Efficient multi-modal high-precision semantic segmentation from MLS point cloud without 3D annotation

Yuan Wang, Pei Sun, Wenbo Chu, Yuhao Li, Yiping Chen, Hui Lin, Zhen Dong, Bisheng Yang, Chao He

Abstract: Highlights•High-precision semantic segmentation of MLS point cloud is achieved without labeling 3D samples.•Multi-frame images fusion and semantic predictions mapping strategy are proposed to provide better coarse segmentation.•A novel hierarchical geometric constraints method is proposed to derive more refined semantic results.