Hyperspectral Image Classification Using Deep Pixel-Pair Features

TLDR

Deep convolutional neural networks have attracted significant interest for hyperspectral image classification, achieving excellent performance when sufficient training samples are available. This study proposes a novel pixel‑pair method that effectively increases the number of training samples, thereby enhancing the discriminative power of deep CNNs. For each test pixel, pixel‑pairs are formed by combining the center pixel with each surrounding pixel, classified by the trained CNN, and the final label is assigned through a voting strategy. Experiments on several hyperspectral datasets show that the proposed method outperforms conventional deep learning approaches in classification accuracy.

Abstract

The deep convolutional neural network (CNN) is of great interest recently. It can provide excellent performance in hyperspectral image classification when the number of training samples is sufficiently large. In this paper, a novel pixel-pair method is proposed to significantly increase such a number, ensuring that the advantage of CNN can be actually offered. For a testing pixel, pixel-pairs, constructed by combining the center pixel and each of the surrounding pixels, are classified by the trained CNN, and the final label is then determined by a voting strategy. The proposed method utilizing deep CNN to learn pixel-pair features is expected to have more discriminative power. Experimental results based on several hyperspectral image data sets demonstrate that the proposed method can achieve better classification performance than the conventional deep learning-based method.