TY  - JOUR
T1  - GDR: A LiDAR-Based Glass Detection and Reconstruction for Robotic Perception
AU  - Taufiqurrahman, Akbar 
AU  - Shin, Soo-Young 
JO  - The Journal of Korean Institute of Communications and Information Sciences
PY  - 2025
DA  - 2025/1/1
DO  - 10.7840/kics.2025.50.6.984
KW  - Glass detection
KW  - glass reconstruction
KW  - robot perception
KW  - LiDAR
KW  - feature extraction.
AB  - This paper proposes GDR, a LiDAR-based glass detection and reconstruction for robotics perception. This method addresses the challenges LiDAR faces as an optical sensor, particularly its difficulty in achieving comprehensive perception of glass planes in its environment. Using a proposed glass detection method based on the unique characteristics of glass points, GDR separates them from non-glass points in the input point cloud. First, feature extraction is performed based on the characteristic distance values of glass points. Then, feature extraction continues on the intensity values. Based on these two feature extractions, glass points are identified and used by the glass reconstruction method to reconstruct the glass plane in the input point cloud, resulting in a corrected point cloud with a comprehensive perception of the glass planes in the environment.
GDR is validated through several experiments, yielding an overall glass plane detection accuracy of 83.67% which outperforms the previous method. Additionally, the overall success rate and accuracy for glass reconstruction are 100% and 96.77%, respectively.