Hyperspectral images provide fine details of the scene under analysis in terms of spectral information. This is due to the presence of contiguous bands that make possible to distinguish different objects even when they have similar colour and shape. However, neighbouring bands are highly correlated, and, besides, the high dimensionality of hyperspectral images brings a heavy burden on processing and also may cause the Hughes phenomenon. It is therefore advisable to make a band selection pre-processing prior to the classification task. Thus, this paper proposes a new supervised filter-based approach for band selection based on neural networks. For each class of the data set, a binary single-layer neural network classifier performs a classification between that class and the remainder of the data. After that, the bands related to the biggest and smallest weights are selected, so, the band selection process is class-oriented. This process iterates until the previously defined number of bands is achieved. A comparison with three state-of-the-art band selection approaches shows that the proposed method yields the best results in 43.33% of the cases even with greatly reduced training data size, whereas the competitors have achieved between 13.33% and 23.33% on the Botswana, KSC and Indian Pines datasets.

This paper addresses the band selection of a hyperspectral image. Considering a binary classification, we devise a method to choose the more discriminating bands for the separation of the two classes involved, by using a simple algorithm: single-layer neural network. After that, the most discriminative bands are selected, and the resulting reduced data set is used in a more powerful classifier, namely, stacked denoising autoencoder. Besides its simplicity, the advantage of this method is that the selection of features is made by an algorithm similar to the classifier to be used, and not focused only on the separability measures of the data set. Results indicate the decrease of overfitting for the reduced data set, when compared to the full data architecture.

No abstract

Hyperspectral image classification by means of Deep Learning techniques is now a widespread practice. Its success comes from the abstract features learned by the deep architecture that are ultimately well separated in the feature space. The great amount of parameters to be learned requires the training data set to be very large, otherwise the risk of overfitting appears. Alternatively, one can resort to features selection in order to decrease the architecture's number of parameters to be learnt. For that purpose, this work proposes a simple feature selection method, based on single-layer neural networks, which select the most distinguishing features for each class. Then, the data will be classified by a deep neural network. The accuracy results for the testing data are higher for the lower dimensional data set when compared to the full data set, indicating less overfitting for the reduced data. Besides, a metric based on scatter matrices shows that the classes are better separated in the reduced feature space.

Hyperspectral images provide rich  spectral details of the observed scene by exploiting contiguous bands.But, the processing of such images becomes heavy, due to the high dimensionality.Thus, band selection is a practice that has been adopted before any further processing takes place.Therefore, in this paper, a new unsupervised method for band selection based on clustering and neural network is proposed. A comparison with six other band selection frameworks shows the strength of the proposed method.