Research And Implementation Of SVM License Plate Recognition Algorithm Based On FPGA

As a comprehensive system, intelligent transportation system(ITS) has attracted much attention, which improves the transportation efficiency, alleviates the traffic congestion and reduce the traffic pollution, and the vehicle license plate recognition(VLPR) technology is the core of it.This paper has studied and compared some commonly used algorithm of each module in the license plate recognition technology, then selected some algorithms which meets the characteristics of FPGA to achieve, and some of the algorithms has be improved. In the module of vehicle license plate location, the mature algorithm is used, which combines edge detection and morphology. It has a stronger ability to resist interference compared with the traditional. In the character segmentation module, a new algorithm is proposed, which is base on projection segmentation, and then segmentation Chinese character by compare with other characters width, it effectively reduces the probability of the Chinese character error segmentation. The character recognition module is based on the theory of support vector machine(SVM) and prior knowledge, then construct three kinds of classifiers to classified different types of character. Finally, achieved the modules of vehicle license plate location, character segmentation and multiplexed binary function in the form of the FPGA implementation in the Quartus II software platform, and embedded Nios II processor with the function of the character recognition, which is achieve by SOPC technology, all of them constitute a complete vehicle license plate recognition system.