Speed-vision: speed measurement by license plate reading and tracking

Abstract

Our proposed approach, Speed Vision by license plate reading is aimed to compute the instantaneous speed of the vehicles using monocular as well as binocular system configurations. By tracking the vehicle in two images it is possible to get a high accurate estimation of the vehicle speed. The features used for matching and tracking are high level features (clusters of matched license plate characters). The processing system is actually connected into a larger distributed architecture based on standard CORBA interfaces to achieve a more efficient interconnection between the peripheral units and the central supervisor. The description of the processing scheme is supported by a simulation model, that is used to identify the most critical parameters of the system. Experimental results in the laboratory environment are discussed, by computing the accuracy in the computation of license plate displacement in different configurations. Finally, a series of experimental results were obtained using a binocular license plate reading system.