Online Mapping from the Weight Matching Odometry and Highly Dynamic Point Cloud Filtering via Pseudo-Occupancy Grid

The efficient locomotion of autonomous driving and robotics suggests a clearer visual-ization and a more precise map. This paper presents a high accuracy online mapping with weight matching LiDAR-IMU-GNSS odometry and object-level highly dynamic point cloud filtering method based on pseudo-occupancy grid. The odometry integrates IMU pre-integration, ground points through progressive morphological filtering(PMF), motion compensation and weight feature point matching. Weight feature point matching enhances alignment accuracy by combining geometric and reflectance inten-sity similarities. By justifying the pseudo-occupancy ratio between the current frame and prior local submaps, grid probability value is updated to identify the distribution of dynamic grids. Object-level point cloud clusters segmentation is obtained using curved voxel clustering method, eventually leading to filtering out the object-level highly dy-namic point clouds during the online mapping process. The proposed odometry, in comparison with LIO-SAM and FAST-LIO2 frameworks, shows superior accuracy in the KITTI, UrbanLoco, and Newer College (NCD) datasets. Meantime, the proposed highly dynamic point cloud filtering algorithm also shows a better detection precision than the performance of Removert and ERASOR. Furthermore, the high accuracy online map-ping is built from a real-time dataset with the comprehensive filtering of driving vehi-cles, cyclists and pedestrians. This research contributes to field of the high accuracy online mapping, especially in filtering the highly dynamic objects in an advancing way.