留言板

尊敬的读者、作者、审稿人, 关于本刊的投稿、审稿、编辑和出版的任何问题, 您可以本页添加留言。我们将尽快给您答复。谢谢您的支持!

姓名
邮箱
手机号码
标题
留言内容
验证码

基于混合网络的异源遥感图像变化检测

周圆 李祥瑞 杨晶

周圆, 李祥瑞, 杨晶等 . 基于混合网络的异源遥感图像变化检测[J]. 北京航空航天大学学报, 2021, 47(3): 451-460. doi: 10.13700/j.bh.1001-5965.2020.0455
引用本文: 周圆, 李祥瑞, 杨晶等 . 基于混合网络的异源遥感图像变化检测[J]. 北京航空航天大学学报, 2021, 47(3): 451-460. doi: 10.13700/j.bh.1001-5965.2020.0455
ZHOU Yuan, LI Xiangrui, YANG Jinget al. Heterogeneous remote sensing image change detection based on hybrid network[J]. Journal of Beijing University of Aeronautics and Astronautics, 2021, 47(3): 451-460. doi: 10.13700/j.bh.1001-5965.2020.0455(in Chinese)
Citation: ZHOU Yuan, LI Xiangrui, YANG Jinget al. Heterogeneous remote sensing image change detection based on hybrid network[J]. Journal of Beijing University of Aeronautics and Astronautics, 2021, 47(3): 451-460. doi: 10.13700/j.bh.1001-5965.2020.0455(in Chinese)

基于混合网络的异源遥感图像变化检测

doi: 10.13700/j.bh.1001-5965.2020.0455
详细信息
    作者简介:

    周圆   女,博士,教授,博士生导师。主要研究方向:机器学习、计算机视觉、3D图像/视频编码与传输、无线传感器网络、网络视频通信等

    通讯作者:

    周圆, E-mail: zhouyuan@tju.edu.cn

  • 中图分类号: TP751.1

Heterogeneous remote sensing image change detection based on hybrid network

More Information
  • 摘要:

    为了更加准确地进行异源遥感图像的变化检测任务,提出了一种基于混合网络的异源遥感图像变化检测算法。利用伪孪生网络提取异源遥感图像块间空间维度的变化特征,利用早期融合网络提取异源遥感图像块间光谱维度的变化特征,将2支网络提取的特征进行融合,并将融合后的变化特征输入到sigmoid层进行二分类检测。同时,在伪孪生网络中加入对比损失函数,通过优化对比损失函数,使得在特征空间中,未变化图像对的空间特征差异更小,变化图像对的空间特征差异更大,以提升网络的区分能力和收敛速度。

     

  • 图 1  基于混合网络的异源遥感图像变化检测

    Figure 1.  Change detection of heterogeneous remote sensing image based on hybrid network

    图 2  变化检测任务中使用LSTM模型

    Figure 2.  LSTM model used in change detection task

    图 3  2D卷积和3D卷积操作的比较

    Figure 3.  Comparison of 2D convolution and 3D

    图 4  Bastrop复杂火灾数据集

    Figure 4.  Bastrop complex fire dataset

    图 5  Onera变化检测数据集

    Figure 5.  Onera change detection dataset

    图 6  HCNN的网络配置示意图

    Figure 6.  Schematic diagram of HCNN network configuration

    图 7  Bastrop复杂火灾数据集的变化检测结果数据分析(Kappa系数)

    Figure 7.  Change detection result data analysis of Bastrop complex fire dataset (Kappa coefficient)

    图 8  L5T1 vs. L5T2实验中的变化检测结果

    Figure 8.  Results of change detection in L5T1 vs. L5T2 experiment

    图 9  L5T1 vs. ALIT2实验中的变化检测结果

    Figure 9.  Results of change detection in L5T1 vs. ALIT2 experiment

    图 10  L5T1 vs. L8T2实验中的变化检测结果

    Figure 10.  Results of change detection in L5T1 vs. L8T2 experiment

    图 11  在香港图像对上使用不同算法获得的变化检测结果

    Figure 11.  Change detection results on Hong Kong image pair obtained by different algorithms

