Detection and Utilization of Reflections in LiDAR Scans through Plane Optimization and Plane SLAM

Area of Science:

• Robotics and Sensor Fusion
• Computer Vision
• 3D Reconstruction

Background:

• Reflections from materials like glass and mirrors in LiDAR sensing lead to inconsistent data.
• These inconsistencies pose significant challenges for robotic localization, mapping, and navigation.
• Existing methods struggle to accurately account for reflective surfaces.

Purpose of the Study:

• To develop a robust method for identifying and mapping reflective planes in LiDAR data.
• To improve the accuracy of LiDAR-based localization and mapping by mitigating reflection-induced errors.
• To enhance the applicability of LiDAR sensing in complex environments with reflective surfaces.

Main Methods:

• Constructing a global, optimized map of reflective planes by refining plane detection parameters across multiple LiDAR scans.
• Integrating reflective plane estimation into a plane SLAM (Simultaneous Localization and Mapping) algorithm.
• Classifying LiDAR readings based on the optimized reflective plane map.

Main Results:

• The proposed method achieves superior classification accuracy for LiDAR readings compared to single-scan approaches.
• Demonstrated the practical applicability of the reflective plane mapping technique within a SLAM framework.
• Experimental validation confirmed significant improvements in handling reflective environments.

Conclusions:

• Optimized mapping of reflective planes effectively addresses data inconsistencies in LiDAR sensing.
• The developed approach enhances the reliability of robotic localization, mapping, and navigation in environments with reflective surfaces.
• Open-source code and data are provided for reproducibility and further research.