Research of License Plate Recognition Based on Rough Sets and Fuzzy SVM
|School||Jiangsu University of Science and Technology|
|Course||Pattern Recognition and Intelligent Systems|
|Keywords||License Plate Recognition Rough Set SVM Fuzzy SVM|
With the rapid increase in the number of city car, license plate recognition technology has become the intelligent transportation research hotspot. Car license plate recognition technology is a core part of intelligent transportation, including the license plate image preprocessing, license plate location, tilt correction, character segmentation and character recognition, and several other parts. This paper analyzes the license plate recognition technology at home and abroad in recent years, based on the latest developments on the license plate recognition system, the key techniques are studied. Support vector machine (SVM) is a 1990s by Vapnik et al A new type of machine learning methods, it can successfully deal with classification and regression problems. As excellent support vector machine learning performance, the technology has become the hotspot of machine learning, and in many areas has been successfully applied. However, as a technology is not yet mature, support vector machines currently there are still many limitations, such as the existence of non-subregional, training time is too long. Fuzzy Mathematics is the study of the problem of many boundaries are not clearly a mathematical tool, the use of fuzzy math and fuzzy logic, can well handle a variety of ambiguity. Rough sets can simplify the training sample set, while retaining the premise of important information to eliminate redundant data and improve the speed of classification. This article will rough sets and fuzzy SVM combination is given based on rough sets and fuzzy SVM license plate character recognition algorithms. In license plate positioning stage, using a support vector machine based license plate location algorithm can accurately locate the license plates. In the license plate character segmentation stage, taking into account the presence of adhesions license plate character case, gives an improved vertical projection segmentation method to improve the character segmentation success rate. In the license plate character recognition stage, the use of rough set of characters to extract features about Jane, effectively reducing the training and testing time, and the traditional SVM multi-classification method has not be problems in the subregion, this paper introduces the theory of fuzzy mathematics effectively solved the multi-classification. Finally, MATLAB R2009a and LIBSVM toolbox presented in this paper based on rough sets and fuzzy SVM license plate character recognition algorithms were validated experimental results show that the algorithm is given in this paper can solve the SVM can not exist in the subregion problems and improve the recognition efficiency, experimental results verify the effectiveness of the algorithm.