Ground Plane Estimation for Robot Navigation Using Homography

For mobile robot navigation algorithms such as obstacle avoidance, path planning, target following, landmark detection, distinguishing the ground surface from the elevated obstacle plane is very crucial from a pair of stereo images. This paper presents ground plane estimation approach based on the concept multi-view relations between the two images of the same scene. The relationship is created over the features extracted from the pair of images using the SIFT (Scale Invariant Feature Transform) algorithm. Planar homography technique is used which assumes that the textured dominant ground plane lies in the lower part of the image. Hence by using pair of stereo images specific features are classified on the basis of ground plane homography matrix as either belonging to ground plane or not.