An adaptive windowing prediction algorithm for vehicle speed estimation

Abstract

A novel integrated system for vehicle speed estimation is presented. The mean traffic speed is estimated by using a sequence of video images from CCD cameras on the road and performing image analysis techniques pixel distance calculation, and inverse perspective transformation with known parameters. Moving vehicles are detected by applying frame differencing methodology: a combination of morphological operations and image region growing techniques. A new fast search algorithm, called adaptive windowing prediction based on determined vehicle speed in previous frame, is proposed for improving real-lime system computational efficiencies. Experimental results show that the proposed system exhibits good performance at a significantly reduced computation complexity.