Research On License Plate Recognication In Complex Scenes

With the rapid socio-economic development, people’s living standards improve, the number of motor vehicles is also increasing. To improve the management efficiency of the vehicle, to ease traffic pressure, to use efficiently parking resources, maintain road safety, must find an efficient motor vehicle management solution. Motor vehicle license plate is "identity card" of the vehicle, is an effective vehicle marked identity. Automobile license plate recognition technology mainly studies on license plate number, with the rise of intelligent transportation, achieved fruitful results after years of development. This paper presents an effective method of license plate recognition on complex scenarios based on previous studies.The main content of this paper includes the following three aspects:(1) The paper proposes a method of license plate location in complex scenes. This method makes use of color and texture information to realize location. Use image enhancement, edge detection and connected component analysis to extract image texture features, and analyze the texture information to obtain license plate candidate regions; finally a multi-stage filtration with many conditions is designed to complete the license plate precise positioning.(2) As to character segmentation, a three-division feedback structure of character segmentation is presented. Template matching method, projection and dynamic threshold segmentation method are used comprehensively in this structure. Adjust segmentation results via character recognition as feedback. Experiments show that the proposed character method is effective for a variety of license plate image quality, and has good robustness.(3) For character recognition problem, sum up the current mainstream feature extraction and classification. Experiments compare neural networks and the SVM classification algorithm for license plate character recognition results. At last, realize the license plate character recognition using SVM and sift feature operator.The proposed method of license plate recognition method for complex scenarios has good adaptability, and acquires a high recognition rate in the situation of the image quality deterioration because of being affected by light, weather, wear and other factors.