Convolutional Neural Network Algorithm based on Improved Support Vector Machine

Abstract

This electronic Aiming at the problem that the traditional Convolutional Neural Network algorithm uses the Softmax function to classify tires with poor accuracy, a convolutional neural network image classification and recognition algorithm based on improved support vector machine is proposed. This algorithm uses the support vector machine instead of the Softmax function to complete the image classification problem in the convolutional neural network algorithm. At the same time, a relaxation variable and a penalty factor are introduced into the traditional support vector machine, thereby changing the support vector machine to be used for classifier for multiple classification problems. This improved algorithm is applied to tire damage image recognition. Through a series of comparison experiments between the improved algorithm and the traditional convolutional neural network algorithm, the classification and recognition effects of the improved algorithm on the classification and recognition of damaged tire images are analyzed. The effectiveness of the proposed improved algorithm is proved.