An Enhanced GC-RANSAC Approach for Subway Indoor Ground Segmentation

Abstract

Addressing the challenges of traditional LiDAR ground segmentation algorithms in processing complex subway scenes captured, an improved GC-RANSAC ground segmentation algorithm is proposed. This method initially uses the RANSAC algorithm to extract several planes. Then, it calculates the slope threshold for each plane to preserve non-ground areas. Finally, the min-cut algorithm is utilized to complete the ground segmentation. Based on the quadruped robots experimental platform, empirical data from subway scenarios are collected and tested. The processing results show the proposed algorithm demonstrates excellent performance compared with other selected methods. The False Negative Rates are 0.56%, 0.84%, and 1.67% for elevators, subway station corridors, and subway station halls respectively, which meets the effectiveness requirements for quadruped robots.