Vision Odometer Based on RGB-D Camera

Abstract

In order to solve the problem of self-localization of robots in unknown environment, a method based on RGB-D camera is proposed. RGB-D camera is a visual sensor that can collect RGB images and depth images simultaneously in recent years. Firstly, the introduction of the mainstream RGB-D camera is introduced. Then, the ORB feature of the current image is extracted from the visual odometer and the features are matched with the key frames. The RANSAC algorithm is used to estimate the interframe motion iteratively. Finally, the key frames are extracted and utilized the ICP is optimized to get the optimal registration of key frame pose. The method can effectively improve the positioning accuracy of mobile robots.