Plane detection for Kinect image sequences

Ability to effectively and efficiently detect planes is very useful for mobile robot navigation. This is because plane structures are abundant in man-made environment. Existing methods for plane detection are derived for different kinds of input data. In this paper, we propose a plane detection method for image sequences from Kinect. The algorithm takes into account the limitation of the Kinect's depth camera by using visual data to help detect planes beyond accessible range of the depth camera. Unlike other plane detection methods, sequence of images from Kinect is also used for plane tracking which helps increase accuracy and automatically provides odometry information. The algorithm is implemented and preliminary experimental results are presented.