手持终端采集全景图像中的车牌定位算法

License plate localization based on the handheld terminals is the basis of the license plate number recognition system based on hand-held terminals. During locating, using a handheld terminal, license plate images are vulnerable to the complex background and light conditions, and the factors such as the shooting angle and distance also have great influence on license plate localization. Coping with the complex background interference, this paper proposes a method combining the theory of Markov random field image segmentation, building a foreground- background segmentation model. Vehicles in the foreground area are extracted, thus adapting to complex background; Considering the impact of different photo angle and distance, we use license plate color characteristic information and find a tonal threshold value interval. By using feedback control algorithm, we get all potential license plate areas. By utilizing character feature of the license plates, we can remove the false license plates, achieving the ultimate localization. Our method can better remove the background interference, overcome the influence of the handset camera angle and distance, achieving the high license plate localization accuracy.