License Plate Image Location Algorithm Study Based On Projection Template Method

License plate recognition system is a key technology in urban intelligent transportation system. It can automatically intake and identify the license number of the car, using the uniqueness of license plate number to identify vehicles. And bring convenience for intelligent transportation system management. With the development of society, the license plate recognition technique (LPR) is used more widely, and technical requirements are also more and more high. Therefore, LPR is an important research topic in modern intelligent transportation management.On the basis of analyzing and summarizing the domestic and foreign scholar's research achievements, the paper research deeply on the key technology of license plate recognition system. The main research contents are as follows:①License plate localization part: Combining edge detection and mathematical morphology, effectively highlighted rich area of vertical edges (candidate with license plate area), then using the geometrical characteristic and color features of the license plate screen the candidate with license plate area to determine the correct license plate area.②License plate characters segmentation part: To eliminate the influence of license plate characters segmentation by license plate frame, interference border of License plate is removed. Combining with license plate of prior knowledge, improved traditional vertical projection segmentation method is used to identify the biggest interval area between the second and the third license plates characters, then segmenting two characters to left and five characters to right separately for realizing the segmentation of license plate characters.③License plate characters recognition part: This part is the core part of the paper. On the basic analysis of the license plate recognition method commonly used, the paper proposes a kind of license plate character recognition algorithm based on hierarchical. The algorithm recognizes the simple structure character only by the first stage rough classification, and recognizes confusing characters through the second stage fine classification. Each stage has designed on the feature extraction and classifier. On the feature extraction, the algorithm combine statistical feature and structural feature; on classifier design, first stage rough classification adopts Support Vector Machines(SVM), and using SVM toolbox- LibSVM, the second stage fine classification adopts decision table as classifier.Finally, the algorithm is simulated and debugged in Visual C++6.0. Experiments show that license plate character recognition algorithm based on hierarchical which is designed in the paper can effectively improve character recognition rate , and the recognition effects for confusing license plate characters is better.