遥感影像智能解译系统EasyFeature的关键技术及应用  

Key Technologies and Applications of EasyFeature--An Intelligent Interpretation System for Remote Sensing Images

作　　者：胡翔云[1] 张觅[1] 张祖勋[1] 李小凯 邓凯 姜慧伟 庞世燕 饶友琢 宫金杞 冯存均 詹远增 王兴坤 HU Xiangyun;ZHANG Mi;ZHANG Zuxun;LI Xiaokai;DENG Kai;JIANG Huiwei;PANG Shiyan;RAO Youzhuo;GONG Jinqi;FENG Cunjun;ZHAN Yuanzeng;WANG Xingkun(School of Remote Sensing and Information Engineering,Wuhan University,Wuhan 430079,China;Wuhan Daspatial Technology Co.Ltd,Wuhan 430223,China;National Geomatics Center of China,Beijing 100830,China;Faculty of Artificial Intelligence in Education,Central China Normal University,Wuhan 430079,China;Wuhan Handleray Technology Co.Ltd,Wuhan 430073,China;School of Geomatics Science and Technology,Nanjing Tech University,Nanjing 211816,China;Zhejiang Academy of Surveying and Mapping,Hangzhou 310012,China;Zhejiang Application Center of Nature Resources Satellite Technology,Hangzhou 310012,China;Key Laboratory of National Geographic Census and Monitoring,Hangzhou 310012,China)

机构地区：[1]武汉大学遥感信息工程学院,湖北武汉430079 [2]武汉大势智慧科技有限公司,湖北武汉430223 [3]国家基础地理信息中心,北京100830 [4]华中师范大学人工智能教育学部,湖北武汉430079 [5]武汉汉达瑞科技有限公司,湖北武汉430073 [6]南京工业大学测绘科学与技术学院,江苏南京211816 [7]浙江省测绘科学技术研究院,浙江杭州310012 [8]自然资源浙江省卫星应用技术中心,浙江杭州310012 [9]自然资源部地理国情监测重点实验室,浙江杭州310012

出　　处：《武汉大学学报(信息科学版)》2025年第3期554-561,共8页Geomatics and Information Science of Wuhan University

基　　金：中央高校自主科研项目(2042022dx0001)。

摘　　要：随着大数据和人工智能技术的迅猛发展,遥感影像自动解译技术取得了显著进步,但现有遥感影像自动解译方法在鲁棒性、可靠性以及精度等方面仍难以与人相媲美。面向实际生产应用需求,创建了场景-目标-像素层次关系的多要素提取模型,形成了遥感影像分类和要素提取成套技术;提出了语义信息增强与虚警再抑制机制、融合先验形状、特征匹配优化、二维-三维联合处理等变化检测新方法,以及人机智能协同的交互式地物采编思路,构建了高性能遥感影像智能解译技术体系,研发了自主知识产权软件系统EasyFeature,并在全球测图、自然资源常态化监测等国家重大工程中取得了广泛应用,降低了中国对国外同类软件的依赖。With the rapid development of big data and artificial intelligence technologies,significant progress has been achieved in automatic interpretation techniques for remote sensing imagery.However,the robustness,reliability,and accuracy of existing automatic interpretation methods still fall short of human-level performance when compared.To address the practical production application needs,this research has developed a multiobject extraction model that establishes a hierarchical relationship among scene-object-pixel,thereby creating a comprehensive suite of technologies for remote sensing image classification and feature extraction.It introduces novel methods for change detection,such as semantic information enhancement coupled with false alarm resuppression mechanisms,fusion of prior shape knowledge,optimized feature matching,and 2D-3D joint processing.Furthermore,it proposes a conceptual framework for interactive land feature extraction and editing through human-computer intelligent collaboration,thus constructing a high-performance intelligent interpretation technology system for remote sensing imagery.This research has resulted in the development of an independent intellectual property software system named EasyFeature.And it has been widely applied in major national projects such as global mapping and normalized monitoring of natural resources,effectively reducing China's reliance on foreign comparable software solutions.

关 键 词：深度学习 自动解译 影像分类 变化检测 交互提取 

分 类 号：P237[天文地球—摄影测量与遥感]