<p>Hyperspectral image (HSI) classification plays a crucial role in remote sensing applications, particularly in urban environment monitoring. However, the high dimensionality of spectral data and complex spatial structures in urban scenes pose significant challenges for accurate classification. This paper introduces Dual-Branch Spectral-Spatial Attention Network with Cross-Modal Integration (DBSSANet-CMI), a novel deep learning architecture that synergistically combines Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs) for enhanced HSI classification. Our model features a unique CNN-to-ViT Integration (C2VI) module that facilitates effective cross-modal learning, and a Multi-scale Convolution Fusion (MCF) module for capturing multi-scale spatial information. Experiments on the University of Pavia dataset demonstrate that DBSSANet-CMI outperforms state-of-the-art methods, achieving an overall accuracy of 98.12%. The model shows robust performance even with limited training data, maintaining high accuracy when trained on only 10% of the dataset. Ablation studies confirm the significant contribution of each proposed component to the model’s performance. DBSSANet-CMI’s ability to effectively integrate spectral and spatial information while balancing local details and global context makes it a powerful tool for accurate urban land cover classification using hyperspectral imagery.</p>