CN115047424A - Millimeter wave radar point cloud target clustering method based on KDE-DBSCAN - Google Patents

Abstract

The invention discloses a KDE-DBSCAN-based millimeter wave radar point cloud target clustering method, which comprises the following steps: s1, for each point in the millimeter wave radar point cloud data set D, calculating the corresponding correlation window width W and the correlation adjacent point RN of the point according to the k value corresponding to the area where the point is located; s2, preliminarily dividing a radar point cloud data set D into different pseudo clusters FC according to the association window width W of each point and the association adjacent point set RN; s3, counting each pseudo cluster FC generated in the step S2 _i Deleting the pseudo clusters of which the number of points is less than lambda; s4, determining the optimal association radius epsilon and the minimum association point number minP of each pseudo cluster; s5, obtaining the optimal association radius epsilon and the minimum association point number minP of each pseudo cluster through the step S4, and then clustering each pseudo cluster by using a DBSCAN algorithm. The method can adapt to the point cloud distribution characteristics of the millimeter wave radar, thereby improving the clustering quality of the millimeter wave radar targetAmount of the compound (A).

Description

Millimeter wave radar point cloud target clustering method based on KDE-DBSCAN

Technical Field

The invention relates to the technical field of point cloud target segmentation clustering and identification, in particular to a millimeter wave radar point cloud target clustering method based on KDE-DBSCAN.

Background

The millimeter wave radar has the functions of measuring distance, speed and angle of a target, and has better robustness under a plurality of working conditions, so that the millimeter wave radar is more and more applied to automatic driving. In the application of millimeter wave radar, radar target point cloud clustering is an important part, and with the improvement of the resolution of the millimeter wave radar, the data volume obtained by the radar from the same target is increased, so that the target point cloud of the millimeter wave radar needs to be clustered by using a proper clustering algorithm.

At present, common clustering algorithms of millimeter wave radars are K-means clustering, DBSCAN clustering and the like, but the algorithms have better effects when processing a data set with uniform data density, but the data density of target point clouds of the millimeter wave radars is not uniform, and because the angular resolution of the millimeter wave radars in certain directions is fixed and the angular resolution of the millimeter wave radars in different directions is different, the density of the target point clouds has great relation with the distance and angle of a target, generally, the closer the distance is, the smaller the angle is, the greater the density of the target point clouds is, the farther the distance is, the larger the angle is, the more sparse the target point clouds are, and at the moment, the common clustering algorithms are difficult to show better performance.

Disclosure of Invention

The invention aims to solve the technical problem of providing a KDE-DBSCAN-based millimeter wave radar point cloud target clustering method, which can adapt to the point cloud distribution characteristics of a millimeter wave radar so as to improve the clustering quality of millimeter wave radar targets.

In order to solve the technical problems, the technical scheme of the invention is as follows: a KDE-DBSCAN-based millimeter wave radar point cloud target clustering method comprises the following steps:

s1, for each point in the millimeter wave radar point cloud data set D, calculating the corresponding correlation window width W and the correlation adjacent point RN of the point according to the k value corresponding to the area where the point is located;

s2, preliminarily dividing a radar point cloud data set D into different pseudo clusters FC according to the association window width W of each point and the association adjacent point set RN;

s3, counting each pseudo cluster FC generated in the step S2 _i Deleting the pseudo clusters of which the number of points is less than lambda;

s4, determining the optimal association radius epsilon and the minimum association point number minP of each pseudo cluster;

s5, obtaining the optimal association radius epsilon and the minimum association point number minP of each pseudo cluster through the step S4, and then clustering each pseudo cluster by using a DBSCAN algorithm.

