain of the first core point is the second core point, traverse the unread data points within the domain of the second core point to determine whether a third core point exists, and mark the second core point as read.

[0025] All unread core points and unread data points in their areas are traversed according to the depth-first traversal algorithm to complete the clustering of all data points.

[0026] Furthermore, the determining whether the number of data points is not less than the minimum sample size, if the number of data points is not less than the minimum sample size, further includes:

[0027] If the number of data points is less than the minimum sample size, the first data point is a noise point, and it is determined whether there are any unread data points. If there are any unread data points, an unread data point is randomly searched for as the first data point for traversal; otherwise, the process ends.

[0028] Furthermore, the step of sequentially determining whether the remaining unread data points within the area of ​​the first core point are second core points, and if an unread data point within the area of ​​the first core point is the second core point, further includes:

[0029] If an unread data point in the area of ​​the first core point is not the second core point, then this point is a critical point, and the judgment of whether other data points in the area of ​​this point are core points is stopped. This point is marked as read, and the currently read data points are counted as the first cluster.

[0030] In a second aspect, the present invention further provides an adaptive vehicle-mounted millimeter-wave radar DBSCAN clustering device, comprising:

[0031] A data acquisition module is used to obtain point cloud data of the vehicle and perform data processing on the point cloud data to obtain the vehicle length factor, vehicle width factor, imaginary radius, and speed deviation;

[0032] A search area setting module is used to use an ellipse as the shape of the search area and calculate the major axis and minor axis of the ellipse search area according to the vehicle length factor, vehicle width factor, imaginary radius, and speed deviation;

[0033] A data set construction module is used to construct an L-dimensional vector data set containing all point cloud data according to the major axis and minor axis of the elliptical search area; wherein L is a positive integer;

[0034] A parameter calculation module, used to calculate the domain radius and the minimum number of samples based on the data set;

[0035] The clustering module is used to perform DBSCAN clustering on the data set according to the domain radius and the minimum number of samples, so as to traverse all data points in the data set. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 1 is a flow chart of DBSCAN clustering of an adaptive vehicle-mounted millimeter-wave radar provided by an embodiment of the present invention;

[0037] Figure 2 Schematic diagram of the structure of the adaptive vehicle-mounted millimeter-wave radar DBSCAN clustering provided by an embodiment of the present invention;

[0038] Figure 31 is a schematic diagram of the results of DBSCAN clustering of the adaptive vehicle-mounted millimeter-wave radar provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0039] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0040] See also Figure 1 , is a schematic diagram of the process of adaptive vehicle-mounted millimeter-wave radar DBSCAN clustering provided by an embodiment of the present invention, including steps S11 to S15, specifically:

[0041] Step S11: Obtain point cloud data of the vehicle, and perform data processing on the point cloud data to obtain the vehicle length factor, vehicle width factor, imaginary radius, and speed deviation of the vehicle.

[0042] Specifically, the calculation of the vehicle length factor and vehicle width factor includes: performing data processing on the point cloud data to obtain the number of data points in the length and width of the point cloud cluster, and calculating the vehicle length factor and vehicle width factor of the vehicle based on the length and width of the rectangular point cloud data.

[0043] It is worth noting that since the raw point cloud data obtained from the vehicle-mounted millimeter-wave radar sensor contains not only vehicles, but also data on bicycles, pedestrians, etc., it is necessary to perform vehicle target recognition on the raw point cloud data based on the traditional DBSCAN algorithm. Since the point cloud data of the vehicle is rectangular, the point cloud cluster obtained after vehicle recognition is also rectangular. Therefore, the number of data points under the length of the point cloud cluster and the number of data points under the width of the point cloud cluster can be obtained. Based on the length and width of the point cloud cluster, the vehicle length factor and vehicle width factor can be calculated.

[0044] Preferably, the formulas for calculating the vehicle length factor and vehicle width factor are:

[0045]

[0046] Among them, ε x and ε y are the vehicle body length and width factor variables, w and l are the width and length of the dataset, M and N are the number of long data points and wide data points of the point cloud cluster, respectively.

[0047] In step S11, the pseudo radius is calculated by calculating the circular area from the reflection cross-sectional radius of the vehicle-mounted millimeter-wave radar. It is worth noting that, assuming that the pseudo area of ​​the data point is a circle, the pseudo radius can be calculated from the circular area. Preferably, the formula for calculating the pseudo radius is:

[0048]

[0049] Among them, ω RCS is the imaginary area of ​​the vehicle's circle.

[0050] In step S11, the speed deviation is calculated based on the speed information of the point cloud data; wherein the speed deviation is proportional to the speed information contained in the point cloud data.

