A Review of Visual SLAM for Dynamic Objects

Abstract

With the wide application of SLAM in fields such as autonomous driving and robot navigation, higher requirements for human-robot interaction capabilities in real scenes have been put forward, so the processing of dynamic targets has developed into one of the research hotspots of SLAM. In this review, visual SLAM is divided into two categories according to the different ways for dynamic target processing: static background map reconstruction, tracking and real scene reconstruction of dynamic targets. This paper discusses the two types of SLAM methods, summarizes the advantages and disadvantages of different methods, compares the experimental results of some methods, and finally provides an outlook on the future trend of visual SLAM and concludes the whole review.