Research On The Key Techniques Of The Vehicle License Plate Recognition System

Intelligent Transportation Systems ITS (Intelligent Transportation System) use of control technology, communication technology, operations research, sensor technology, computer technology and the effective integration of artificial intelligence, the purpose is to fully utilize the existing road infrastructure and resources to improve the person, road, car interaction, improve system security, efficiency and comfort, and then enhance the overall transport economy.Aiming at limitations of license plate image quality, this paper proposes an improved algorithm for image enhancement. Trough analyzing the traditional image enhancement algorithm limitations based on brightness variations, which are vulnerable to the impact of light intensity, light intensity changes will affect the license plate location accuracy. So this paper processes the edge density as the key factors of image enhancers and applied on plate positioning system.The license plate location is the difficulty for the license plate recognition, integrated the advantages and disadvantages of traditional algorithms, this paper proposes an improved algorithm which base on scan-line. Firstly, the up and down boundaries of the license plate is determined by gray jumping, then the license plate binary image is processed by edge detection, feature images are then projected on the characteristics of the image combined with prior knowledge to determine the license plate car cards around the border region.The license plate character segmentation is an important part of the license plate recognition system, the situation of the license plate character segmentation has great impact on the recognition results. First of all, the license plate image is binarizated by a local threshold, and then change the tilt of the plate by Radon transform, and then carried out the border of the plate and rivet handle, lastly character segmentation is realized by gray level projection.Characters recognition is divided into alphabetic-digit characters recognition and Chinese characters recognition. Alphabetic-digit characters are recognized based on edge of the Hausdorff distance, Chinese characters recognition extracted original features of character by using the wavelet transform firstly,, and reduce the dimension of the original features to obtain the final feature. Finally input the final features to the minimum distance classifier and make template match to get the final result of recognition.