Novel boundary determination algorithm for lane detection

Abstract

A novel boundary determination algorithm (BDA) for robust lane detection is presented. The algorithm can track a complete lane boundary based on partial lane edge points provided there exist sufficient orientation similarity there between. BDA mainly comprises two parts: (a) utilizing an edge detection method to extract orientations and coordinates of edge points; (b) repeatedly applying two circular masks, i.e., a probing circular mask followed by a tracing circular mask, to collect two separate sets of orientations of edge points and from which to sift out a primary orientation for tracking a curve as perceived by human eyes. The algorithm is characterized in that the probing circular mask acts to probe ahead for the tracing circular mask in order to avoid deviation from the right track.