A radar target clustering method based on target features and kmeans

Abstract

The present invention discloses a radar target clustering method based on target features and kmeans, which obtains all target point information detected by a radar frame, including but not limited to the distance R i , angle A i ; Preset the grid size R_TH in direction R and the grid size X_TH in direction X; X i =R i sI i ; Find R i The maximum value R max , minimum value R min , X i The maximum value X max , minimum value X min ; #imgabs0# Calculate the number of effective grids in two directions respectively; select one of them as the number of center points of kmeans algorithm; select a target point from each grid, initialize multiple center points as cluster centers, and perform kmeans calculation. The present invention divides radar target points into grids according to target characteristics, which is more specific than randomly selecting cluster centers, can reduce calculation time and reduce errors.

Description

Technical Field

[0001] The invention belongs to the technical field of radar signal processing, and in particular relates to a radar target clustering method based on target features and kmeans. Background Art

[0002] Millimeter-wave radar and other radars play a very important role in the process of detecting targets. When the radar detects a target, when the radar scans an object with a large reflective surface, it often reflects back multiple points, which constitute a point group. At the same time, when encountering a larger background target that is not a target, it will also reflect back multiple points. In the subsequent radar data processing, it is necessary to adopt a reasonable clustering strategy to cluster these points into a certain target to reduce the target calculation amount and form the trajectory information of the determined target object. Among the current clustering algorithms, kmeans is the most commonly used clustering algorithm based on Euclidean distance. Its steps are to pre-divide the data into K groups, randomly select K objects as the initial cluster centers, and then calculate the distance between each object and each seed cluster center, and assign each object to the cluster center closest to it. The cluster center and the objects assigned to them represent a cluster.

[0003] However, in the common kmeans classification method, the number of cluster centers K and the initialization value of K are set manually and randomly selected. Different K values ​​will produce different results, which will also lead to excessively long calculation time and increase the probability that the K value is a noise point. Summary of the invention

[0004] In view of the above problems existing in the prior art, the purpose of the present invention is to provide a radar target clustering method based on target features and kmeans, which performs preliminary grid division on radar target points according to target characteristics, is more specific than randomly selecting the number of cluster centers, can reduce the calculation time of kmeans, can reduce the error caused by the local optimum, and eliminates the probability of selecting an outlier as the center compared with the conventional method of selecting the center. When initializing the center point, it is selected from the divided grid. The method preliminarily determines the center point information. Compared with randomly initializing the center point information, the calculation time of kmeans of the method is shorter.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is:

[0006] A radar target clustering method based on target features and kmeans comprises the following steps:

[0007] Step S1: Get all target points T detected by the radar in one frame i The information includes but is not limited to the target point T i The distance R i , angle Ai , speed V i , where i∈(1, 2, ...m);

[0008] Step S2: preset the size R_TH of the grid in direction R and the size X_TH of the grid in direction X according to the actual test scenario and test target; i =R i sina i ;

[0009] Step S3: Find all target points T i The distance R in the direction R i The maximum value R max and the minimum value R min ;

[0010] Step S4: Calculate the number of grids R in the R direction num ,

[0011] Step S5: According to R num and the effective number of grids K in the R_TH calculation direction R Rnum ;

[0012] Step S6: Find all target points T i Horizontal information X in direction X i The maximum value X max and minimum value X min ;

[0013] Step S7: Calculate the number of grids X in direction X num ,but

[0014] Step S8: According to X num and X_TH calculates the number of effective grids K in the direction X Xnum ;

[0015] Step S9: Select K Rnum or K Xnum K is the number of center points of the kmeans algorithm num ;

[0016] Step S10: Select a target point T from each grid at random i , initialize K num Center points, K num The center points are used as clustering centers and the kmeans algorithm is performed.

[0017] As a further improvement of the above technical solution:

[0018] X i The target point T iHorizontal information, horizontal information X i The direction is direction X, the target point is T i The straight-line distance between the radar transmitting point and the i , target point T i The line between the radar transmitting point and i , connect L i The angle between the central axis Z of the radar transmission wave is angle A i , connect L i The direction is direction R.

[0019] The kmeans algorithm in step S10 includes the following steps:

[0020] Step S111: For each target point T in the data i , calculate it to K = K1, K2, K3, ..., K num The distance between the cluster centers is calculated and divided into the cluster center with the smallest distance.

[0021] Step S112: For each category K j , recalculate its cluster center

[0022] Step S113: Repeat steps S111 and S112 until the set termination condition is reached.

[0023] The termination conditions include but are not limited to the number of iterations reaching a set value and the error change reaching a set value.

[0024] The radar detection is performed at 10 frames per second.

[0025] The valid grid is a grid with target point K i of grid.

[0026] In step S5, all target points T obtained in step S1 are i The information of and the division of the grid in direction R are used to calculate each target point T i Which grid it falls into in direction R, and the target point T in each grid i The number of .

[0027] In step S8, all target points T obtained in step S1 are i The information and the division of the grid in direction X are used to calculate each target point T i Which grid it falls into in direction X, and the target point T in each grid i The number of .

