High Performance Online Loop Closure Detection for Topological Mapping

Abstract

In simultaneous localization and mapping (SLAM) system, loop closing is defined as the correct identification of a previously visited location. Loop closing is essential for the precise self-localisation of the robot; however, the performance of loop closure detection is seriously affected by dynamic objects and perceptual aliasing in the environment. In the traditional likelihood matching methods, the number of matching words and the difference between them are not considered. This paper proposes a method based on mixed similarity to calculate the similarity score, thereby improving the performance of closed-loop detection. Experiments are performed on datasets from dynamic environments and visual repetitive environments, and then this method can produce a higher recall rate with 100% accuracy compared to the latest methods.