Spatial First Hyperspectral Image Classification With Graph Convolution Network

This paper introduces a novel and efficient Graph Convolutional Network (GCN) Spatial Supporting Modification (SSM) method for classifying Hyperspectral Images (HSI), called First (SPA-F). The proposed can utilize spatial information fully based on the assumption that neighboring pixels are more likely to belong same category. uses from two levels consists of following steps. Firstly, graph is constructed all non-background pixels. Secondly, GCN trained tested perform node classification tasks graph. Thirdly, supporting located according their position in HSI. Finally, SSM strategy used fine-tune initial results obtained second step. superiorities verified Indian Pines Salinas-A scene datasets even when using small number training samples.