Color image analysis for vehicle speed measurement

Abstract

Image analysis for traffic data collection has been studied throughout the world for more than a decade. A survey of existing systems shows that research was focused mainly on the monochrome image analysis and that the field of color image analysis was rarely studied. With the application of color image analysis in mind, this paper proposes a new algorithm for vehicle speed measurement in daytime. The new algorithm consists of four steps: (i) image input, (ii) pixel analysis, (iii) single image analysis, and (iv) image sequence analysis. It has three significant advantages. First, the algorithm can distinguish the shadows caused by moving vehicles outside the detection area from the actual vehicles passing through the area, which is a difficult problem for the monochrome image analysis technique to handle. Second, the algorithm significantly reduces the image data to be processed; thus only a personal computer is required without the addition of any special hardware. The third advantage is the flexible placement of detection spots at any position in the camera's field of view. The accuracy of the algorithm is also discussed. Key words: speed measurement, vehicle detection, image analysis, image processing, traffic control, traffic measurement and road traffic.