As a preferable technical solution, the step S1 is specifically as follows:

s1-1, dividing the image into N areas according to the radial distance, wherein the N areas are respectively marked as r ₀ ，r ₁ ，……，r _N-1 ；

S1-2, traversing each point x in the point cloud data set D, and searching k nearest to the point x according to the region where the x is located _ri Adjacent points to generate Euclidean distance set D _nn (x)＝{d _j ＝d(x，x _j )|j＝[1，......，k _ri ]Where ri is 0, 1, … …, N-1, and d _j Arranged in ascending order;

s1-3, selecting kernel density function K (x), using K (x) and Euclidean distance set D in step S1-2 _nn (x) Generating distance probability density distribution function

Where h is the bandwidth of the kernel function k (x), i ═ 1, 2, … …, k _ri ；

S1-4, according to the distance probability density distribution function

Find the first maximum point of the function { (x) _j ，f(x _j ))|1＜j≤h，h≤k _ri Is then x _{j＝[1，...，h]} Associated neighbor point RN belonging to the point having an associated window width W equal to d _h 。

As a preferred technical solution, the step S2 is specifically as follows:

s2-1, carrying out ascending arrangement on the associated window width W of each point in the step S1;

s2-2, selecting the point with the minimum correlation window width W, traversing each point in the data set D, if the point is not attributed to any pseudo cluster, recursively adding the correlation adjacent point belonging to the point x to the pseudo cluster FC _i Performing the following steps;

s2-3, removing FC belonging to the pseudo cluster from the radar data set D _i And i ═ i + 1;

s2-4, repeating the steps S2-2 to S2-3 until all the points in the data set D are traversed.

As a preferred technical solution, the step S4 is specifically as follows:

s4-1, for a given pseudo cluster FC _i ＝{x _j |j≤n _i W, the associated window widths W for all points form a set { W _m |m≤n _i Recording the point x in the dummy cluster _j At a radius w _m Number of adjacent points in range is m _j Then define the function

Wherein, RN _j Is a pseudo intra-cluster point x _j By searching for w that minimizes the function f _m Then w is _m Is the optimal correlation radius of this pseudo cluster;

s4-2, FC for a given pseudo cluster _i ＝{x _j |j≤n _i Can be represented by the formula

To determine the minimum number of associated points threshold minP,

wherein, k is 1, 2.. said., n'; n' is the number of dummy clusters, n _k As a dummy cluster FC _k Inner pointNumber, totalVolume _k For a volume containing the smallest hyper-rectangle for all points in the pseudo-cluster, the hyper-rectangle may be determined by the largest and smallest value of each dimension of the data; ε Volume _k Can be determined according to the optimal association radius epsilon in the fourth step, and epsilon Volume for 1-D data _k Is 2 ∈; for 2-D data, the value is π ε ² For 3-D data, the value is

For higher dimensional data, the values are the volumes of hypercubes with side lengths of 2 ∈.

Due to the adoption of the technical scheme, the invention has the beneficial effects that: the method can adapt to the point cloud distribution characteristics of the millimeter wave radar, thereby improving the clustering quality of the millimeter wave radar target.

Drawings

The drawings are only for purposes of illustrating and explaining the present invention and are not to be construed as limiting the scope of the present invention. Wherein:

FIG. 1 is a flow chart of an embodiment of the present invention;

Detailed Description

The invention is further illustrated below with reference to the figures and examples. In the following detailed description, certain exemplary embodiments of the present invention are described by way of illustration only. Needless to say, a person skilled in the art realizes that the described embodiments can be modified in various different ways without departing from the spirit and scope of the present invention. Accordingly, the drawings and description are illustrative in nature and not intended to limit the scope of the claims.

