RGB-D Visual Odometry Algorithm Based on Improved ORB for Indoor Environments

Pose estimation and map reconstruction are necessary conditions for autonomous robots. Considering the low positioning accuracy of autonomous robots in indoor environments, this paper presents an RGB-D visual odometry algorithm for indoor environment, which based on an improved ORB. The improved ORB algorithm based on quadtree form is used to extract the features of visual odometry. Then, optimizing camera poses by ICP algorithm. In addition, this paper selects the key frame to obtain higher accuracy of position estimation and constructs a partial map. Experiments demonstrate that the proposed RGB-D visual odometry can obtain accurate and robust estimation results in indoor environments.