License Plate Image Location Algorithm Study Based On Projection Template Method

Vehicle license plate recognition system is quite an important part of intelligent traffic system, which can automatically provide vehicles inspecting, validation, registering and alarming, so it can be used in highway management systems, parking fees systems, uptown vehicle management systems and electronic police systems. Vehicle license plate recognition system consists of 4 parts, which are license plate image preprocessing, license plate localization, character segmentation and character recognition. And, image preprocessing and license plate location has an important impact on the character segmentation and character recognition.This paper analyzes and summarizes the recent achievement of scholars in and abroad in this field. Furthermore, combined with license plate characters and image location technology, the location algorithm for vehicle license plate is studied deeply and the main research work is summarized as follows:1. Select the appropriate algorithms to preprocess each original vehicle image, which includes color image graying, enhancing image contrast, image binarization and image denoising. Then get the binary image of vehicle.2. Develop a license plate location algorithm which based on projection template method. The main steps of this algorithm are: Considering the difference of horizontal projection and vertical projection on the plate image, apply the processing steps of separate location. First, use vertical projection and gray jump sorting method to get the vertical coordinates. Then use the improved 3σcriteria to correct these coordinates, and remove the mutation errors. Thus, the plate length and the accurate vertical position of the plate image are obtained. After that, set a template to match the image according to the plate length and inherent aspect ratio of plate. Finally, compute out the horizontal position of the plate image. So, the exact location of license plate can be gotten.3. For some plate images with serious distortion, count both half of the average height of the images, and get the distortion rate of the plate images. In this way, the distortion plate images can be well rectified.This algorithm was achieved in Visual C plus plus 6.0 programming environment. Experiments show that when using this algorithm, the license plate information can be effectively extracted in a complex context. With fast computation speed and great robustness, it can satisfy the accuracy and real-time requirements of the vehicle license plate recognition.