计算机工程与应用 ›› 2022, Vol. 58 ›› Issue (8): 58-66.DOI: 10.3778/j.issn.1002-8331.2111-0197

• 热点与综述 • 上一篇    下一篇

SAR图像建筑物目标检测研究综述

石颉,袁晨翔,丁飞,孔维相   

  1. 苏州科技大学 电子与信息工程学院,江苏 苏州 215009
  • 出版日期:2022-04-15 发布日期:2022-04-15

Survey of Building Target Detection in SAR Images

SHI Jie, YUAN Chenxiang, DING Fei, KONG Weixiang   

  1. School of Electronic & Information Engineering, Suzhou University of Science and Technology, Suzhou, Jiangsu 215009, China
  • Online:2022-04-15 Published:2022-04-15

摘要: 面对日益剧增的城市建筑物,合成孔径雷达(synthetic aperture radar,SAR)图像的建筑物检测作为SAR图像解译的一个分支逐渐成为一项重要的研究课题。对现有的研究方法进行了分类,从基于传统方法的建筑物检测和基于深度学习的建筑物检测两方面入手,对现有SAR图像的建筑物目标检测算法进行了梳理。简述了SAR图像的特点和SAR图像建筑物检测任务的整体流程,介绍了基于建模、纹理特征和机器学习的方法以及深度学习的目标检测方法。重点论述了当前基于候选区域和回归的主流检测方法。对各类方法的优势和局限性进行对比分析,总结了当前SAR图像建筑物检测技术存在的主要问题和发展瓶颈,并给出相应建议。最后对该领域未来的研究方向进行了展望。

关键词: 合成孔径雷达(SAR), 统计模型, 建筑物, 深度学习, 目标检测

Abstract: With the increasing city buildings, building detection based on synthetic aperture radar(SAR) images, as a branch of SAR image interpretation, has gradually become an important research topic. In this paper, the existing research methods are classified, and the existing building target detection algorithms of SAR images are sorted out from the perspective of building detection based on traditional methods and on deep learning. In addition, the characteristics of SAR images and the whole process of building detection task with SAR images are described. The methods based on modeling, texture features and machine learning, and the target detection method based on deep learning are introduced. Moreover, emphasis is placed on expounding the current mainstream detection methods based on candidate regions and regression. The advantages and limitations of various methods are compared and analyzed, main problems and development bottlenecks of the current building detection technology based on SAR images are summarized, and corresponding suggestions are put forward. Finally, the future research interests in this field are explored.

Key words: synthetic aperture radar(SAR), statistical model, building, deep learning, target detection