[0051] Preferably, the calculation formula of the speed deviation is:

[0052]

[0053] Among them, K is the scale coefficient, and V is the velocity information contained in the point cloud data.

[0054] It is worth noting that ε v It is directly related to different data points. The speed information in the point cloud data is proportional to the speed deviation. The speed deviation can be used to avoid misidentifying two adjacent data points with different speed information as the same cluster.

[0055] Step S12: Using an ellipse as the shape of the search area, and calculating the major axis and minor axis of the ellipse search area according to the vehicle length factor, vehicle width factor, imaginary radius, and speed deviation.

[0056] It is worth noting that since the search area is represented by an ellipse, the major axis and minor axis of the elliptical search area need to be calculated; wherein the major axis is linearly represented by the vehicle length factor, the imaginary radius and the speed deviation, and the minor axis is linearly represented by the vehicle width factor, the imaginary radius and the speed deviation.

[0057] Preferably, the calculation formulas for the major axis and minor axis of the elliptical search area are respectively:

[0058]

[0059] Among them, ε x and ε y are vehicle length factor, vehicle width factor, ε v is the velocity deviation of different data points, ε RCS is the target illusion radius.

[0060] Step S13: constructing a data set of L-dimensional vectors containing all point cloud data according to the major axis and minor axis of the elliptical search area; wherein L is a positive integer.

[0061] Preferably, each data point in the data set is represented by a two-dimensional vector of the major axis and the minor axis of the elliptical search area of ​​the data point. The data set can be represented as:

[0062] D=(d1,d2,…,dn), (5)

[0063] Among them, d i The two-dimensional vectors of the major and minor axes of the elliptical search area for the data point can be expressed as:

[0064] d i =[ε a,m ,ε b,m ],1≤m≤n, (6)

[0065] Among them, m and n are the number of data points in the width and length of the data set, respectively, and m <n,ε a,m and ε b,m are the major and minor axes of the elliptical search area of ​​the i-th data point respectively.

[0066] It is worth noting that the vectors of the data set are not limited to the major axis values ​​and minor axis values ​​of the ellipse, and may also include other data feature values ​​of the data points, which are not limited here.

[0067] Step S14: Calculate the domain radius and the minimum number of samples based on the data set.

[0068] Specifically, the domain radius is related to the vectors of the data set, including: calculating the pairwise distances between a data point and the remaining data points based on the vectors in the data set, selecting the first k smallest distances as the k-shortest distances between the data point and its k nearest neighboring data points, and selecting the inflection point of the distance value from the k-shortest distances as the optimal domain radius, that is, the optimal domain radius is the distance with the largest numerical difference between the front and rear distances in the sorted k-shortest distance series; wherein k is a positive integer, and k is in direct proportion to the value of the L-dimensional vector.

[0069] Exemplarily, the k-nearest distances are sorted from small to large, and the value of the i-th data is subtracted from the i+1-th data in the sorted k-nearest distances to obtain a distance difference series of pairwise distances. The distance corresponding to the maximum value in the distance difference series is selected as the optimal field radius, where i is initially 1 and not greater than k; or linear fitting is used to determine the optimal field radius, which is not limited here.

[0070] Preferably, the distance between two data points is calculated using the Mahalanobis distance formula, which can be expressed as:

[0071]

[0072] Among them, d i and d j are all data points in the data set; S is the covariance matrix of the two data points, which is composed of the covariance of the two random variables in the set of random variables. When the two vectors are uncorrelated, the covariance matrix is ​​a diagonal matrix, and the covariance matrix is ​​expressed as:

[0073]

[0074] in, is the variance of the vector of the ith data point of the population.

[0075] The present invention uses Mahalanobis distance to calculate the similarity measure of two data points, which is not affected by rigidity. The Mahalanobis distance between two data points is independent of the measurement unit of the original data points and can eliminate the interference of the correlation between data points.

[0076] Preferably, the calculation formula for the value of k is:

[0077] k=2*L-1, (9)

[0078] In step S14, the minimum number of samples is in direct proportion to the value of the L-dimensional vector.

[0079] Preferably, the minimum sample size is calculated as follows:

[0080] MinPts=2*L-1, (10)

[0081] Wherein, L is the vector dimension of the data set; it is worth noting that the value of MinPts may be the same as or different from the value of k, which is not limited here.

