A novel urban LC change classification using spatial and channel dual mountaineering team conditional generative adversarial network

Urban Land Cover (LC) change classification is vital for monitoring environmental impacts and urban expansion, but traditional methods face challenges such as noise, class imbalance, and LC variability. This research introduces a multi-step framework to address these issues. Gabor Filtering is applied to enhance texture and boundary details to achieve clearer features. Segmentation is done using a U-Net Transformer Encoder, with the capability to extract spatial and contextual information. A Spatial and Channel Dual Conditional Generative Adversarial Network (SCDCGAN) extracts features and represents intricate relationships of LC for classification. The Mountaineering Team-based Optimization Algorithm (MTOA) optimizes SCDCGAN parameters to reduce classification errors and improve accuracy. The suggested method exhibits better performance on Dataset A (99.33% accuracy) and Dataset B (95.67% accuracy). These performances show significant improvements in classification and segmentation accuracy, especially for skewed urban datasets, providing a strong solution for urban LC change classification.