An Unsupervised Band Selection Based on Band Similarity for Hyperspectral Image Target Detection

Abstract

In remote sensing data processing, band selection is very important for hyperspectral image processing and analysis, which utilize the most distinctive and informative band subset of original bands to reduce data dimensionality. Although band selection can significantly alleviate the computational burden, the process itself may cause additional computation complexity. In this paper, an unsupervised band selection method based on band similarity is proposed for hyperspectral image target detection. Several selected pixels are used for unsupervised band selection instead of using all the pixels to reduce computational complexity. The number of bands to be selected is determined by adjusting the threshold of similarity metric, to ensure target detection operator have the best performance with selected bands. The experimental results show that our method can yield a better result in target detection.