[0082] Step S15: Perform DBSCAN clustering on the data set based on the domain radius and the minimum number of samples to traverse all data points in the data set, specifically:

[0083] Mark all data points in the data set as unread, and randomly select an unread data point as the first data point;

[0084] determining, based on a domain radius of the first data point, whether the number of data points within the domain radius is not less than the minimum sample size; if the number of data points is not less than the minimum sample size, determining that the first data point is a first core point, and marking the first core point as read;

[0085] Sequentially determine whether the remaining unread data points within the domain of the first core point are second core points. If an unread data point within the domain of the first core point is the second core point, traverse the unread data points within the domain of the second core point to determine whether a third core point exists, and mark the second core point as read.

[0086] All unread core points and unread data points in their areas are traversed according to the depth-first traversal algorithm to complete the clustering of all data points.

[0087] Specifically, the determining whether the number of data points is not less than the minimum sample number, if the number of data points is not less than the minimum sample number, further includes: if the number of data points is less than the minimum sample number, the first data point is a noise point, and determining whether there are any unread data points, if there are any unread data points, then randomly searching for unread data points as the first data point for traversal; otherwise, ending.

[0088] The method sequentially determines whether the remaining unread data points in the field of the first core point are the second core point. If an unread data point in the field of the first core point is the second core point, it also includes: if an unread data point in the field of the first core point is not the second core point, then this point is a critical point, and the determination of whether other data points in the field of this point are core points is stopped, the point is marked as read, and the currently read data points are counted as the first cluster.

[0089] The present invention processes point cloud data to obtain the vehicle's length factor, vehicle width factor, phantom radius, and speed deviation. The vehicle's length factor, vehicle width factor, phantom radius, and speed deviation are comprehensively considered to calculate an elliptical search area. Unread data points are searched and clustered through the elliptical search area, which can effectively identify the vehicle's point cloud data and is less likely to cause data confusion. The obtained field radius and minimum number of samples can be adaptively adjusted according to the density of the vehicle without manual adjustment, thereby improving the accuracy of system clustering.

[0090] The present invention provides a schematic diagram of the structure of the adaptive vehicle-mounted millimeter wave radar DBSCAN clustering, see Figure 2 , is a schematic diagram of the structure of the adaptive vehicle-mounted millimeter wave radar DBSCAN clustering provided by an embodiment of the present invention, including: a data acquisition module 21, a search area setting module 22, a data set construction module 23, a parameter calculation module 24, and a clustering module 25, wherein,

[0091] The data acquisition module 21 is used to obtain the point cloud data of the vehicle and perform data processing on the point cloud data to obtain the vehicle length factor, vehicle width factor, imaginary radius, and speed deviation;

[0092] A search area setting module 22 is configured to use an ellipse as the shape of the search area and calculate the major axis and minor axis of the ellipse search area according to the vehicle length factor, vehicle width factor, imaginary radius, and speed deviation;

[0093] The data set construction module 23 is used to construct an L-dimensional vector data set containing all point cloud data according to the major axis and minor axis of the elliptical search area; wherein L is a positive integer;

[0094] A parameter calculation module 24 is used to calculate the domain radius and the minimum number of samples based on the data set;

[0095] The clustering module 25 is configured to perform DBSCAN clustering on the data set according to the domain radius and the minimum number of samples, so as to traverse all data points in the data set.

[0096] Specifically, the data acquisition module 21 also includes:

[0097] a vehicle body factor calculation module, configured to process the point cloud data to obtain the number of data points in the length and width of the point cloud cluster, and calculate the vehicle length factor and vehicle width factor of the vehicle based on the length and width of the rectangular point cloud data;

[0098] The speed deviation calculation module is used to calculate the speed deviation according to the speed information of the point cloud data; wherein the speed deviation is proportional to the speed information contained in the point cloud data.

[0099] The search area setting module 22 further includes:

[0100] A long axis calculation module is used for performing linear representation according to the vehicle length factor, the imaginary radius and the speed deviation.

[0101] The short axis calculation module is used to perform linear representation according to the vehicle width factor, the imaginary radius and the speed deviation.

[0102] In the dataset construction module 23, a dataset of L-dimensional vectors containing all point cloud data is constructed. The dataset is constructed as follows: each data point in the dataset is represented by a two-dimensional vector of the major axis and the minor axis of the elliptical search area of ​​the data point.

[0103] The parameter calculation module 24 also includes:

[0104] The community radius calculation module is used to calculate the community radius through Mahalanobis distance.

[0105] The minimum sample quantity calculation module is used to calculate that the minimum sample quantity is in direct proportion to the value of the L-dimensional vector of the data set.