[0028] Step S6 may be placed before step S3, and step S6 and step S3 may also be performed simultaneously.

[0029] The beneficial effects of the present invention are as follows: preliminary grid division of radar target points is performed according to target characteristics, which is more specific than randomly selecting the number of cluster centers, can reduce the calculation time of kmeans, can reduce the error caused by the local optimum, and compared with the conventional method of selecting the center, eliminates the probability of selecting an abnormal value as the center. When initializing the center point, it is selected from the divided grid. The method preliminarily determines the center point information. Compared with randomly initializing the center point information, the calculation time of kmeans of the present method is shorter. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 It is a schematic diagram of the relationship between the radar transmission wave and the target point position of the present invention.

[0031] Figure 2 It is a schematic diagram of the process of the present invention;

[0032] Figure 3 It is a schematic diagram of the grid in the R direction and the X direction of the present invention;

[0033] Figure 4 It is a schematic diagram of the R direction grid of the present invention;

[0034] Figure 5 It is a schematic diagram of the X-direction grid of the present invention. DETAILED DESCRIPTION

[0035] The specific implementation of the present invention is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the present invention, and is not used to limit the present invention.

[0036] For ease of description, spatially relative terms such as "above", "above", "on the upper surface of", "above", etc. may be used here to describe the spatial positional relationship between a device or feature and other devices or features as shown in the figure. It should be understood that spatially relative terms are intended to include different orientations of the device in use or operation in addition to the orientation described in the figure. For example, if the device in the accompanying drawings is inverted, the device described as "above other devices or structures" or "above other devices or structures" will be positioned as "below other devices or structures" or "below other devices or structures". Thus, the exemplary term "above" can include both "above" and "below". The device can also be positioned in other different ways (rotated 90 degrees or in other orientations), and the spatially relative descriptions used here are interpreted accordingly.

[0037] A radar target clustering method based on target features and kmeans, such as Figures 1 to 5As shown, the following steps are included:

[0038] Step S1: Get all target points T detected by radar D in one frame i The information includes but is not limited to the target point T i The distance R i , angle A i , speed V i . i ∈(1, 2,…, m), where m is the number of target points in one frame detection.

[0039] Target point K i Horizontal information X i , X i =R i s i nA i . Let the horizontal information X i The direction is direction X.

[0040] In this embodiment, 10 frames per second.

[0041] Assume that in a certain transmission, the central axis Z of the radar D transmitting wave is parallel to the direction Y, and the target point T i The radial distance between the transmitting point and radar D is R i , that is, the target point T i The straight-line distance between the radar D emission point and the radar D emission point is R i Target point T i The line connecting the transmitting point of radar D is L i , then the central axis Z and the connecting line L i The angle between them is angle A i . Connect L i The direction is direction R. Figure 2 As shown, the figure schematically illustrates a radar D coordinate system in a two-dimensional plane according to an embodiment.

[0042] Step S2: preset the size R_TH of the grid in the direction R and the size X_TH of the grid in the direction X according to the actual test scenario and test objectives.

[0043] In this step, the size of the grid refers to the size of a single grid.

[0044] Step S3: Find all target points T i The distance R in the direction R i The maximum value R max and the minimum value R min .

[0045] Step S4: Calculate the number of grids R in the R direction num ,

[0046] Step S5: According to R num and the effective number of grids K in the R_TH calculation direction R Rnum The valid grid is the grid with the target point T i Specifically, according to all the target points T obtained in step S1 i The information of and the division of the grid in direction R can be used to calculate the target points T i Which grid it falls into in direction R, and the target point T in each grid i The number of .

[0047] Step S6: Find all target points T i Horizontal information X in direction X i The maximum value X max and minimum value X min .

[0048] Step S7: Calculate the number of grids X in direction X num ,but

[0049] Step S8: According to X num and X_TH calculates the number of effective grids K in the direction X Xnum Specifically, according to all the target points T obtained in step S1 i The information of and the division of the grid in direction X can be used to calculate the target point T i Which grid it falls into in direction X, and the target point T in each grid i The number of .

[0050] Step S9: Select K Rnum or K Xnum K is the number of center points of the kmeans algorithm num .

[0051] It should be noted that step S6 can be placed before step S3, and step S6 and step S3 can also be performed simultaneously, that is, step S3 searches for all target points T i The distance R in the direction R i The maximum value R max and the minimum value R min , and step S6 finds all target points T i Horizontal information X in direction X i The maximum value X max and minimum value X min , the order of the two does not affect the detection.

[0052] Step S10: Select a target point T from each grid at random i As the center point, initialize Knum Center point K j =T i(Xab) or K j =T i(Rab) , where j∈(1, 2, …, num), Xab represents the b-th point of the a-th grid in the X direction, and Rab represents the b-th point of the a-th grid in the R direction.

[0053] When K num =KR num When K j =T i(Rab) , which means that when the number of center points K of the kmeans algorithm num =K Rnum When K is initialized num The center points are located in the R direction. Rnum in a grid.

[0054] When K num =K Xnum When K j =T i(Xab) , which means that when the number of center points K of the kmeans algorithm num =KX num When K is initialized num The center points are located in the X direction. Xnum in a grid.

