Improving the Data Association in Monocular Loop-Closing Detection

Loop-closing detection is one of the most challenging issues in the monocular Simultaneous Location and Mapping (SLAM). Reliable data association is crucial to this issue as spurious matches could prove to be fatal. In this paper, a new approach is proposed to provide a robust data association for the monocular loop-closing detection. The proposed approach is characterized by a combination of the active search algorithm and the Joint Compatibility Branch and Bound (JCBB) algorithm. Experiments are carried out on indoor loop-closing image sequences and results show that the new approach could improve the data association in monocular loop-closing detection.