Refinement of Edge Extraction Based on Lidar Point Cloud

摘要

Lidar has been used in many areas of computer vision and robotics, like 3D reconstruction, autonomous navigation and positioning. Extracting accurate edges for 3d point clouds plays a key role in some applications. However, few people are concerned about this issue. In this paper, we present a novel pipeline for refining the position of the edge using the rotation of the lidar itself and its depth discontinuity at the edges. Our method consists of two parts. First, edge points are extracted according to the depth difference in the reference frame of point cloud. Then, we apply our refining strategy to obtain the more accurate edge points. The novelty is that our approach compensates for hardware deficiencies through algorithms, improving the accuracy of lidar edge points. The experiment provides comparisons of the edge points extracted by the direct method and our method.