P‐17.3: Obstacle detection based on the fusion of multiple solid‐state lidars

Abstract

In automatic driving, mechanical lidar is often used to detect the surrounding environment. Due to cost‐effective reasons, solidstate lidar is now also a good choice. It can obtain a larger number of point clouds. But this means that there are more point clouds that need to be processed, and the requirements for real‐time are more stringent. At the same time, when the vehicle is moving, it is difficult to segment obstacles, which is prone to missed detection and false detection. exist. To solve the above problems, this paper adopts the detection method of multi‐solidstate lidar fusion, and proposes a RANSAC ground detection algorithm with adaptive iteration times. At the same time, the Euclidean clustering method is improved to make it based on the change of the distance radius detected by the point cloud. The threshold is automatically corrected, so that the automatic driving can detect the surrounding environment quickly and accurately. The method in this paper has also been put into actual vehicle experiments, and the test results show that the method in this paper can indeed achieve the expected goal.