🤖 AI Summary
This work addresses the challenge of ground segmentation in LiDAR point clouds caused by reflective noise and isolated ground points. The authors propose a polar grid map–based method that employs intra-segment and inter-segment ground labeling strategies to identify visible, occluded, and isolated ground cells. Reflective noise is explicitly modeled through the introduction of “noisy ground cells,” while an uncertainty-aware adaptive slope mechanism enhances robustness in complex terrains. Point-level segmentation is achieved via fine-grained elevation estimation. Evaluated on four public datasets, the method outperforms existing non-learning approaches, achieving state-of-the-art F1 score, accuracy, and mean Intersection over Union (mIoU). It runs at 135 Hz for 64-beam and 487 Hz for 32-beam LiDAR scans on a single CPU thread.
📝 Abstract
In LiDAR-based environment perception systems, ground segmentation is a key preprocessing step supporting various applications such as mapping and navigation. Although extensively studied, problems such as reflection noise and isolated ground remain challenging. To address these issues, we propose FugSeg, a fast uncertainty-aware ground segmentation method. A polar grid map is adopted as the point cloud representation to ensure generalizability across LiDAR types. Building on that, we develop a within- and cross-segment ground labeling strategy that identifies not only directly visible ground cells but also those that are isolated or occluded. During this process, an adaptive slope is introduced, which incorporates measurement uncertainties to enhance its reliability under complex terrain. Finally, to achieve point-level ground segmentation, a fine-grained ground elevation estimation method is introduced. Throughout the complete workflow, reflection noise is explicitly handled via the proposed noisy ground cells. We conduct comprehensive evaluations on four public datasets covering both structured and unstructured environments. Results show that FugSeg outperforms state-of-the-art non-learning methods, achieving the highest F1, accuracy, and mIoU across all datasets, while maintaining the fastest runtime (135 Hz and 487 Hz for 64- and 32-layer LiDARs) using a single CPU thread, making it suitable for resource-limited systems. The code will be available at https://github.com/Leo-YuLi/FugSeg.