A PCNN-based method for vehicle license localization

In the automatic vehicle license plate recognition (LPR) system, the license plate location is the key step which directly affects the following recognition modules. This paper proposes a novel algorithm for vehicle license localization. This algorithm uses morphological operations on a preprocessed, edge image yielded by PCNN. Then connected component analysis follows to get some candidate regions which probably contain license plate. Finally horizontal and vertical projections are used to work on each candidate region for an accurate plate extraction. Experiments have been carried out on images photographed in different scenes, and achieved very encouraging results.