As shown in fig. 1, a KDE-DBSCAN-based millimeter wave radar point cloud target clustering method includes the following steps:

s1, for each point in the millimeter wave radar point cloud data set D, calculating the corresponding correlation window width W and the correlation adjacent point RN of the point according to the k value corresponding to the area where the point is located;

in step S1, the following is specifically made:

s1-1, according to radial distanceIs divided into N areas which are respectively marked as r ₀ ，r ₁ ，……，r _N-1 ；

S1-2, traversing each point x in the point cloud data set D, and searching k nearest to the point x according to the area where the x is located _ri Adjacent points to generate Euclidean distance set D _nn (x)＝{d _j ＝d(x，x _j )|j＝[1，......，k _ri ]Where ri is 0, 1, … …, N-1, and d _j Arranged in ascending order;

s1-3, selecting kernel density function K (x), using K (x) and Euclidean distance set D in step S1-2 _nn (x) Generating distance probability density distribution function

Where h is the bandwidth of the kernel function k (x), i ═ 1, 2, … …, k _ri ；

S1-4, according to the distance probability density distribution function

Find the first maximum point of the function { (x) _j ，f(x _j ))|1＜j≤h，h≤k _ri Is then x _{j＝[1，...，h]} Associated neighbor point RN belonging to the point having an associated window width W equal to d _h 。

S2, preliminarily dividing the radar point cloud data set D into different pseudo clusters FC according to the correlation window width W of each point and the correlation adjacent point set RN;

in step S2, the following is specifically made:

s2-1, carrying out ascending arrangement on the associated window width W of each point in the step S1;

s2-2, selecting the point with the minimum correlation window width W, traversing each point in the data set D, if the point is not attributed to any pseudo cluster, recursively adding the correlation adjacent point belonging to the point x to the pseudo cluster FC _i The preparation method comprises the following steps of (1) performing;

s2-3, removing FC belonging to the pseudo cluster from the radar data set D _i And i ═ i + 1;

s2-4, repeating the steps S2-2 to S2-3 until all the points in the data set D are traversed.

S3, counting each pseudo cluster FC generated in the step S2 _i Deleting the pseudo clusters of which the number of points is less than lambda;

s4, determining the optimal association radius epsilon and the minimum association point number minP of each pseudo cluster;

in step S4, the following is specifically made:

s4-1, for a given dummy cluster FC _i ＝{x _j |j≤n _i W, the associated window widths W for all points form a set { W _m |m≤n _i Recording the point x in the dummy cluster _j At a radius w _m The number of adjacent points in the range is m _j Then define the function

Wherein, RN _j Is a pseudo intra-cluster point x _j By searching for w that minimizes the function f _m Then w is _m Is the optimal correlation radius of this pseudo cluster;

s4-2, FC for a given pseudo cluster _i ＝{x _j |j≤n _i Can be represented by the formula

To determine the minimum number of associated points threshold minP,

wherein, k is 1, 2.. said., n'; n' is the number of dummy clusters, n _k As a dummy cluster FC _k Number of points in, totalVolume _k For a volume containing the smallest hyper-rectangle of all points within the pseudo-cluster, the hyper-rectangle may be determined by the largest and smallest value of each dimension of the data; ε Volume _k Can be determined according to the optimal association radius epsilon in the fourth step, and epsilon Volume for 1-D data _k Is 2 ∈; for 2-D data, the value is π ε ² For 3-D data, the value is

For higher dimensional data, the value is a hyper-scale with a side length of 2 εVolume of the cube.

S5, obtaining the optimal association radius epsilon and the minimum association point number minP of each pseudo cluster through the step S4, and then clustering each pseudo cluster by using a DBSCAN algorithm.

The foregoing shows and describes the general principles, essential features, and advantages of the invention. It will be understood by those skilled in the art that the present invention is not limited to the embodiments described above, which are described in the specification and illustrated only to illustrate the principle of the present invention, but that various changes and modifications may be made therein without departing from the spirit and scope of the present invention, which fall within the scope of the invention as claimed. The scope of the invention is defined by the appended claims and equivalents thereof.

