Improved closed-loop detection and Octomap algorithm based on RGB-D SLAM

Abstract

In order to solve the problems of the inaccuracy of RGB-D SLAM closed-loop and the map sparse outliers, this paper proposes an improved algorithm of Closed-loop Detection and Octomap mapping. In the improved algorithm, the curvature of the robot's motion trajectory is combined with the cyclic closure detection algorithm to eliminate the difficulties of the front-end cumulative error to the back-end Closed-loop Detection; in the aspect of map sparse outliers, in order to make the map more compact and easy to adjust, the two side confidence interval of Gaussian distribution is combined with statistical filtering to give the initial statistical value. We have done a series of experiments in the open TUM RGB-D data set. The memory and outliers of point cloud map are reduced by 11.4%, 11.3% respectively, and the memory and outliers of Octomap are reduced by 26.7%, 27.3% respectively, and the validity of accurate closed-loop is verified.