Sample Inherent Correlation Mining for Hyperspectral Image Classification (opens in new tab)

Graph convolutional networks (GCNs) have recently shown great promise in hyperspectral image classification (HIC) by effectively modeling long-range correlations among samples. However, existing methods often rely on sample similarity for graph construction while neglecting intra-class variability, thereby limiting the robustness of the extracted features. This article introduces a novel HIC method that leverages sample inherent correlations to enhance feature representation and improve class...