Vehicle Plate Recognition System Of Research And Application

Research and application of license plate intelligent recognition system is very important in modern traffic surveillance and control link, it is an important premise of the realization modernization and intelligent traffic management while it is the one of modern traffic engineering field and hotspot. The international advanced license intelligent identification can already based on computer vision technology and artificial intelligence technology, it can automatically assigned lane various news photographers peccancy vehicles, automatically recognize the license plates, compared data from a database, find goal and record its color, speed, vehicle travelling time and alarm model, it reminds enforcement personnel to inspect the suspicious vehicle, it can reducing human labor and identify plate quickly. Combining database support, network support, alarm, remote transmission and on-site command center network dispatching management auxiliary functions, vehicle plate recognition system can realize real-time vehicle license plate, automatic match vehicle information and data filtering unreasonable, on-site alarm, after testing, remote transmission, center management process automation. Based on the process of license plate character recognition, license plate location, the image preprocessing tilt correction, character feature extraction and character recognition process, the work mainly include the following aspects:(1)In the course of pretreatment, digital image processing technology and math morphology technology are used fully. Experiment results show that the method is very good to overcome the illumination and background noise and can improve the license plate location accuracy.(2)In the process of license plate location, summarizing the plate background texture characteristics, this article based on one texture characteristics method and combined the Hough transform make the plate tilt correction method which can achieve good dynamic vehicle image positioning and improve accuracy of the localization licence.(3) In the process of license plate character segmentation, A method based on image of vertical and horizontal projecture will be used, experiment results show that the method can achieve perfect result.(4)In the course of feature extraction, combining with the thick gridding feature which is widely used now. Experiment results show that the method can improve accuracy of feature identification.(5)In the character recognition phase, BP neural network method is be used, this paper designs the number network and the letter network using character's point feature and the center of gravity feature to recognize.