The Design And Implementation Of Vehicle License Plate Recognition System For Toll Station

Vehicle license plate recognition (LPR) system, as an important component of intelligent transportation system, plays a crucial role in traffic management automation and intelligent, and can be widely used in charging management, urban transport vehicles, traffic flow statistics, electronic police and so on. Therefore, resesarch on vehicle license plate recognition system has great application value. This paper combines environment of the station, aims at applying LPR technology to toll station, and completes the design and realization of vehicle license plate recognition system for toll station.Vehicle license plate recognition system mainly contains three modules: license plate location, character segmentation and character recognition. In the location module, combing with edge and texture features of license plate, in the gray-scale images, a fast location method based on mathematical morphology is proposed. This method makes full use of simple background of the image acquisition site (ie, toll station). Reducing the processing area after edge detection by projection, the method consumes around 100ms.In the character segmentation module, projection method is used. First, Pretreatment is carried out on the license plate. That is plate binarization and plate frame removal, etc.. Then for each character, use the smallest rectangular box to carry out the coarse segmentation. Reference to the average width, height, spacing of character character and consistency of character position, fine segmentation of characters is finished, including fragmented characters merging and connected character splitting. For tilted characters only in horizontal direction, even without tilt correction, this method can achieve better segmentation. In addition, in the binarization process, traditional Bernsen method is improved and experiments show better results.In the character recognition module, a recognition approach based on support vector machine is applied. According to different position of license plate characters, different classifiers are constructed. After characters normalization and feature extraction, SVM classifiers are applied to recognition experiment. Then experiments are analyzed from two aspects: parameters of SVM and feature extraction methods.