Research On Key Techniques Of License Plate Recognition System

The automatic recognition of vehicle license plate (VLPR) is an important step in Intelligent Transportation system. Its key technology involves vehicle license plate location, characters segmentation and characters recognition. A concentrated research has been made on this subject, and the solutions fall into three categories.Firstly, vehicle plate location technique, as the most critical part of VLPR, has been studied. A synthetic approach under a complex background for vehicle license plate location based on segmentation and texture analysis has been proposed. The proposed approach combined the color feature, texture feature and geometric feature of vehicle license plate, which employed HSV color model, fixed matching of background color and character color of vehicle license plate to segment the car image with an effective algorithm. Aimed at the segmented image, line scanning has been used to determine the upper and lower boundary of vehicle license plate on the basis of the characteristic of consecutive grey beeps in horizontal direction. Then differentiate sciagraphy has been adopted to determine the left and right boundary of vehicle license plate, based on projection characteristic in vertical direction. Finally, true vehicle license plate has been accurately located by calculating the HVR. This algorithm employed synthetic characteristic to locate vehicle license plate, which overcomed defects and limitations of single characteristic locating method, and was slightly confined to the size or position of the license plate and illumination condition of the car image. Experiment results showed that the location approach was effective, high-precision and adaptable under a complex background.Secondly, character segmentation means to segment the license image into single character for recognition. Rather than the traditional approach, projection and connectivity-based character segmentation algorithm has been adopted, which involves three steps: plate image binary, top and bottom boundary determination by projection-based algorithm and characters segmentation by connectivity-based algorithm. Some elementary results have been given.Thirdly, Chinese characters, letters and numbers in the license plate are considered as the character recognition process of VLPR system. Template-matching method has been used for recognition of letters and digits in license plate and BP neural network-based method has been also adopted to recognize Chinese character, letters and digits in license plate. At last, its validity has been proved by experiments.