Search

Recently, a series of deep learning methods based on the convolutional neural networks (CNNs) have been introduced for classification of hyperspectral images (HSIs). However, in order to obtain the optimal parameters, a large number of training samples are required in the CNNs to avoid the overfitting problem. In this paper, a novel method is proposed to extend the training set for deep learning based hyperspectral image classification. First, given a small-sample-size training set, the principal component analysis based edge-preserving features (PCA-EPFs) and extended morphological attribute profiles (EMAPs) are used for HSI classification so as to generate classification probability maps. Second, a large number of pseudo training samples are obtained by the designed decision function which depends on the classification probabilities. Finally, a deep feature fusion network (DFFN) is applied to classify HSI with the training set consists of the original small-sample-size training set and the added pseudo training samples. Experiments performed on several hyperspectral data sets demonstrate the state-of-the-art performance of the proposed method in terms of classification accuracies. [ABSTRACT FROM AUTHOR]

In this letter, a novel edge-preserving filtering based approach is proposed for feature extraction of hyperspectral images, which consists of the following steps. First, the dimension of the hyperspectral image is reduced with an averaging based method. Then, the resulting features are obtained by performing edge-preserving filtering on the dimension reduced image, in which a learned edge detection map serves as one of the major cues in the filtering process. The advantage of the proposed method is that it makes full use of the learned edge information in the feature extraction process, and thus, able to improve the performance with respect to other traditional feature extraction methods. Experiments on two real hyperspectral data sets demonstrate the outstanding performance of the proposed method especially when the number of training samples is limited. [ABSTRACT FROM AUTHOR]