EnhanceERASOR: Two-Stage Static 3D Point Cloud Mapping in Dynamic Scenes
Shuyang Yu, Yi Wu, Xiaoqing Guan, Song Jin, Haoxiang Liu, You Wang, Guang Li
AI summary
Problem
Dynamic objects in real-world environments create ghost traces in accumulated 3D point cloud maps, which degrade localization and path planning for autonomous systems. Existing online mapping methods often suffer from false positives and negatives due to limited temporal context and occlusion.
Approach
The method uses a lightweight online scan-to-scan stage with region-wise pseudo occupancy descriptors and consistency checks to filter dynamic points in real-time, followed by an offline stage that refines the map using submap consistency analysis and voxel-guided dense mapping.
Key results
- 0.981 F1 score on SemanticKITTI Seq 05
- Submap consistency checks suppress semi-dynamic objects
- Robust generalization across diverse LiDAR sensors and scenarios
- Real-time operation with low computational overhead
Why it matters
Provides autonomous vehicles and robots with reliable, high-fidelity static maps essential for safe navigation and planning in dynamic environments.
Abstract
A clean map of the surrounding environment is essential for autonomous driving systems to ensure reliable localization and safe path planning. However, the existence of dynamic objects introduces ghost traces into the map, significantly degrading its quality. To address this issue, we propose EnhanceERASOR, a two-stage framework for static 3D point cloud mapping, consisting of a lightweight Online- ERASOR stage for real-time static mapping and an Offline- Refinement stage for global optimization. The Online-ERASOR stage utilizes the egocentric ratio of pseudo occupancy between consecutive scans to identify dynamic points, followed by verifi- cation and post-processing strategies to suppress false positives and false negatives. The Offline-Refinement stage introduces a submap-to-map consistency check to suppress semi-dynamic and slow-moving objects, and adopts a voxel-guided strategy for dense static mapping. Extensive experiments on diverse datasets with different scenarios and sensors demonstrate the superior performance, robustness, and generalization ability of our proposed method in static map construction.