Loop Detection Method Based on Multi-Frame Background Point Distance Histogram Encoding

Loop closure detection is a crucial component in Simultaneous Localization and Mapping (SLAM) for constructing high-precision maps, as it can eliminate cumulative errors generated during the operation of SLAM systems. To address the shortcomings of traditional loop closure detection methods in reverse loop closure detection, we propose a method based on multi-frame background points. Firstly, we segment the current point cloud into regions and identify the farthest point as the background point. Then, we encode the relative distance relationships of all background points into histograms. Finally, by comparing the scores of the current histogram with those of candidate histograms, as well as the scores from the previous five frames, we determine whether a loop closure has occurred. We conducted comparative experiments on the KITTI dataset, which is widely used in the autonomous driving field, against various loop closure detection algorithms. The results demonstrate that our method effectively identifies loop closures while ensuring real-time performance.