Stereo Tracking of Objects with respect to a Ground Plane

Abstract

In this paper, a novel algorithm for tracking 3D cricket ball locations in relation to an estimated ground plane is proposed. This algorithm proposes using Random Sample Consensus to find a ground plane from disparity images, and then estimating 3D ball positions with respect to the ground plane using fast contour tracing and stereo depth values. In that no prior research has used stereo depth values to locate a ground plane and track an object with respect to that plane, the 90.7% successful tracking rate with only 1.95% false positives suggests that the proposed novel algorithm is a useful contribution for enabling automatic setup and ease of use for such tracking.