The Design And Implementation Of The Car License Plate Recognition System

With the acceleration of the process of China’s urbanization, the number of motorvehicles in cities is growing rapidly, which presented new challenges to the traditionalmethod of traffic management. At present, the issue of traffic management hasbecome an important issue of urban management. Modern intelligent trafficmanagement system is a new research topic which is proposed to deal with urban roadtraffic management. It can achieve the intelligent traffic management and highefficiency. And the automatic license plate recognition system is an important partof the intelligent traffic management system, which has the very broad prospect ofapplication. Since license plate number is the most effective motor vehicleidentification information to distinguish vehicle, automatic license plate recognitionsystem can automatically identify license plate number from the image of vehicle,which provide the necessary basis for the follow-up vehicle intelligent management.Automatic license plate recognition system can obtain useful information quicklyfrom a large number of image information on the license plates, which avoid thetedious manual search, collecting process.Automatic license plate recognition technology is a frontier research subject basedon image processing, pattern recognition, and automation technology. Typicalautomatic license plate recognition system comprises image pre-processing, licenseplate location, character segmentation and character recognition, etc. This articlestudies the automatic license plate recognition technology from these four parts, andfinally determines a series of effective algorithms for vehicle license plate recognitionquickly and effectively. The main research content and results are described below:(1) In the image preprocessing section, because of the presence of noise and otherinterference information for vehicle image which input system, this paper first doimage grayscale and image binarization for the vehicle image in order to improve thefollow-up recognition accuracy. The image binarization uses global binarizationmethod, and uses the Otsu algorithm to get global threshold. Based on this threshold,we can get the binarization vehicle image. The plate region which obtained by the global binarization method are clear, and the character boundaries are definite. Thismethod avoids the character strokes fracture phenomena by using local binarization.For the image which is inadequate of brightness and contrast, we use gray stretchingto enhance the brightness and contrast of the image. Image noise removal uses medianfilter algorithm, which removes the noise effectively, and preserves the edges of theimage information maximally.(2) In the license plate location section, we analysis and compare several platelocation methods, such as the method based on edge detection and Hough transform,method based on artificial neural network, method based on mathematicalmorphology. And then a license plate location method is proposed which is based onthe license plate prior knowledge and line scan. This method locates the license platecharacter area by using a sliding window to detect the pixel hopping points of licenseplate character area. This algorithm which can locate the license plate boundaryaccurately has a low computation, and it meets the requirements of real-time andaccuracy of the system for license plate location.(3) In character segmentation part, we use Hough transform to correct the inclinationof the license plate. According to the characteristics of the license plate character, wepropose a character segmentation algorithm based on the license plate priorknowledge and the projection method. By using this algorithm, we can get thecharacter boundaries accurately and achieve good segmentation results.(4) In the character recognition part, we first analysis and compare several commonlicense plate character recognition methods, and then propose a character recognitionmethod based on the combination of template matching-feature points matching. Theidentification system has two subsystems. When a recognition procedure begins,license character image first get into the No.1subsystem. If there are a lot of characterfeature differences between two characters, then the No.1subsystem which using thetemplate matching method can complete the identification. For confusing characters,we must put them into No.2subsystem for further identification, which using the13feature points matching method. Experimental results show that the system designed by proposed algorithms hasachieved the desired results in license plate location, character segmentation,character recognition and other parts. So this system has certain guiding significancefor practical application.