3D Point Cloud Stitching for Object Detection with Wide FoV Using Roadside LiDAR

Light Detection and Ranging (LiDAR) is widely used in the perception of physical environment to complete object detection tracking tasks. The current methods datasets are mainly developed for autonomous vehicles, which could not be directly roadside perception. This paper presents a 3D point cloud stitching method with wide horizontal field view (FoV) using LiDAR. Firstly, base model trained by KITTI dataset has achieved accuracy 88.94. Then, new range 180° can inferred break limitation camera’s FoV. Finally, multiple sets results from single LiDAR stitched build 360° solve problem overlapping objects. effectiveness proposed approach been evaluated collected clouds. experimental show that offers cost-effective solution achieve larger FoV, number output objects increased 77.15% more than model, improves performance