A new method for obstacle detection in ITS

Abstract

In a 2D road image, vertical edges of a vehicle still keep vertical and do not converge, therefore obstacle detection can be performed in the 2D image plane. With the help of the detected lane structure, two obstacle detection schemes are given. One uses edge detection and binary rank order filtering (BROF) techniques for the cases of normal luminance, the other utilizes weighted mean filter and region identification to detect regions of vehicle tail lights in the evening or at night. Experimental analyses are. given, and longitudinal distance of a vehicle is obtained.