测绘学报 ›› 2023, Vol. 52 ›› Issue (2): 283-296.doi: 10.11947/j.AGCS.2023.20220202

• 摄影测量学与遥感 • 上一篇    下一篇

联合UNet++和多级差分模块的多源光学遥感影像对象级变化检测

王超1, 王帅1, 陈晓1,4, 李俊勇1, 谢涛2,3   

  1. 1. 南京信息工程大学电子与信息工程学院, 南京 210044;
    2. 南京信息工程大学遥感与测绘工程学院, 南京 210044;
    3. 青岛海洋科学技术国家实验室区域海洋学与数值模拟实验室, 青岛 266237;
    4. 南京信息工程大学江苏省大气环境与装备技术协同创新中心, 南京 210044
  • 收稿日期:2022-03-18 修回日期:2022-07-15 发布日期:2023-03-07
  • 通讯作者: 谢涛 E-mail:xietao@nuist.edu.cn
  • 作者简介:王超(1984-),博士,副教授,硕士生导师,主要研究方向为高分辨率遥感影像处理。E-mail:chaowang@nuist.edu.cn
  • 基金资助:
    国家重点研发计划(2021YFC2803302);国家自然科学基金(42176180);江苏省应急管理科技项目(YJGL-YF-2020-16);江苏省自然资源发展专项(JSZRHYKJ202114);江苏省博士后基金(2021K013A);江苏省研究生科研与实践创新计划(SJCX22_0335);江苏省六大人才高峰工程(2019XYDXX135)

Object-level change detection of multi-sensor optical remote sensing images combined with UNet++ and multi-level difference module

WANG Chao1, WANG Shuai1, CHEN Xiao1,4, LI Junyong1, XIE Tao2,3   

  1. 1. School of Electronics and Information Engineering, Nanjing University of Information Science and Technology, Nanjing 210044, China;
    2. School of Remote Sensing & Geomatics Engineering, Nanjing University of Information Science & Technology, Nanjing 210044, China;
    3. Laboratory for Regional Oceanography and Numerical Modeling, Qingdao National Laboratory for Marine Science and Technology, Qingdao 266237, China;
    4. Jiangsu Provincial Collaborative Innovation Center of Atmosphere Environment and Equipment Technology, Nanjing University of Information Science and Technology, Nanjing 210044, China
  • Received:2022-03-18 Revised:2022-07-15 Published:2023-03-07
  • Supported by:
    The National Key Research and Development Program of China (No. 2021YFC2803302);The National Natural Science Foundation of China project (No. 42176180);The Natural Science Foundation of Jiangsu Province(No. YJGL-YF-2020-16);The Natural Science Foundation of Jiangsu Province(No. JSZRHYKJ202114);The Post-doctoral fund of Jiangsu Province(No. 2021K013A);Postgraduate Research & Practice Innovation Program of Jiangsu Province(No. SJCX22_0335);The Six Talent-peak Project in Jiangsu Province (No. 2019XYDXX135)

摘要: 随着传感器技术的飞速发展,基于多源光学遥感影像的变化检测已成为遥感领域中的研究热点。由于传感器成像差异,同一景象在多源光学遥感影像中通常呈现出不同的表现形式,因此面临着更加突出的“伪变化”问题。为此,本文提出了一种联合UNet++和多级差分模块的多源光学遥感影像对象级变化检测方法。该方法首先提出了一种多尺度特征提取差分(multi-scale feature extraction difference,MFED)模块,以增强模型对“伪变化”的识别能力;在此基础上,利用UNet++网络输出的多尺度特征对变化区域进行多角度精细刻画,并提出了一种自适应证据置信度指标(adaptive evidence credibility indicators,AECI);最后结合影像分割与Dempster-Shafer (DS)理论设计了加权DS证据融合策略(weighted dempster shafer evidence fusion,WDSEF),从而实现了深度网络像素级输出至对象级结果的映射。通过对不同地区的4组高分多源光学影像数据集进行试验,并与多种先进的深度学习方法进行对比分析,结果表明:在不同空间分辨率和时相差异条件下,本文方法的总体精度(overall accuracy,OA)和F1 score分别可达91.92%和63.31%以上,在目视分析和定量评价均显著优于对比方法。

关键词: 多源光学遥感影像, 变化检测, UNet++, 多尺度特征提取差分, 自适应证据信度指标, 加权DS证据融合

Abstract: With the rapid development of sensor technology, change detection based on multi-sensor optical remote sensing images has become a research hotspot in the field of remote sensing. Due to the differences of sensor imaging, different patterns of manifestation are shown in multi-sensor optical remote sensing images for one scene, leading to a more obvious problem of "pseudo change". Therefore, an object-level change detection method for multi-sensor optical remote sensing images combining UNet++ and multi-stage difference module is proposed in this paper. Firstly, multi-scale feature extraction difference (MFED) module is proposed by this method to enhance the ability of the model to identify "pseudo change". On this basis, multi-scale feature outputs by UNet++ network are used for multi-angle meticulous depiction. Adaptive evidence credibility indicator (AECI) is proposed as well. At last, image segmentation and Dempster-Shafer (DS) theory are combined to design weighted Dempster-Shafer evidence fusion (WDSEF), so as to achieve mapping from pixel-level output of deep network to object-level results. Experiment was conducted to four sets of high-resolution multi-sensor optical image datasets from different regions, and contrastive analysis was conducted to multiple methods of advanced deep learning. The results revealed that, the overall accuracy (OA) and F1 score of the proposed method reached more than 91.92% and 63.31%, respectively, under different conditions of spatial resolution and temporal phase difference, which were significantly better than the comparison methods in both visual analysis and quantitative evaluation.

Key words: multi-sensor optical remote sensing image, change detection, UNet++, multi-scale feature extraction difference, adaptive evidence reliability index, weighted DS evidence fusion

中图分类号: