An improved algorithm of point cloud extraction for automotive radar

Abstract

In this paper, a point cloud extraction method based on short-range vehicle millimeter wave radar is proposed. This method is based on Cluster CLEAN algorithm. Compared with the traditional peak extraction algorithm, it can obtain more abundant point cloud data, which is conducive to the subsequent target classification. In this paper, the Cluster CLEAN algorithm is improved, and a improved incremental Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm is proposed to improve the clustering process of the Cluster CLEAN algorithm. It is not necessary to repeatedly Cluster the past scattering point data in each iteration, and the computational efficiency of the algorithm can be improved by more than 98%. In addition, in this paper, a modified order statistics based multidimensional clustering (OSMC) algorithm is proposed to directly Cluster the clustering results of Cluster CLEAN algorithm in angle dimension, without the need to perform range-velocity-angle clustering after angle dimension estimation, which accelerates the efficiency of the algorithm and reduces the redundancy of the algorithm. It is verified that compared with the traditional peak extraction algorithm, the signal processing algorithm in this paper can increase the number of scattering points from the same target by 2-7 times while maintaining real-time performance.