Traffic Character Recognition Theory And Application Based On CS

With the ever-increasing number of cars, unban traffic is increacingly attentioned by peo-ple. How to carry out traffic management effectively more and more become the focus. Therefore, the ITS has made rapid development, license plate recognition system as an important branch of ITS has also been fast development, and highways, urban roads and car parks occupy an irreplaceable project management position.This paper summarizes the consolidation and license plate recognition in recent years, some of the methods and the latest progress, and will a new theory-the theory of compressed sensing applied to character recognition on. License plate recognition system can be basically divided into preprocessing, license plate localization, character segmentation, character recognition five modules, this paper studies a simple static background image under the license plate recognition technology, on the license plate location, character segmentation and character recognition, and improvements and some new methods. the main task is to:First of all, the car image pretreatment to prepare for the follow-up treatment, including the color gray, gray stretch so. Then according to the characteristics of plates, character-based texture, edge detection and morphology and vector quantization of the plate positioning, experiments show that the application meets the requirements. And then after the license plate on the positioning was correct, binary and other treatment, then the method of direct projection character segmentation, due to projection may lead to adhesion of characters and a character to be separate phenomena, the paper then uses a template segmentation methods, results were better. In the character recognition part, the article first extract the characteristics of the characters, and then used on the BP neural network character recognition methods were introduced and analyzed. Finally, a new theory, both the compressed sensing theory were applied to character recognition, that this is an efficient way possible.This research can more accurately locate license plate, segmentation, and recognition on the tilt of the plate, deformation, etc. have strong anti-interference ability, good research and development value.