Fusion Scan Context: A Global Descriptor Fusing Altitude, Intensity and Density for Place Recognition

Loop closure detection is a necessary and challenging problem in simultaneous localization and mapping (SLAM). It is often processed by using light detection and ranging (LiDAR) or camera sensor. It is essential to describe laser point cloud information with structural information due to the lack of texture of laser point clouds. This paper presents a comprehensive global descriptor, fusion scan context (FSC), that integrates various information including altitude, intensity and density. To improve the efficiency of loop closure detection, a three-stage matching method is designed, including ring key searching stage, sector key matching stage and similarity comparison stage. Comprehensive experiments based on public datasets have been performed to evaluate the proposed descriptor, and they indicate that the descriptor has better performance than similar ones in loop closure detection.