摘要
建筑物的变化信息提取是遥感影像提取的重要内容之一,对土地调查、城市规划、土地执法等具有重要意义。针对原始U-Net模型预测效果较差、存在漏检等问题,本文提出了一种融合聚合残差卷积块和注意力模块的改进U-Net模型。结果表明,改进后的U-Net模型在建筑物变化信息提取上相比原始的U-Net模型,精度有很大的提升,可为建筑物变化监测提供一定的技术支持。
The extraction of building change information is one of the important contents of remote sensing image extraction,which is of great significance to land survey,urban planning and land law enforcement.Aiming at the problems of poor prediction effect and omission detection in the original U-Net model,this paper proposes an improved U-Net model which integrates the aggregated residual convolution block and attention module.Compared with the original U-Net model,the accuracy of the improved U-Net model in building change information extraction is greatly improved.This study can provide some technical support for building change monitoring.
作者
凡建林
姚辉
高叶
FAN Jianlin;YAO Hui;GAO Ye(Zhejiang Provincial Institute of Surveying and Mapping Science and Technology,Hangzhou 310000,China;Ningbo Metallurgical Survey,Design and Research Co.,Ltd.,Ningbo 315000,China)
出处
《测绘与空间地理信息》
2024年第9期218-220,224,共4页
Geomatics & Spatial Information Technology
关键词
深度学习
建筑物提取
聚合残差卷积块
注意力机制
deep learning
building extraction
aggregated residual convolution block
attention mechanism
