基于神经网络的遥感图像变化检测系统  

Change detection system of remote sensing images based on neural network

作　　者：王祥[1] 张志鑫 牟浩东 刘天航 张玉松 张玉柱[1] WANG Xiang;ZHANG Zhixin;MOU Haodong;LIU Tianhang;ZHANG Yusong;ZHANG Yuzhu(China Yangtze Power Co.,Ltd.,Wuhan 443000,China;Network&Information Center,Changjiang Water Resources Commission,Wuhan 430010,China;Center of Technology Innovation for Digital Enablement of River Basin Management,Wuhan 430010,China;Smart Changjiang Innovation Team of Changjiang Water Resources Commission,Wuhan 430010,China;School of Artificial Intelligence and Automation,Huazhong University of Science and Technology,Wuhan 430074,China)

机构地区：[1]中国长江电力股份有限公司,湖北武汉443000 [2]长江水利委员会网络与信息中心,湖北武汉430010 [3]长江水利委员会流域管理数字赋能技术创新中心,湖北武汉430010 [4]长江水利委员会智慧长江创新团队,湖北武汉430010 [5]华中科技大学人工智能与自动化学院,湖北武汉430074

出　　处：《人民长江》2024年第9期230-237,共8页Yangtze River

基　　金：国家自然科学基金项目(62071456);长江水利委员会流域管理数字赋能技术创新中心基金项目。

摘　　要：变化检测是计算机视觉领域极具研究价值的技术分支。为提高长江沿岸生产建设活动的监管效率,以影像配准为基础,结合自注意力与特征信息提取,提出了长江沿岸建筑物变化检测系统。该系统集影像处理、影像配准、变化检测为一体,实现了全流程自动化处理的功能。结果表明:①设计的系统在影像配准上将精度误差控制在2个像素内,达到了像素级精度。②该系统在变化检测测试数据集上的F 1值取得了优于75%的结果,证明该系统的精度较高。研究成果也可应用于大型水库岸线保护监测等场景中,提升库区治理管理的效率。Change detection is a highly valuable research branch in the field of computer vision.In order to improve the monitoring efficiency of production and construction activities along the Changjaing River,a building change detection system along the Chanjiang River was proposed on the basis of image registration,combined with self-attention and feature extraction.It integrated image processing,image registration and change detection,achieving automatic operation across overall processes.The results show that the designed system minimizes the registration accuracy error within two pixels,achieving pixel-level accuracy.Moreover,the system achieves a result of more than 75%in terms of F 1 score on the proposed change detection test dataset,demonstrating its high accuracy.The research can be further applied in the shoreline management of large reservoirs,improving the efficiency of governance of reservoirs.

关 键 词：遥感图像 人工智能 影像配准 变化检测 岸线监测 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]