Optical flow based plane detection for mobile robot navigation

Abstract

Dominant plane detection is an essential task for an autonomous navigation of mobile robots equipped with a vision system, as we assume that robots move on the dominant plane. In this paper, based on optical flow, the image motions of points which located in a plane were described by planar flow. After the parameters of planar flow are estimated, the distance error of image points can be used for detecting the plane region from monocular image sequence. The proposed method includes three steps. First, the robust optical flow is estimated between two consecutive images from the image sequence. Then a geometry constraint provided by planar flow is applied to detect the plane regions in optical flow. And finally, some iterative procedures can be used for improving the results. Compared with the feature point based method which sometimes is invalid because of the lacking of texture, the optical flow based method can detect the ground plane even if there are not much more image features in the plane region. Some experiments with outdoor images have been conducted.