Research on Rotor Winding Detection Based on Texture Information and Convolution Neural Network

Abstract

To achieve the rotor winding qualification test, a new detection method based on texture information and convolution neural network is proposed. This method could effectively eliminate the effect of illumination by using LBP to extract image texture information in advance which is calculated by circular radius, and then using convolution neural network to extract features in further to discriminate the workpiece is qualified or not, which can put the discrimination accuracy to a higher stage. The experimental results prove it, which shows that the new method in this paper performs better than traditional methods no matter of the accuracy rate or the time it costs.