A remote sensing image change detection method based on multi-scale feature cross-fusion is proposed to address the issues of missed detections and false alarms for small targets,as well as insufficient segmentation of fine details in traditional change detection methods.The method adopts an encoder-decoder structure with the introduction of Multi-Scale Feature Fusion (MSFF) module and Multi-Scale Attention (MSA) mechanism between the encoder and decoder.The former is used to aggregate information from different scales,while the latter captures differences and correlations between different scales.A Refinement Output (RO) module is introduced at the end of the encoder,using parallel convolutions with dilation and atrous convolutions to further refine the fine details and reduce edge information loss.Experimental results on the LEVIR-CD dataset demonstrate that the proposed method can effectively identify small targets while preserving fine details,leading to significant improvements in accuracy,F1 score,intersection-over-union,and overall accuracy.