[0106] The clustering module 25 also includes a traversal module for marking all data points in the data set as unread, and randomly selecting an unread data point as the first data point; judging whether the number of data points within the domain radius of the first data point is not less than the minimum sample number, if the number of data points is not less than the minimum sample number, the first data point is the first core point, and the first core point is marked as read; judging whether the remaining unread data points within the domain of the first core point are second core points in turn, if an unread data point within the domain of the first core point is the second core point, traversing the unread data points within the domain of the second core point to determine whether there is a third core point, and marking the second core point as read; traversing all unread core points and the unread data points within their domains according to the depth-first traversal algorithm to complete the clustering of all data points.

[0107] Among them, the judging whether the number of data points is not less than the minimum sample number, if the number of data points is not less than the minimum sample number, further includes: if the number of data points is less than the minimum sample number, the first data point is a noise point, and judging whether there are any unread data points, if there are any unread data points, then randomly searching for unread data points as the first data point for traversal; otherwise, ending.

[0108] The method sequentially determines whether the remaining unread data points in the field of the first core point are the second core point. If an unread data point in the field of the first core point is the second core point, it also includes: if an unread data point in the field of the first core point is not the second core point, then this point is a critical point, and the determination of whether other data points in the field of this point are core points is stopped, the point is marked as read, and the currently read data points are counted as the first cluster.

[0109] The present invention processes point cloud data to obtain the vehicle's length factor, vehicle width factor, phantom radius, and speed deviation. The vehicle's length factor, vehicle width factor, phantom radius, and speed deviation are comprehensively considered to calculate an elliptical search area, thereby obtaining accurate vehicle point cloud data without causing data confusion. The obtained field radius and minimum number of samples can be adaptively adjusted according to the distribution density of the data set without the need for manual adjustment, thereby improving the accuracy of system clustering.

[0110] The present invention also provides a schematic diagram of clustering results, see Figure 3 , is a schematic diagram of the results of the adaptive vehicle-mounted millimeter-wave radar DBSCAN clustering provided by an embodiment of the present invention. In the figure, data points of the same cluster are divided into the same shaded area. It can be seen from the figure that this algorithm can accurately cluster data points with few noise points.

[0111] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0112] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the steps in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0113] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0114] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0115] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. An adaptive vehicle-mounted millimeter-wave radar DBSCAN clustering method, characterized in that: The method comprises: Obtaining point cloud data of the vehicle, and performing data processing on the point cloud data to obtain a vehicle length factor, a vehicle width factor, a phantom radius, and a speed deviation of the vehicle; Taking an ellipse as the shape of the search area, and calculating the major axis and minor axis of the ellipse search area according to the vehicle length factor, vehicle width factor, imaginary radius, and speed deviation; Constructing a data set of L-dimensional vectors containing all point cloud data according to the major axis and minor axis of the elliptical search area; wherein L is a positive integer; Based on the data set, calculate the domain radius and the minimum number of samples; Performing DBSCAN clustering on the data set according to the domain radius and the minimum number of samples to traverse all data points in the data set; The point cloud data of the vehicle is obtained and data processing is performed on the point cloud data to obtain the vehicle length factor, vehicle width factor, imaginary radius, and speed deviation of the vehicle. The calculation of the vehicle length factor and vehicle width factor includes: The point cloud data is processed to obtain the number of data points of the point cloud cluster in terms of length and width, and the vehicle length factor and vehicle width factor of the vehicle are calculated based on the length and width of the rectangular point cloud data; the vehicle length factor and vehicle width factor are specifically: Among them, ε x and ε y are the vehicle body length and width factor variables, w and l are the width and length of the dataset, M and N are the number of long data points and wide data points of the point cloud cluster, respectively; The pseudo radius is calculated by using the circular area from the reflection cross-sectional radius of the vehicle-mounted millimeter-wave radar. Specifically, the pseudo radius is: Among them, ω RCS is the imaginary area of ​​the vehicle's circle; The speed deviation is calculated based on the speed information of the point cloud data; wherein the speed deviation is proportional to the speed information contained in the point cloud data; the speed deviation is specifically: e v =KV; Among them, K is the scale coefficient, and V is the velocity information contained in the point cloud data; The major axis is linearly represented by the vehicle length factor, the imaginary radius, and the speed deviation, and the minor axis is linearly represented by the vehicle width factor, the imaginary radius, and the speed deviation; the major axis and the minor axis are specifically: e a =e x +e v +e RCS ,he b =e y +e v +e RCS ; Among them, ε x and ε y are vehicle length factor, vehicle width factor, ε v is the velocity deviation of different data points, ε RCS is the target illusion radius.

2. The adaptive vehicle-mounted millimeter-wave radar DBSCAN clustering method according to claim 1, characterized in that: The method constructs a data set of L-dimensional vectors containing all point cloud data. The data set is constructed as follows: each data point in the data set is represented by a two-dimensional vector of the major axis and the minor axis of the elliptical search area of ​​the data point.

