Research And Design Of License Plate Recognition System

With the increase of car ownership and the development of the Internet and electronic technology,License Plate Recognition System(LPRS)has become a very important part in Intelligent Transportation Systems(ITS),and has successfully used in various aspects such as highway video surveillance and traffic violation inspection.Therefore,it's necessary to identify license plates effectively and accurately.The core algorithm of license plate recognition system mainly consists of three parts: the license plate location,character segmentation and character recognition.In this paper,we analyze in depth on the core algorithm of license plate recognition,and develop the license plate recognition system.This paper mainly contains the following aspects.1.Study on license plate location algorithm.For the phenomenon that the license plate recognition is slow,the illumination is uneven and it's difficult to extract the license plate features quickly and efficiently when the license plate is deformed,we adopt the grayscale and color location combination algorithm for a rough location in this paper.Then we obtain the license plate candidate regions by image preprocessing algorithms such as adaptive image banalization,Gaussian blur and erosion and dilation.According to the national standard license plate and the extracted license plate features,the system determines the location of the license plate.After that the post-processing is performed to get the right size of the license plate,such as skew correction and affine transformation.Finally,we enter the license plate images into a support vector machine to verify the reasonableness of the position of the license plate.2.Study on the license plate character segmentation algorithm.We analyze the common license plate character segmentation algorithms and use the grayscale and color location combination algorithm for preprocessing the images containing the license plates.Then we split non-Chinese characters by looking for the circumscribed rectangle.In this paper,special rectangular positioning method is used for Chinese characters segmentation.3.Study on license plate character recognition algorithm.We conclude the common character recognition algorithms,and analyze in depth on the characteristics of BP neural network classifier.At last we choose the neural network for license plate character recognition.4.The implementation of license plate recognition system.Based on Jetson TK1 Embedded Development Kit,we adopt the above related algorithms and implement the license plate recognition system with the open source vision library OpenCV.