Hyperspectral image (HSI) classification is pivotal for remote sensing applications, including environmental monitoring, precision agriculture, and urban land-use analysis. However, its accuracy is often limited by scarce labeled data, class imbalance, and domain discrepancies between standard RGB and HSI imagery. Although recent deep learning approaches, such as 3D convolutional neural networks (3D-CNNs), transformers, and generative adversarial networks (GANs), show promise, they struggle with spectral fidelity, computational efficiency, and cross-domain adaptation in label-scarce scenarios. To address these challenges, we propose the Transformer–Graph Convolutional Network–Diffusion with Hybrid Domain Adaptation (TGDHTL) framework. This framework integrates domain-adaptive alignment of RGB and HSI data, efficient synthetic data generation, and multi-scale spectral–spatial modeling. Specifically, a lightweight transformer, guided by Maximum Mean Discrepancy (MMD) loss, aligns feature distributions across domains. A class-conditional diffusion model generates high-quality samples for underrepresented classes in only 15 inference steps, reducing labeled data needs by approximately 25% and computational costs by up to 80% compared to traditional 1000-step diffusion models. Additionally, a Multi-Scale Stripe Attention (MSSA) mechanism, combined with a Graph Convolutional Network (GCN), enhances pixel-level spatial coherence. Evaluated on six benchmark datasets including HJ-1A and WHU-OHS, TGDHTL consistently achieves high overall accuracy (e.g., 97.89% on University of Pavia) with just 11.9 GFLOPs, surpassing state-of-the-art methods. This framework provides a scalable, data-efficient solution for HSI classification under domain shifts and resource constraints.

Hyperspectral image classification (HSIC) plays a critical role in agriculture, environmental monitoring, and urban planning. Despite extensive research, existing methods face persistent challenges, including high spectral dimensionality, class imbalance, and limited integration of local and global features. Convolutional neural networks (CNNs) are constrained by weak long-range dependency modeling, while vision Transformers (ViTs) and graph convolutional networks (GCNs) often suffer from data hunger and high computational overhead. To address these limitations, we propose HyperViTGCN, a hybrid framework that synergistically combines 3D-CNNs, lightweight ViTs, and attention-guided GCNs (AGCNs). The proposed architecture incorporates: (1) a self-attention-driven dynamic graph for capturing fine-grained inter-pixel relationships, (2) adaptive multiscale patch embedding for improved scene variability handling, and (3) streamlined Transformer-GCN modules optimized through pruning and distillation for computational efficiency. In addition, autoencoder-based normalization and GAN-driven minority class augmentation mitigate data imbalance and improve representation learning. Extensive experiments on multiple benchmark datasets, including Indian Pines, Houston2013, PaviaU, Botswana, HJ-1A, ZY-1-02D, and WHU-OHS (large-scale urban scenes with 32 bands from eight satellites at 10-meter resolution, enhancing land use change analysis and classification fidelity), demonstrate that HyperViTGCN achieves state-of-the-art performance whilereducing computational complexity by up to 30% compared to existing hybrid approaches. These results, particularly on diverse sensors like ZY-1-02D and large-scale urban scenes in WHU-OHS, highlight the model's robustness and scalability for domain-adaptive HSIC.