Learning Planar Geometric Scene Context Using Stereo Vision

A reliable method for detecting planar regions in a video/stereo scene would be of great use to the field of computer vision. Solutions to this problem are applicable to object recognition, scene identification, and robot-related applications. In this paper we present a plane-finding algorithm that uses data from a binocular stereo camera system to produce labeled output images showing the major planes in a video scene. The algorithm is based on the three-dimensional Hough Transform but also presents many useful approaches and heuristics applicable to general plane-finding.