Video image processing for traffic analysis

Abstract

In recent years the application of computer-based image processing techniques to a range of traffic data collection tasks has been successfully demonstrated. In a similar field of research carried out by the author at the University of Wales College of Cardiff, a system based on commercial image processing hardware, a 80486 IBM PC-AT and a video recorder was assembled. The main aim was to develop a system for automatic vehicle data measurement and to extend its application to the collection and analysis of pedestrian data. This paper will focus on the development of the system for vehicle detection and measurement. A direct segmentation technique on the video images was adopted as a standard method of vehicle identification. The identification of the presence of an individual vehicle based on brightness information at relatively few sample points within the images was possible. Double threshold values were applied to the area of interest for the conversion of the area into a binary form. To compensate for the ambient lighting changes, a method of updating threshold values sequentially was introduced. The suitability of the approach and detection algorithm was assessed by analysing a video tape containing a traffic scene for the measurement of vechicle movement. This tape is typical of tapes from which data have been extracted manually using event recorders. Although the performance of the implementation algorithms needs to be further assesed, the preliminary results have demonstrated the success of collecting data for vehicle counts, speeds and headways with reasonable accuracy.