A Dynamic Points Removal Benchmark in Point Cloud Maps

• DOI: 10.48550/arxiv.2307.07260
• Publication Date: 2023-01-01
• Authors: Qingwen Zhang; Daniel Duberg; Ruoyu Geng; Mingkai Jia; Lujia Wang; Patric Jensfelt
• Keywords: Benchmarking; Benchmark (surveying); CLARITY; Representation; Code (set theory)

In the field of robotics, point cloud has become an essential map representation. From perspective downstream tasks like localization and global path planning, points corresponding to dynamic objects will adversely affect their performance. Existing methods for removing in clouds often lack clarity comparative evaluations comprehensive analysis. Therefore, we propose easy-to-extend unified benchmarking framework evaluating techniques maps. It includes refactored state-of-art novel metrics analyze limitations these approaches. This enables researchers dive deep into underlying reasons behind limitations. The benchmark makes use several datasets with different sensor types. All code related our study are publicly available further development utilization.