Research And Application Of License Plate Recognition In Electronic Police System

With the development of intelligent traffic management, research and the construction of electronic police system is paid more and more attention by our country. License Plate Recognition system as the central component of electronic police system, makes a direct impact on the product quality and work efficiency of it.The relevant algorithm of License Plate Location has been discussed; License Plate Character segmentation and License Plate Character Recognition and how to carry them out. In the process of License Plate Location, line scan is used to determine the up and bottom boundary of the license plate. In the process of License Plate Character segmentation, the left and right boundary is determined according to the vertical projection of the line scan result. Thereby the region of license plate is determined. In the process of character segmentation, obtain the vertical projection of License Plate, and then according to the vertical projection and the size of license plate characters, the characters can be picked out separately.On the aspect of Character Recognition, the principle of Artificial Neural Network and BP Neural Network has been discussed. As the user-defined license plate is allowed to be used in some city of our country, the number array of license plate no longer have certain order. 3 subnets (Chinese character subnet, letter subnet, number & letter subnet) have been designed to be used on Character Recognition. Use the methods that pick out 1 pixel from every 3 pixels of each row and column to finish feather extraction.Matlab is used to carry out the algorithm which has been discussed. More than 200 images captured in different environment are used to train and test the BP net. The experiment shows that the whole system recognition rate has reached 94.5%,and has good identification speed, the high success rate of identification.