    图 12  在孟买图像对上使用不同算法获得的变化检测结果

    Figure 12.  Change detection results on Mumbai image pair obtained by different algorithms

    图 13  PS网络与HCNN网络是否添加对比损失函数的二元交叉熵损失随着网络训练的变化曲线

    Figure 13.  Binary cross entropy loss curves of PS and HCNN networks with and without contrast loss function binary in network training

    表  1  Bastrop复杂火灾数据集的变化检测结果

    Table  1.   Change detection results on Bastrop complex fire dataset

    算法 OA/%
    L5T1 vs.L5T2 L5T1 vs.ALIT2 L5T1 vs.L8T2
    MAD+CVA 87.233 79.991 77.075
    KCCA+CVA 92.614 90.175 91.671
    DCCA+CVA 95.556 94.676 93.943
    DCCAE+CVA 96.598 95.713 94.308
    EF 97.358 97.115 94.314
    PS 98.382 98.136 94.526
    HCNN 99.286 98.764 95.581
    下载: 导出CSV

    表  2  在香港图像对上使用不同算法获得的变化检测定量结果

    Table  2.   Quantitative results of change detection obtained by different algorithms on Hong Kong image pairs

    算法 OA/% AA/% Kappa系数
    S 96.542 68.511 0.424
    PS 97.900 76.090 0.631
    EF 97.134 74.943 0.598
    HCNN 97.617 81.200 0.656
    下载: 导出CSV

    表  3  在孟买图像对上使用不同算法获得的变化检测定量结果

    Table  3.   Quantitative results of change detection obtained by different algorithms on Mumbai image pairs

    算法 OA/% AA/% Kappa系数
    S 95.928 69.293 0.457
    PS 96.325 80.217 0.635
    EF 95.449 80.599 0.628
    HCNN 97.293 82.583 0.673
    下载: 导出CSV

    表  4  L5T1 vs.L8T2实验中EF网络与HCNN网络使用2D卷积和3D卷积对检测精度的影响

    Table  4.   Effect of EF and HCNN networks with 2D convolution and 3D convolution on detection accuracy in L5T1 vs. L8T2 experiment

    算法 OA/% Kappa系数
    EF(2D卷积) 94.314 0.711
    EF(3D卷积) 94.938 0.723
    HCNN(2D卷积) 95.258 0.731
    HCNN(3D卷积) 95.581 0.759
    下载: 导出CSV

    表  5  L5T1 vs.ALIT2实验中PS网络与HCNN网络是否添加对比损失函数对检测精度的影响

    Table  5.   Effect of PS and HCNN networks with and without contrast loss function on detection accuracy in L5T1 vs. ALIT2 experiment

