To address the issues of blurred boundaries, weak semantic modeling, and limited anomaly response in GIS-based remote sensing change detection, this paper proposes a change detection and anomaly recognition method based on a Double-Layer Multi-Coding (DLMC) structure. The proposed model integrates cross-layer encoding, structural differencing, and multi-temporal semantic fusion modules to construct a progressive representation path for change modeling. In addition, a joint modeling mechanism combining a saliency map and anomaly factor is introduced to enhance sensitivity to local mutations and spatial inconsistencies. Experiments conducted on the LEVIR-CD, WHU-CD, and GF-Change-GIS datasets demonstrate that DLMC achieves the best performance with an F1-score of 0.902, a Kappa coefficient of 0.849, and a SAR (Structural Anomaly Response) score of 0.781, significantly outperforming representative models such as FC-EF and ChangeFormer. Further ablation studies and error map visualizations confirm the model’s superiority in boundary-level change detection and localized anomaly identification.