The Research Of The Chinese Character Recognition Of License Plate Based On Discrete Hopfield Neural Network

With the rapid development of the economy and the transportation industry,higher requirements are needed for intelligent transportation systems. Automaticlicense plate recognition system is the core part of the intelligent transportationsystems and the character recognition is the core part of the automatic license platerecognition system, and the Chinese character recognition is the most important anddifficult part.A number of methods are proposed by domestic researchers to identify thelicense plate characters in recent years. Better recognition effect and certainrecognition rate can be achieved for those character images with less noise andinterference. But we don’t have a satisfactory result for characters with more noiseand the case of blur, deformation and inclination, especially the Chinese characters.So there is theoretical significance and application value to carry out the research onrecognition methods for Chinese characters in license plates and improve therecognition rate.This article has made further research of the Chinese character recognition oflicense plate with some methods aiming at the image pre-processing and discreteHopfield neural network’s associative memory function. First, the basic theories ofHopfield neural network are introduced and a license plate character recognitionnetwork is designed based on the theories. Then its associative memory function ispreliminary tested. A certain effect has been achieved for characters with moreinterference using one image filtering method based on connected area introduced inthis paper. Finally, the proposed method to remove rotation based on the statisticalcharacteristics for the character images which have certain deformation and rotation,has improved the recognition rate when a character recognized by the trained network.The template matching and Hopfield neural network are both used to recognize onepre-processed pattern. The results show that Hopfield neural network only hasadvantage of its associative memory function to crippled modes.