基于深度图的三维激光雷达点云目标分割方法

Target Segmentation Method for Three-Dimensional LiDAR Point Cloud Based on Depth Image

三维激光雷达广泛应用在智能车系统中,点云目标分割是智能车环境感知中的关键技术。针对目前三维激光雷达点云目标分割算法实时性和准确性不高的问题,提出一种基于深度图的点云目标快速分割方法。将点云数据表示为深度图,建立深度图与点云数据的映射关系。利用激光雷达扫描线的角度阈值去除地面点云数据,结合深度图和自适应参数改进的DBSCAN(Density-Based Spatial Clustering of Applications with Noise)算法对非地面点云进行聚类分割。实验结果表明该方法相对于传统聚类算法在时间效率上有很大的提升,且能较好地降低欠分割错误率,分割准确度提升10%,达到了85.02%。

Point cloud target segmentation is the key to perceive targets for a smart car using three-dimensional(3D) LiDAR. Aiming at the problems of poor real-time and low accuracy of the existing in 3D LiDAR point cloud target segmentation algorithms, an approach based on a depth map is proposed in this paper to realize fast and accurate segmentation for point cloud target segmentation. The original data are transformed into a depth map, and the mapping relationship between point cloud data and a depth map is established. After removing the ground point cloud data by using the angle threshold of the LiDAR scanning line, the non-ground point cloud is clustered and segmented by the improved DBSCAN(Density-Based Spatial Clustering of Applications with Noise) algorithm combined with the depth map and the adaptive parameters. Experimental results show that the proposed method has a significant improvement in time efficiency compared with the traditional clustering algorithms. Moreover, the under-segment error rate is decreased while the segmentation accuracy is increased by 10% to 85.02%.