航空学报 > 2026, Vol. 47 Issue (10): 632103-632103   doi: 10.7527/S1000-6893.2025.32103

航天遥感图像智能处理与分析专刊

知识引导下的遥感影像智能解译方法综述

任若天1,2, 赵理君1(), 赵旭阳1,2, 张正1, 李宏益1, 薛新华3, 唐娉1   

  1. 1.中国科学院 空天信息创新研究院,北京 100094
    2.中国科学院大学 电子电气与通信工程学院,北京 100049
    3.中国电子科技集团 第二十八研究所,南京 210007
  • 收稿日期:2025-04-10 修回日期:2025-05-08 接受日期:2025-05-29 出版日期:2025-06-11 发布日期:2025-06-10
  • 通讯作者: 赵理君 E-mail:zhaolj201934@aircas.ac.cn
  • 基金资助:
    民用航天技术预先研究项目(D040404);中国科学院青年创新促进会项目(2022127)

A review of knowledge-guided intelligent interpretation methods for remote sensing imagery

Ruotian REN1,2, Lijun ZHAO1(), Xuyang ZHAO1,2, Zheng ZHANG1, Hongyi LI1, Xinhua XUE3, Ping TANG1   

  1. 1.Aerospace Information Research Institute,Chinese Academy of Sciences,Beijing 100094,China
    2.School of Electronic,Electrical and Communication Engineering,University of Chinese Academy of Sciences,Beijing 100049,China
    3.The 28th Research Institute,China Electrnics Technology Group Corporation,Nanjing 210007,China
  • Received:2025-04-10 Revised:2025-05-08 Accepted:2025-05-29 Online:2025-06-11 Published:2025-06-10
  • Contact: Lijun ZHAO E-mail:zhaolj201934@aircas.ac.cn
  • Supported by:
    Civil Aerospace Technology Pre-research Project of China(D040404);Youth Innovation Promotion Association Project, Chinese Academy of Sciences(2022127)

摘要:

随着遥感技术的快速发展和应用需求的不断提升,遥感影像的智能解译已成为研究的热点。而知识作为对特定领域或现象的理解、经验和信息,能帮助解译模型更好地解释和处理遥感数据,除了能够提升解译精度、减少对标注数据的依赖之外,还能在复杂、不确定场景下增强模型的鲁棒性,为应对多源异构遥感数据的智能处理提供了关键支撑。首先介绍了知识引导下的遥感影像智能解译方法的发展历程,接着归纳了当下常用于解译任务中的知识类型,进而从多种层次总结和比较了不同知识驱动方法的有效性和先进性;最后对知识引导下的遥感影像智能解译方法进行了总结与展望。

关键词: 遥感影像, 智能解译, 领域知识, 深度学习, 人工智能

Abstract:

With the rapid advancement of remote sensing technology and the growing demand for its applications, intelligent interpretation of remote sensing imagery has emerged as a prominent research focus. Knowledge, defined as the comprehension, experience, and information about specific domains or phenomena, plays a crucial role in enhancing interpretation models’ capacity to analyze and process remote sensing data. In addition to improving interpretation accuracy and reducing dependence on annotated data, knowledge significantly strengthens model robustness in complex and uncertain scenarios, thereby providing essential support for the intelligent processing of multi-source heterogeneous remote sensing data. This paper first reviews the evolutionary trajectory of knowledge-guided intelligent interpretation methods for remote sensing imagery,subsequently summarizes the commonly used types of knowledge in interpretation tasks, and then proceeds to compare the effectiveness and advancement of different knowledge-driven approaches at multiple levels. Finally, the paper provides a summary and outlook on the future development of knowledge-guided intelligent interpretation methods for remote sensing imagery.

Key words: remote sensing imagery, intelligent interpretation, domain knowledge, deep learning, artificial intelligence

中图分类号: