A Novel Convolutional Neural Network for Image Classification

Abstract

Image classification technology is the most basic task in computer vision, and it is widely used in medical treatment, online shopping, face recognition, etc. An improved convolutional neural network model is proposed for image classfication in this article. Firstly, the novelty of this model is that principal component analysis is used to replace the pooling layer in convolutional neural networks, which mainly calculates and selects the eigenvectors with large eigenvalues to make linear table of the original data to achieve the purpose of rapid dimensionality reduction. Secondly, a three-layer convolutional layer is used to extract the main features of the image. Through classification experiments on the MNIST data set, kaggle leaves and other data sets, the desired accuracy is achieved. It indicates that the improved convolutional neural network model has a perfect recognition effect, and the improved model is suitable for classification processing of different types of data sets. In addition, the classification of noise images can also achieve higher accuracy.