Image classification method rationally utilizing spatial information of the image

In order to improve the accuracy of image classification problem, this paper proposes a new classification method based on image feature extraction and neural network. The method consists of two stages: image feature extraction and neural network classification. At the stage of feature extraction, spatial pyramid matching (SPM) feature, local position feature and global contour feature are extracted. The utilization of spatial information in SPM feature is effectively improved by combining the above three features. At the stage of neural network classification, a multi-hidden layer feedforward Histogram Intersection Kernel Weighted Learning Network (HWLN) is proposed to take advantage of three features to improve the classification accuracy. In the structure of network, the hidden layer output features are used as the bias of the input features to weight the coefficients of the features. The input layer and the hidden layers are directly connected with the output layer to realize the combination of linear mapping and nonlinear mapping. And the Histogram Intersection Kernel is used instead of random initialization of the input weight matrix. Taking combined features as input information, HWLN can realize the mapping relationship between input information and target category, so as to complete the image classification task. Extensive experiments are performed on Caltech 101, Caltech 256 and MSRC databases respectively. The experimental results show that the proposed method further utilizes the spatial information, and thus improves the accuracy of image classification.