 Our proposed method is able to generate change maps with higher accuracy than other state-of-the-art CD methods. It uses an improved UNET++ network to fuse multiple side outputs and combine global and fine-grained information from core-registered image pairs. This allows us to generate more reliable change maps with fewer errors compared to previous methods. This article was offered by Daifeng Peng, Yongjun Zhang, and Haiyang Guan.