The Apply Research Of License Plate Recognition Technology

License Plate Recognition System is the most critical and practical part of Intelligent Transportation System. It is an integration of computer vision, image processing and pattern recognition technology. License plate recognition technology generally includes four parts of license plate image processing, license plate location, character segmentation and character recognition. Although the license plate recognition technology has achieved some results, the current license plate recognition system still exist some problem of license plate location in the complex background and similar characters recognizable errors. Therefore, it is necessary to research license plate recognition technology further.In order to solve the license plate location in the complex background, we should make full use of the license plate texture characteristics and geometric features during the license plate location. Firstly, do the color image binarization processing, then use bottom-line scanning method to scan the binary image to achieve the horizontal orientation of the license plate image. Then most of the complex background and pseudo-plate region have already been removed. Finally, do edge detection and vertical orientation of the license plate image in the horizontal orientation according to the characteristics of license plate at the edge mutation. Then do tilt correction of the locate license plate image which has large tilt angle, and remove frame and rivets of the license plate to get accurate license plate image.During character recognition, License plate character recognition is divided into two categories to identify based on the characteristics of the distribution of license plate characters, including one characters classifier and another numbers and letters mixed classifier. Each category are multi-class classification problem in the two classifiers, then we can transformed multi-class classification problem into binary classification problem, and the SVM is good for the two types of classification problems, it not only applies to the classification of the finite samples but also guarantee to find the global optimal solution. Here, we use one-to-one vote mechanism to implement the classification and recognition of license plate characters.The experimental results show that we can locate the license plate image from the complex background by using the above license plate location method. We compare with the character recognition method based on template matching and the character recognition based on SVM method, and the result show that license plate character recognition method based on SVM can solve similar character recognition errors more effectively. The accuracy of license plate location and the improving of the recognition rate of similar characters improve the recognition rate of the entire license plate recognition system.