A Fast and Robust Triangle Descriptor for Loop Closure Detection in 3D LiDAR SLAM

Loop closure detection is one of the important steps in simultaneous localization and mapping (SLAM) to eliminate accumulated errors. To successfully detect loops, the capability of 3D place recognition is required. In this work, we design a 3D place recognition descriptor called FR-STD based on the global descriptor Stable Triangle Descriptor (STD). FR-STD proposes a local descriptor Fast Point Feature Vector (FPFV) designed for triangles, which is used to enhance the differentiation between triangle descriptors with similar side lengths and reduce the probability of mismatching. FR-STD is a robust descriptor that combines global and local information. We evaluate its performance on publicly available datasets, and the results demonstrate that, whether in spinning LiDAR or solid-state LiDAR scenarios, FR-STD outperforms STD.