Semantics Aware Loop Closure Detection in Visual SLAM

Abstract

Loop closure detection is of vital importance to simultaneous localization and mapping for robot motion in an unknown environment. This research reviews the deep learning approaches using semantic information for loop closure detection. In view of shortcomings of the existing research, an improved loop closure detection method is proposed in this research fusing semantic information with a feature-based Bag-of-Words model. RefineNet is used for high-resolution semantic segmentation and dense semantic feature extraction. Semantics being invariant to viewpoint changes and dynamic environment can improve the overall performance of the environment.