especially in scenarios with limited labeled samples. To tackle these problems, we propose a novel approach called the dual-semantic graph convolution network (DSGCN) for HSI classification in this ...