Research Into Static License Plate Image Recognition System

Being applied in areas like toll payment and traffic management, the license plate recognition technology has the functions of reducing manpower, increasing efficiency and improving management system, etc. As the number of cars in our country continues to grow, the technology will show us its great value, both economically and realistically.Some license plate recognition products have been put into service in our country. They can get high recognition rate only under an environment of enough light, and the license plate also must be clean enough. This dissertation deals mainly with the illegibility of license plates that arises from bad weather, insufficient light and awful road condition, and tries to make the algorithm more adaptable to the environment. Software for the recognition of static license plate image has also been worked out. Examined by the indistinct images collected at different time and places, and under different weather, the algorithm is proved to be considerably adaptable to different conditions, and can be applied in real situations.The three steps included in the algorithm are license plate segmentation, which consists of rough segmentation and accurate segmentation, character segmentation and character recognition. The rough segmentation transforms the original image into a texture image, and locates the plate-like areas through the intensity of textures in these areas, and thus finds out the only true plate area according to their gray vertical projection. As a necessary supplement to rough segmentation, accurate segmentation improves its accuracy. A closed-loop feedback controller is designed to adjust parameters of the texture analysis algorithm. In the chapter of character segmentation, the traditional method of finding minimum vertical projections combined with platelicense format is used. The use of space mark matching method proposed in this dissertation to ensure the first segmentation position obtains a very high rate of success. For the reason that it's difficult to identify indistinct license plate Chinese characters through a conventional measure to generic print character, the method of fractal dimension is proposed in the character recognition chapter, and is used as the feature values of Chinese characters by which to fulfill the target through the training of B-P neural network. For letter and figure recognition, the conventional method that is a combination of topologic analysis and template matching is used and has been improved for the particularity of different license plates.