Genetic Neural Network and Its Application in Recognition of Tire Code

Genetic arithmetic can search whole best key of question. In this paper, it is applied to choose network initial weight value, then make use of back propagation arithmetic to search optimization weights, and design a multiplayer neural network based on genetic arithmetic. This paper introduces the structure of genetic neural network, imposes the entropy error function to speed up the learning rate, and introduces into learning rate and momentum gene to adjust network weights self-adaptively so as to avoid neural network training surge, and the inferring course is also brought forth. Combine tire code character recognition, distill the eigenvector of character image which is segmented to be recognized, and train the genetic neural network and exhibit recognition result. Practice proves that the method can improve the constringency and the recognition nicety of network.