License Plate Recognition Based on Genetic Algorithm

Abstract

A license plate location method based on projection and genetic algorithm was proposed in this paper. Projection located the left and right edges of license plate. Genetic algorithm searched license plate region with second order moment as objective function which described texture in license plate region quantitatively. Four sub-nets were designed to classify Chinese characters, alphabet characters, alphabet & numeral characters and numeral characters toughly. Small nets were designed to recognize several characters with similar structure precisely. Experiments indicates that locating accuracy of algorithm proposed in this paper is 97%. The algorithm can anti-interference strongly and has a well perform on poor quality image. BP classifier designed in this paper has well stability, high fault tolerant capability and its accuracy is 96.5%.