An Improved Fast Ground Segmentation Algorithm for 3D Point Cloud

Abstract

In this paper, we propose an improved algorithm to divide the point cloud collected from Lidar into ground points and non-ground points. The primary purpose of our method is to improve the accuracy and guarantee data processing speed. In every frame of the point cloud, we divide the point cloud into different blocks basing on the vertical lines, and threshold points and ground initial points are then analyzed in each block. According to the priori approach, we complement two more restrictions to pick the threshold points and the ground initial points, which are height threshold and vertical distance. These conditions take the vertical extreme distance information into account. After the threshold points and ground initial points are labelled, all points between ground initial points and threshold points are ground points, and the other points are non-ground points. The whole process is completed after all points are labelled. We do some experiments based on our proposed method, and the algorithm can reach 22 frames per second. The results prove that our method has higher accuracy and the efficiency meets the real-time requirements.