Low-rank sparse feature selection for image classification

Abstract

There is a lot of redundancy in the high dimensional raw images, which not only greatly increases the computational burden of image classification process, but also inevitably degrades the classification performance of the model. High-performance dimensionality reduction algorithms are in urgent need of development. To solve this problem, we develop a novel feature selection model for dimension reducing. It greatly reduces redundant features and selects the most representative features for classification. Besides, we also design a novelty version of the lightweight convolutional neural network (newCNN). The newCNN can enhance the classification performance of the system. To improve the classification accuracy, we build a hybrid classification (HC) model with the newCNN and Support Vector Machines (SVM). This model not only solves the problem of overfitting in the training process, but also has excellent generalization ability and robustness. The experimental results verify the effectiveness of our proposed methods.