Study Of Vehicles Recognition Based On Image Process And BP Neural Network

The vehicle license automatic recognition system is developed in recent years based on the image and the character recognition technology intellectualization traffic control system, and it is a hot spot in patterns recognition applied research domain both at home and abroad at present, becoming an important research topic in the information processing technology. This system is mainly made up of vehicle license picture gathering and the pretreatment, the vehicle license region localization and withdraws, the vehicle license character division and the recognition .The article has made a quite comprehensive elaboration about the vehicle license localization and the new algorithm of character recognition, the new technology as well as the system overall design. In view of existing research situation at home and abroad, I have made thorough research about two essential technologies: the vehicles license region localization and the license character recognition.This paper has studied a lot of method and theory about vehicle plate location, character divide and character recognition systemically. In the aspect of vehicle plate location, author adopt the algorithm of the frequency of gray scale alternate to seek approximately location. Use the characteristic of rich vehicle plate regional veins to find out every possibility vehicle plate area. Then for these candidate regional gray scales that make vertical direction projection. According to this, differentiate really fake vehicle plate. Eventually, get only vehicle plate area. After correct position approximately, according to HSI color space theory, get accurate vehicle plate area.About character divide, author uses 2 value image vertical projection methods with known plate knowledge to divide character apart. So can eliminate vehicle plate around two end positional accuracies the influence for that character divides into shelf. The aspect of character recognition, this paper has used Chinese character, English character and digital & English 3 neural networks to carry out character recognition. Experiment result shows that this recognition algorithm has gotten better recognition effect.