Vehicle detection in open-world scenes using a Hough transform technique

Demonstrates a technique incorporating a Hough transform algorithm to detect arc/circle line features in cluttered open-world scenes, as for example may be presented by vehicle wheel and wheel-arches. Detection of the wheels and wheel arches is achieved by searching for concentric arcs/circles and therefore is intended to detect primarily side views of vehicles, although the method will accept some change in orientation as the wheels present ellipses of limited eccentricity. The technique is proposed as a cueing mechanism for guiding more detailed and robust object identification algorithms onto a subset of the complete image. An integrated set of algorithms has been developed taking as input a grey-level image and outputting a cued region containing a putative vehicle defined from an extracted wheel-base