A Vertical Distribution Descriptor Based on Static Point Cloud Map Building from 3D LiDAR Data

Static maps are essential for the proper functioning of autonomous mobile platforms. However, dynamic objects may occur during mapping, leaving artifacts that can affect downstream tasks such as navigation and localization. We propose a dynamic objects removal algorithm based on vertical distribution descriptors (VDD) to solve this problem. Firstly, the regions of interest in the current LiDAR frame and submap are divided into multiple unit spaces to ensure low execution time of the algorithm. The VDD is computed for each unit space, and the Bhattacharyya coefficient is used to compare the differences between corresponding unit spaces and identify dynamic regions. Finally, to reduce the misidentification of static points, the ground points of the dynamic region are extracted to restore it to the static map. Using the Semantic KITTI dataset for validation, the results indicate that our method can effectively and robustly filter dynamic points.