TRAFFIC QUEUE DETECTION USING VIDEO IMAGE-PROCESSING
Arianne Jennifer Rourke, Mgh Bell · 1991
The automatic detection and analysis of queuing and congested traffic as well as the detection of obstacles at critical locations, such as railway level-crossings, has proven to be difficult using current point-based sensor technology, such as inductive loops. In this paper, an image processing system is described which provides accurate detection and analysis of these types of traffic situations. Using the Fast Fourier Transform (FFT) method, the presence of vehicles in a video image are detected. Further processing determines whether these vehicles are moving or stationary, and their positions along the length of road under surveillance is recorded. If a calibrated scene is employed, these measurements can be related to actual ground locations within a road network. Information on queuing, such as the existence of a queue, queue lifetimes, queue length, queue growth and decay and, very importantly, the location of the rear of the queue can all be obtained. The area over which the road can be monitored is flexible, and can be controlled by the focussing of the camera and/or the size of the detection window defined in software. It is therefore possible to tailor the system to suit a particular application, whether that be to monitor a small area, such as a level-crossing, or a large area such as a section of motorway. Comparative testing of the automatic response of the system with manual analysis of queuing traffic scenes has shown it to be a reliable method of queue detection. (A) For the covering abstract see IRRD 851414.