Low relative speed moving vehicle detection using motion vectors and generic line features

Abstract

This paper presents a new approach to the detection of a vehicle with low relative speed to a monocular moving camera, for complementing moving vehicle detection using motion vectors from H.264/AVC encoder. This method makes use of the generic horizontal line features that exist on most vehicles as a clue of localizing moving vehicles. Further filtering and grouping of these detected lines followed by ego motion compensation can effectively detect moving vehicle with low relative speed for application in advanced driver assistance system. Our test results show a high detection rate of over 90%.