Study On Fuzzy Logic-based License Plate Recognize System

Following the development of science and economy, intelligent transport system has become a focus of research attention. The license plate recognition system is the one of its important parts. It has been widely used in road monitoring, vehicle management and highway automatic fee, etc. Fuzzy Reasoning is the algorithm which can imitate the way of human deal with the problem. It can solve the problem directly by using Fuzzy rules and unnecessary to construct the mathmatic model of the problem. Since it is good at dealing with the fuzzy information, it has been used in many fields, for example, fuzzy control, fuzzy neural network and fuzzy decision-making etc.License plate recognition system can be divided into to four parts: image pre-processing, vehicle license plate location, character segmentation and character recognition. By analysing the fuzzy reasoning algorithm and vehicle plate recognition system at home and abroad, we present a Fuzzy Logic-based license plate recognition system. The experimental results show that the method is effective for car license plate location.The major research of this paper can be follow as:1. In the image pre-processing, a method of binary-valuation based on concept of intuition fuzzy entropy is proposed, and we compared it with the Otsu algorithm. Assumed that the standard deviation of the pixels in the license plate within a certain range, based on which a method to simplify the license plate image is proposed.2. By researching the color-based and the texture-based method of the location license plate, respectively. A new algorithm which based on similarity measure template matching is presented. This method assumed the license plate image is consisted by candidate license plates. In this method we construct the template for representing license plates, then we construct the fuzzy sets for representing the template car pictures, respectively. The similarity for each template size location of each photo is estimated by using the method of sliding windows, choose the one with biggest value of similarity as the range of interesting.3. By research the Traditional Back-Propagation(TBP) algorithm, a improve BP(Back-Propagation) neural network has been presented. The learning rate of this new BP neural network is adjusted by fuzzy reasoning method and the Genetic algorithm is used for getting the fuzzy rules. The simulation results indicate that the new algorithm has improved TBP algorithm effectively.4. Using the improve BP neural network for character recognition, and compare the effectively with different feature extraction methods.