LiDAR Filtering in 3D Object Detection Based on Improved RANSAC

At present, the LiDAR ground filtering technology is very mature. There are fewer applications in 3D-object detection due to limitations of accuracy and efficiency. If can be removed quickly accurately, algorithm detect objects more accurately quickly. In order meet application requirements detection, inspired by Universal-RANSAC, we analyze detailed steps RANSAC propose a precise efficient RANSAC-based method. The principle GroupSAC analyzed, sampled points grouped attributes make it easier sample correct point. Based on this principle, devise method for limiting that applicable point clouds. We describe preemptive detail. Its breadth-first strategy adopted obtain optimal plane without complex iterations. use International Society Photogrammetry Remote Sensing (ISPRS) datasets KITTI dataset testing. Experiments show our has higher efficiency compared with currently widely used methods. explore methods experimental results improve object affecting