Abstract :
The potentially severe consequences of incidents and congestion in road tunnels has resulted in a high level of traffic surveillance being specified. Closed circuit television (CCTV) systems usually form an important part of the surveillance system, with operators viewing the images either on an ad hoc basis, or in response to an alarm raised by some other means. The typical picture from a camera in a tunnel has both advantages and disadvantages when trying to computerise the analysis of the images. On the positive side, the lighting conditions are generally fixed, except near the portals and the camera view is usually fixed as well. Against this, the camera position is low and oblique resulting in a large amount of occlusion, that is the masking of large sections of the image by vehicles in the foreground, and the camera can be subject to a rapid build up of grime. The author describes the use of a novel technique called image processing for automatic computer traffic surveillance (IMPACTS), which is tolerant of the difficult camera views, and can provide the sort of surveillance information that is needed as part of a tunnel management system. Some results are given for a scene from a camera in the Blackwall Tunnel which indicate how the system performs