Semisupervised Classification of Hyperspectral Image Based on Graph Convolutional Broad Network

Hyperspectral image (HSI) classification has attracted much attention in the field of remote sensing. However, lack sufficient labeled training samples is a huge challenge for HSI classification. To face this challenge, we propose semisupervised method based on graph convolutional broad network (GCBN). First, to avoid underfitting problem caused by insufficient linear sparse feature representation ability learning system (BLS), convolution operation applied extract nonlinear and discriminative spectral-spatial features from original replace mapping traditional BLS. Second, solve model limited samples, combinatorial average (CAM) proposed use valuable paired generate sample expansion set GCBN training. Third, BLS used perform extracted GCN extended CAM, which further enhances ability. Finally, output weights can be easily calculated ridge regression theory. Experimental results three real datasets demonstrate effectiveness our GCBN.