Vision based measurements of traffic parameters

Basic concepts of image processing are reviewed first with relation to traffic related image sequences. Next, two real-time applications that use vision based sensors for the extraction of traffic related measurements are described. The first corresponds to the detection of incidents in the traffic of road vehicles and the second attempts to measure some characteristics of crowds in confined spaces, such as railway station concourses. Both applications deal with a different type of input images but in both cases spatial relationships between features are sought. The paper concludes that real-time cost-effective image processing algorithms can be found if the implicit complexity of the processing of sequences of images is reduced by taking advantage of any natural constraint to discard or simplify parts of the input data while keeping and using those image characteristics that provide the maximum amount of required information per data unit