[0055] For steps S4 to S10, in this embodiment, Figure 3 and 4 As shown, the horizontal axis direction represents the direction R, and the vertical axis direction represents the direction X. In the figure, the dark grids marked with numbers represent valid grids, that is, grids with target points or objects, and the numbers in the dark grids represent the number of target points in this grid. The light grids are invalid grids, which means that there are no target points or objects in this grid.

[0056] like Figure 4 As shown, in this embodiment, the calculated R num = 10, in direction R, the effective number of grids K Rnum =7, where R min To R max In the direction of , the first four grids and the last three grids are valid grids, and the middle three grids are invalid grids. The number of target points in the first four valid grids is 1, 3, 2, 2, respectively, and the number of target points in the last three valid grids is 3, 6, 3, respectively.

[0057] like Figure 5 As shown, in this embodiment, the calculated X num =6, in direction X, the effective number of grids K Xnum =5, where from Xmin To X max In the direction of , the first five grids are valid grids, and the last grid is an invalid grid. The number of target points in the first five valid grids is 6, 3, 3, 6, and 2 respectively.

[0058] When K is selected Rnum =7 as K num When the value of K is selected, a target point is selected as the center point in each valid grid in the direction R and initialized. Xnum =5 as K num When the value of is equal to , a target point is selected as the center point in each valid grid in direction X and initialized. That is, for direction X or direction R, there is a center point in each valid grid in each dimension.

[0059] Step S11: K num The center points are used as clustering centers and the kmeans algorithm is performed.

[0060] Step S11 includes the following steps:

[0061] Step S111: For each target point T in the data i , calculate it to K num The distance between the cluster centers is calculated and the cluster is divided into the class with the smallest distance.

[0062] Step S112: For each category K j , recalculate its cluster center Among them, ci are all the j The formula in this step calculates the target point belonging to category K. j The centroid of all samples.

[0063] Step S113: Repeat steps S111 and S112 until the set termination condition is reached.

[0064] In step S113, the termination condition is that the number of iterations reaches a set value, the error change reaches a set value, etc.

[0065] Finally, it is necessary to explain here that the above embodiments are only used to further illustrate the technical solution of the present invention in detail and cannot be understood as limiting the scope of protection of the present invention. Some non-essential improvements and adjustments made by technicians in this field based on the above content of the present invention all fall within the scope of protection of the present invention.

Claims

1. A radar target clustering method based on target features and kmeans, characterized in that: The steps include: Step S1: Get all target points T detected by the radar in one frame i The information includes the target point T i The distance R i , angle A i , speed V i , where i∈(1,2,...m); Step S2: preset the size R_TH of the grid in direction R and the size X_TH of the grid in direction X according to the actual test scenario and test target; i =R i sina i ; Among them, X i The target point T i Horizontal information, horizontal information X i The direction is direction X, the target point is T i The straight-line distance between the radar transmitting point and the i , target point T i The line between the radar transmitting point and i , connect L i The angle between the central axis Z of the radar transmission wave is angle A i , connect L i The direction is direction R; Step S3: Find all target points T i The distance R in the direction R i The maximum value R max and the minimum value R min ; Step S4: Calculate the number of grids R in the R direction num , Step S5: According to R num and the effective number of grids K in the R_TH calculation direction R Rnum ; Step S6: Find all target points T i Horizontal information X in direction X i The maximum value X max and minimum value X min ; Step S7: Calculate the number of grids X in direction X num ,but Step S8: According to X num and X_TH calculates the number of effective grids K in the direction X Xnum ; Step S9: Select K Rnum or K Xnum K is the number of center points of the kmeans algorithm num ; Step S10: Select a target point T from each grid at random i , initialize K num Center points, K num The center points are used as clustering centers and the kmeans algorithm is performed.

2. The target clustering method according to claim 1, characterized in that: The kmeans algorithm in step S10 includes the following steps: Step S111: For each target point T in the data i , calculate it to K = K1, K2, K3, ..., K num The distance between the cluster centers is calculated and divided into the cluster center with the smallest distance. Step S112: For each category K j , recalculate its cluster center Among them, c i are all the j The target point; Step S113: Repeat steps S111 and S112 until the set termination condition is reached.

3. The target clustering method according to claim 2, characterized in that: The termination conditions include the number of iterations reaching a set value and the error change reaching a set value.

4. The target clustering method according to claim 1, characterized in that: The radar detection is performed at 10 frames per second.

5. The target clustering method according to claim 1, characterized in that: The valid grid is a grid with target point K i of grid.

6. The target clustering method according to claim 1, characterized in that: In step S5, all target points T obtained in step S1 are i The information of and the division of the grid in direction R are used to calculate each target point T i Which grid it falls into in direction R, and the target point T in each grid i The number of .

7. The target clustering method according to claim 1, characterized in that: In step S8, all target points T obtained in step S1 are i The information and the division of the grid in direction X are used to calculate each target point T i Which grid it falls into in direction X, and the target point T in each grid i The number of .

Images