3. The adaptive vehicle-mounted millimeter-wave radar DBSCAN clustering method according to claim 1, characterized in that: The calculating of the domain radius and the minimum number of samples based on the data set includes: the minimum number of samples is in direct proportion to the value of the L-dimensional vector of the data set.

4. The adaptive vehicle-mounted millimeter-wave radar DBSCAN clustering method according to claim 1, characterized in that: The DBSCAN clustering is performed on the data set according to the domain radius and the minimum number of samples to traverse all data points in the data set, including: Mark all data points in the data set as unread, and randomly select an unread data point as the first data point; determining, based on a domain radius of the first data point, whether the number of data points within the domain radius is not less than the minimum sample size; if the number of data points is not less than the minimum sample size, determining that the first data point is a first core point, and marking the first core point as read; Sequentially determine whether the remaining unread data points within the domain of the first core point are second core points. If an unread data point within the domain of the first core point is the second core point, traverse the unread data points within the domain of the second core point to determine whether a third core point exists, and mark the second core point as read. All unread core points and unread data points in their areas are traversed according to the depth-first traversal algorithm to complete the clustering of all data points.

5. The adaptive vehicle-mounted millimeter-wave radar DBSCAN clustering method according to claim 4, characterized in that: The determining whether the number of data points is not less than the minimum sample size, if the number of data points is not less than the minimum sample size, further includes: If the number of data points is less than the minimum sample size, the first data point is a noise point, and it is determined whether there are any unread data points. If there are any unread data points, an unread data point is randomly searched for as the first data point for traversal; otherwise, the process ends.

6. The adaptive vehicle-mounted millimeter-wave radar DBSCAN clustering method according to claim 4, characterized in that: The step of sequentially determining whether the remaining unread data points within the area of ​​the first core point are second core points, and if an unread data point within the area of ​​the first core point is the second core point, further comprises: If an unread data point in the area of ​​the first core point is not the second core point, then this point is a critical point, and the judgment of whether other data points in the area of ​​this point are core points is stopped. This point is marked as read, and the currently read data points are counted as the first cluster.

7. An adaptive vehicle-mounted millimeter-wave radar DBSCAN clustering device, characterized in that: include: A data acquisition module is used to obtain point cloud data of the vehicle and perform data processing on the point cloud data to obtain the vehicle length factor, vehicle width factor, imaginary radius, and speed deviation; A search area setting module is used to use an ellipse as the shape of the search area and calculate the major axis and minor axis of the ellipse search area according to the vehicle length factor, vehicle width factor, imaginary radius, and speed deviation; A data set construction module is used to construct an L-dimensional vector data set containing all point cloud data according to the major axis and minor axis of the elliptical search area; wherein L is a positive integer; A parameter calculation module, used to calculate the domain radius and the minimum number of samples based on the data set; A clustering module, configured to perform DBSCAN clustering on the data set according to the domain radius and the minimum number of samples, so as to traverse all data points in the data set; The point cloud data of the vehicle is obtained and data processing is performed on the point cloud data to obtain the vehicle length factor, vehicle width factor, imaginary radius, and speed deviation of the vehicle. The calculation of the vehicle length factor and vehicle width factor includes: The point cloud data is processed to obtain the number of data points of the point cloud cluster in terms of length and width, and the vehicle length factor and vehicle width factor of the vehicle are calculated based on the length and width of the rectangular point cloud data; the vehicle length factor and vehicle width factor are specifically: Among them, ε x and ε y are the vehicle body length and width factor variables, w and l are the width and length of the dataset, M and N are the number of long data points and wide data points of the point cloud cluster, respectively; The pseudo radius is calculated by using the circular area from the reflection cross-sectional radius of the vehicle-mounted millimeter-wave radar. Specifically, the pseudo radius is: Among them, ω RCS is the imaginary area of ​​the vehicle's circle; The speed deviation is calculated based on the speed information of the point cloud data; wherein the speed deviation is proportional to the speed information contained in the point cloud data; the speed deviation is specifically: e v =KV; Among them, K is the scale coefficient, and V is the velocity information contained in the point cloud data; The major axis is linearly represented by the vehicle length factor, the imaginary radius, and the speed deviation, and the minor axis is linearly represented by the vehicle width factor, the imaginary radius, and the speed deviation; the major axis and the minor axis are specifically: e a =e x +e v +e RCS ,he b =e y +e v +e RCS ; Among them, ε x and ε y are vehicle length factor, vehicle width factor, ε v is the velocity deviation of different data points, ε RCS is the target illusion radius.

Images