Visual SLAM with RGB-D Cameras

This paper focuses on visual SLAM with RGB-D cameras (abbreviated as RGB-D SLAM), which has been an actively studied issue in the robotics community since RGB-D cameras can obtain depth information of environments simply. Firstly, two types of RGB-D cameras are introduced according to the principle of depth measurement. Secondly, the typical RGB-D SLAM algorithm framework is normally divided into four parts: visual odometry, optimization, loop closing and mapping. Thirdly, a series of landmark achievements on algorithm, open source libraries and tools, and performance evaluation of RGB-D SLAM are summarized. Finally, the advantages and the disadvantages, as well as the development trends of RGB-D SLAM are discussed.