Optimization algorithm for closed-loop detection based on similarity fusion mechanism

To solve the problems of low closed-loop accuracy and error-prone loopback in the closed-loop detection thread of traditional SLAM algorithm, this article proposes a new similarity fusion mechanism for closed-loop detection optimization algorithm, which calculates the fused similarity of key frames of suspected loopback frames for further validation, and selects high-quality loopback key frames to improve the efficiency of closed-loop detection while reducing the occurrence of error loopback. The algorithm is validated in the KITTI public dataset and can effectively improve the accuracy of the SLAM system and generate globally consistent maps.