Hyperspectral image classification based on convolutional neural network and dimension reduction

One of the most important ways to explore the information in hyperspectral images (HSIs) is accurate classification of targets. Deep learning algorithm has made a great breakthrough in many areas due to its strong ability of data mining. Typical deep learning models such as convolutional neural network (CNN), deep belief network (DBN) and so on, not only combines the advantages of unsupervised and supervised learning but also have a good performance in large data classification. In this paper, a hybrid classification method combined CNN with dimension reduction (DR) operated by principal component analysis (PCA) is proposed, which fully takes the spatial information and the spectral characteristics of HSI into account. Furthermore, in order to solve the problem of sample imbalance, virtual samples are introduced to the experiments. Numerical results show that the proposed DR-CNN method, especially with virtual samples (named as DR-CNN-vs method in this paper) has promising prospect in the field of HSI classification.