Detection and Utilization of Reflections in LiDAR Scans Through Plane Optimization and Plane SLAM
Yinjie Li, Xiting Zhao, Sören Schwertfeger
TL;DR
The paper tackles the challenge of LiDAR reflections from glass and mirrors that degrade localization and mapping. It introduces a global reflective-plane map built from multi-frame observations and two pipelines: Reflection Detection via Plane Optimization (using external pose) and Reflection SLAM (plane-based SLAM without external poses). The work provides a complete methodology for per-scan glass detection, plane map maintenance, and global point classification, complemented by a plane registration framework and Gauss-Newton pose optimization, and demonstrates improved reflection removal and mapping, including the ability to map around corners. The authors validate their approach on the 3DRef dataset, compare against deep-learning baselines, and release open-source code and data, highlighting practical impact for robust indoor navigation and 3D reconstruction in reflective environments.
Abstract
In LiDAR sensing, glass, mirrors and other material often cause inconsistent data readings, because the laser beams may report the distance of the glass, the distance of the object behind the glass or the distance to a reflected object. This causes problems in robotics and 3D reconstruction, especially with respect to localization, mapping and thus navigation. With dual-return LiDARs and other methods, one can detect the glass plane and classify the points in a single scan. In this work we go one step further and construct a global, optimized map of reflective planes, in order to then classify all LiDAR readings at the end. As our experiments will show, this approach provides superior classification accuracy compared to the single scan approach. The code and data for this work are available as open source online.