Claims (4)

1. A KDE-DBSCAN-based millimeter wave radar point cloud target clustering method is characterized by comprising the following steps:

s1, for each point in the millimeter wave radar point cloud data set D, calculating the corresponding correlation window width W and the correlation adjacent point RN of the point according to the k value corresponding to the area where the point is located;

s2, preliminarily dividing the radar point cloud data set D into different pseudo clusters FC according to the correlation window width W of each point and the correlation adjacent point set RN;

s3, counting each pseudo cluster FC generated in the step S2 _i Deleting the pseudo clusters of which the number of points is less than lambda;

s4, determining the optimal association radius epsilon and the minimum association point number minP of each pseudo cluster;

s5, obtaining the optimal association radius epsilon and the minimum association point number minP of each pseudo cluster through the step S4, and then clustering each pseudo cluster by using a DBSCAN algorithm.

2. The KDE-DBSCAN-based millimeter wave radar point cloud target clustering method of claim 1, wherein with respect to step S1, the following are specific:

s1-1, dividing the image into N areas according to radial distance, wherein the N areas are respectivelyIs recorded as r ₀ ，r ₁ ，……，r _N-1 ；

S1-2, traversing each point x in the point cloud data set D, and searching k nearest to the point x according to the area where the x is located _ri Adjacent points to generate Euclidean distance set D _nn (x)＝{d _j ＝d(x，x _j )|j＝[1，......，k _ri ]Where ri is 0, 1, … …, N-1, and d _j Arranged in ascending order;

s1-3, selecting kernel density function K (x), using K (x) and Euclidean distance set D in step S1-2 _nn (x) Generating distance probability density distribution function

Where h is the bandwidth of the kernel function k (x), i ═ 1, 2, … …, k _ri ；

S1-4, according to the distance probability density distribution function

Find the first maximum point of the function { (x) _j ，f(x _j ))|1＜j≤h，h≤k _ri }, then x _{j＝[1，...，h]} Associated neighbor point RN belonging to the point having an associated window width W equal to d _h 。

3. The KDE-DBSCAN-based millimeter wave radar point cloud target clustering method according to claim 1, wherein step S2 specifically includes the following steps:

s2-1, carrying out ascending arrangement on the associated window width W of each point in the step S1;

s2-2, selecting the point with the minimum correlation window width W, traversing each point in the data set D, if the point is not attributed to any pseudo cluster, recursively adding the correlation adjacent point belonging to the point x to the pseudo cluster FC _i Performing the following steps;

s2-3, removing FC belonging to the pseudo cluster from the radar data set D _i And i ═ i + 1;

s2-4, repeating the steps S2-2 to S2-3 until all the points in the data set D are traversed to the end.

4. The KDE-DBSCAN-based millimeter wave radar point cloud target clustering method according to any one of claims 1 to 3, wherein with respect to step S4, the following is specifically made:

s4-1, for a given dummy cluster FC _i ＝{x _j |j≤n _i W, the associated window widths W for all points form a set { W _m |m≤n _i Recording the point x in the dummy cluster _j At a radius w _m The number of adjacent points in the range is m _j Then define the function

Wherein, RN _j Is a pseudo intra-cluster point x _j Is associated with the number of neighboring points, search for w that minimizes the function f _m Then w is _m Is the optimal correlation radius of this pseudo cluster;

s4-2, for a given pseudo cluster FC _i ＝{x _j |j≤n _i Can be represented by the formula

To determine the minimum number of association points threshold minP,

wherein, k is 1, 2.. said, n'; n' is the number of dummy clusters, n _k As a dummy cluster FC _k Number of points in, totalVolume _k For a volume containing the smallest hyper-rectangle of all points within the pseudo-cluster, the hyper-rectangle may be determined by the largest and smallest value of each dimension of the data; ε Volume _k Can be determined according to the optimal association radius epsilon in the fourth step, and epsilon Volume for 1-D data _k Is 2 ∈; for 2-D data, the value is π ε ² For 3-D data, the value is

For higher dimensional data, the values are the volumes of hypercubes with side lengths of 2 ∈.

Images