A visual odometry system based on Kinect sensor

In this paper we present a system that estimates the motion of a Kinect camera based on video input. Motion is estimated by Kinect sensor itself. First of all, the infrared camera of Kinect is calibrated by clicking corners manually and the corners position are optimized, the relative pose of RGB camera and infrared camera for Kinect is calibrated by the optimized points, furthermore, Harris Point features are matched between pairs of frames and linked into image trajectories at video rate by proposed method, calculate the 3D coordinate of matched points in IR image, the outliers are deleted by RANSAC. At last, the robot trajectory is estimated by improved ICP method (Iterative Closest Point). The experiment shows that the proposed method has more precision in contrast with conventional method.