Abstract: Point cloud clustering is a key step in the realization of environment perception of unmanned vehicle by lidar. It clusters the discrete points in the point cloud map constructed by lidar into individual whole, which provides the necessary basis for subsequent classification and tracking. In this paper, the clustering algorithms applied to unmanned vehicle point cloud clustering are divided into six categories, which are the existing clustering algorithm based on partition, clustering algorithm based on hierarchy, clustering algorithm based on density, clustering algorithm based on grid, clustering algorithm based on distance and hybrid clustering algorithm. Based on the systematic analysis of various clustering algorithms, the problems, solutions and performance in the process of point cloud clustering are compared and analyzed. Considering the accuracy and real-time requirements of unmanned vehicle point cloud clustering, the combination of edge algorithm, hybrid clustering algorithm and new clustering algorithm will be the research hotspot of unmanned vehicle point cloud clustering, and also the research focus of unmanned vehicle point cloud clustering in the future.

Key words: Point Cloud Clustering, Lidar, Driverless, Clustering Algorithm