Key Tech Research Of License Plate Recognition System

As a key technology in improving the intelligent traffic control, and an important part of ITS, License Plate Recognition (LPR) System has a lots of useful applications, such as intelligent management in Parking lots, road traffic surveillance and automatic charge in highway. Domestic and foreign scholars have already had a deep and extensive research in LPR, and acquired some preliminary achievements. Some exploited products have been put into use, but they haven't met the peoples'requirement yet.Three parts make up a typical LPR system: license plate location, license plate character segment, license plate character recognition. This article concentrates on"license plate location"and"license plate character recognition":1. After deeply researching the cascaded license plate location method raised by Wang[10], we found that when there exits complex upright borderline around license plate, it can not be applicable. So we brought an improved license plate candidates extraction method. Keeping the real time character, our method improved the detection ratio of formal cascaded license plate location method.2. Concerning the license plate character recognition, we comprehensively researching three main license plate character recognition methods: template matching, neural network and statistical classifier. According to the character array of Chinese license plate, we designed number classifier, letter classifier and number & letter classifier specifically. After comprehensively comparing three license plate character recognition methods, we found that the SVM based character recognition method can fully meet the needs of real-time and detection rate of the LPR system.