Active loop closing based on laser data in indoor environment

In Artificial Intelligence, Loop Closing is a key issue of Simultaneous Localization and Mapping (SLAM) that helps SLAM work efficiently and robustly. To solve the problem of the quantity of scene matching increasing linearly with time when loop closing in SLAM for mobile robots, a laser range finder based method is proposed for classifying the geometric scene and determining the set of similar frames for indoor corridor environment. Firstly, based on the turning function, a new line segment extraction method is designed to swiftly obtain the segment feature of scenes. Secondly, the set of basically similar frames is constructed according to scene entropy, scan area and close scan area. Finally, the method describing geometric scenes with the sequences of turning angle is proposed based on the turning angle histogram. By using Longest Common Subsequence matching and Hu-moment-based contour matching, the compact set of similar frames and the best matched frames of query frame are found. The experiment proves that the compact set and the best matched frames can be obtained efficiently and accurately, and based on that, the quantity of scene matching could be greatly reduced.