License Plate Recognition Technology Research

As an important part of the intelligent transportation systems, license plate recognition (LPR) technology becomes more and more important role in the life of people. License plate localization, Character segmention and character recognition consists of the three main parts of license plate identification, and they are the focus of license plate identification technology, and I would introduce the three aspects of LPR technology in this paper.In license plate localization part, a image contains a vehicle and the background, so in this paper I puts forward the methd of extracting the vehicle from a image, for the nature of the gray jumps of a image in vertical direction, with the use of image vertical edge detection and image mathematical morphology method, we could extract the license plate candidate areas; Because a license plate in a image is darkly, we enhance the potential license plate image before the follow-up treatment, and then we use morphological operation to locate the license plate based on the geometric characteristics of a license plate. Experiments show that the algorithm is simple, efficient, and location rate is97%. Before character segmention, in view of the located license plate may contain redundant parts such as borders, I introduced the license plate boundary resection based on projection method, and this operation lays a good foundation for follow-up character segmention.In character segmention part, we divided adhesive characters based on characters envelop and vertical projection, and then using a traditional vertical projection algorithm combination with template matching method to segment characters. Experiments show the method is effective, and also has good adaptability for the license with redundant boundariesIn the license plate identification, I used the template matching method and BP neural network algorithm, I have used feature point template matching method for the similar characters to improve the template matching method, and for the situation of slow convergence, long learning time of BP network,I have made the corresponding improvement too.In this paper the experimental environment is the matlab platform, and I have simulated the corresponding license plate location, character segmentation and character recognition.