现代电子技术2025，Vol.48Issue(23)：58-64,7.DOI:10.16652/j.issn.1004-373x.2025.23.008
基于改进Swin Transformer的遥感图像建筑物变化检测
Improved Swin Transformer based change detection of buildings in remote sensing images
罗季 1王晓红 1杨祎斐 1肖剑1
作者信息
- 1. 贵州大学 矿业学院,贵州 贵阳 550025
- 折叠
摘要
Abstract
In view of the blurred boundary between changing object and background,serious feature loss,and false detection and missed detection in the change detection of buildings in remote sensing images,an improved Swin Transformer based change detection of buildings in remote sensing images is proposed.In this method,the global channel spatial attention(GCSA)module is embedded behind the four Swin Transformer blocks to capture the dependence between the channel and the spatial dimension,and reduce the information loss in feature extraction,so as to make full use of global features and improve the detection ability of boundary division.The proposed method is trained and tested on the public datasets LEVIR-CD and WHU-CD.In comparison with that of the original Swin Transformer,the overall accuracy of the new network on the two datasets is increased by 0.51%and 0.49%,respectively,and its IoU is increased by 6.79%and 6.40%,respectively.It can be seen that the proposed method can improve the recognition accuracy of building change areas effectively.关键词
遥感图像/变化检测/Swin Transformer/通道注意力/空间注意力/特征提取/信息损失/边界划分Key words
remote sensing image/change detection/Swin Transformer/channel attention/spatial attention/feature extraction/information loss/boundary division分类
信息技术与安全科学引用本文复制引用
罗季,王晓红,杨祎斐,肖剑..基于改进Swin Transformer的遥感图像建筑物变化检测[J].现代电子技术,2025,48(23):58-64,7.
