Abstract: This research addresses the limitations of existing remote sensing change detection methods in feature extraction and difference modeling, investigating approaches to enhance bi-temporal feature interaction quality and change region detection accuracy.A novel dual-branch feature interaction method for change detection is proposed, comprising two core modules: The Patch Channel Exchange module, which implements fine-grained bi-temporal feature interaction through channel-dimension exchanges at the patch level; The Dual Branches Feature Difference module, which employs a dual-branch decoding architecture and explicit difference modeling strategy to enhance change region perception.Experimental validation was conducted on two public datasets: SYSU, LEVIR-CD and WHU. The experiments demonstrate that the proposed method achieves F1 scores of 82.22%, 92.50% and 93.11% on SYSU, LEVIR-CD and WHU datasets, respectively. Compared to existing methods, the PCE module improves bi-temporal feature interaction efficiency, while the DBFD module enhances change region detection accuracy.The research demonstrates that patch-level feature interaction and dual-branch difference modeling effectively improve the performance of remote sensing change detection. The method enhances model robustness while maintaining detection accuracy, providing a new technical solution for remote sensing change detection.