    算法 OA/% Kappa系数
    PS(未添加对比损失函数) 98.136 0.903
    HCNN(未添加对比损失函数) 98.66 0.932
    PS(添加对比损失函数) 98.356 0.917
    HCNN(添加对比损失函数) 98.764 0.943
    下载: 导出CSV
  • [1] ALBERGA V. Similarity measures of remotely sensed multi-sensor images for change detection applications[J]. Remote Sensing, 2009, 1(3): 122-143. doi: 10.3390/rs1030122
    [2] FERNANDEZ-PRIETO D, MARCONCINI M. A novel partially supervised approach to targeted change detection[J]. IEEE Transactions on Geoscience and Remote Sensing, 2011, 49(12): 5016-5038. doi: 10.1109/TGRS.2011.2154336
    [3] LAI P L, FYFE C. Kernel and nonlinear canonical correlation analysis[J]. International Journal of Neural Systems, 2000, 10(5): 365-377. doi: 10.1142/S012906570000034X
    [4] VOLPI M, DE MORSIER F, CAMPS-VALLS G, et al. Multi-sensor change detection based on nonlinear canonical correlations[C]//2013 IEEE International Geoscience and Remote Sensing Symposium. Piscataway: IEEE Press, 2013: 1944-1947.
    [5] ZHU Z, WOODCOCK C E. Object-based cloud and cloud shadow detection in Landsat imagery[J]. Remote Sensing of Environment, 2012, 118: 83-94. doi: 10.1016/j.rse.2011.10.028
    [6] JIAN M, LAM K, DONG J, et al. Visual-patch-attention-aware saliency detection[J]. IEEE Transactions on Systems, Man, and Cybernetics, 2015, 45(8): 1575-1586. http://europepmc.org/abstract/med/25291809
    [7] JIAN M, QI Q, DONG J, et al. Integrating QDWD with pattern distinctness and local contrast for underwater saliency detection[J]. Journal of Visual Communication and Image Representation, 2018, 53: 31-41. doi: 10.1016/j.jvcir.2018.03.008
    [8] JIAN M, ZHANG W, YU H, et al. Saliency detection based on directional patches extraction and principal local color contrast[J]. Journal of Visual Communication and Image Representation, 2018, 57: 1-11. doi: 10.1016/j.jvcir.2018.10.008
    [9] MERCIER G, MOSER G, SERPICO S B. Conditional copulas for change detection in heterogeneous remote sensing images[J]. IEEE Transactions on Geoscience and Remote Sensing, 2008, 46(5): 1428-1441. doi: 10.1109/TGRS.2008.916476
    [10] GAO F, LIU X, DONG J, et al. Change detection in SAR images based on deep Semi-NMF and SVD networks[J]. Remote Sensing, 2017, 9(5): 435. doi: 10.3390/rs9050435
    [11] MOU L, BRUZZONE L, ZHU X X. Learning spectral-spatial-temporal features via a recurrent convolutional neural network for change detection in multispectral imagery[J]. IEEE Transactions on Geoscience and Remote Sensing, 2019, 57(2): 924-935. doi: 10.1109/TGRS.2018.2863224
    [12] ZHANG P, GONG M, SU L, et al. Change detection based on deep feature representation and mapping transformation for multi-spatial-resolution remote sensing images[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2016, 116: 24-41. doi: 10.1016/j.isprsjprs.2016.02.013
    [13] ZHANG Z, VOSSELMAN G, GERKE M, et al. Change detection between multimodal remote sensing data using Siamese CNN[EB/OL]. (2018-07-25)[2020-08-01]. https://arxiv.org/abs/1807.09562.
    [14] LIU J, GONG M, QIN K, et al. A deep convolutional coupling network for change detection based on heterogeneous optical and radar images[J]. IEEE Transactions on Neural Networks and Learning Systems, 2016, 29(3): 545-559. http://ieeexplore.ieee.org/document/7795259/citations
    [15] HUGHES L H, SCHMITT M, MOU L, et al. Identifying corresponding patches in SAR and optical images with a pseudo-Siamese CNN[J]. IEEE Geoscience and Remote Sensing Letters, 2018, 15(5): 784-788. doi: 10.1109/LGRS.2018.2799232
    [16] NIU X, GONG M, ZHAN T, et al. A conditional adversarial network for change detection in heterogeneous images[J]. IEEE Geoscience and Remote Sensing Letters, 2018, 16(1): 45-49. http://it.ckcest.cn/portal.php?mod=viewaid=3338420
    [17] HOCHREITER S, SCHMIDHUBER J. Long short-term memory[J]. Neural Computation, 1997, 9(8): 1735-1780. doi: 10.1162/neco.1997.9.8.1735
    [18] LI Y, ZHANG H, SHEN Q. Spectral-spatial classification of hyperspectral imagery with 3D convolutional neural network[J]. Remote Sensing, 2017, 9(1): 67. doi: 10.3390/rs9010067
    [19] CHOPRA S, HADSELL R, LECUN Y. Learning a similarity metric discriminatively, with application to face verification[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE Press, 2005, 1: 539-546.
    [20] DAUDT R C, LE SAUX B, BOULCH A, et al. Urban change detection for multispectral earth observation using convolutional neural networks[C]//2018 IEEE International Geoscience and Remote Sensing Symposium. Piscataway: IEEE Press, 2018: 2115-2118.
    [21] VOLPI M, CAMPS-VALLS G, TUIA D. Spectral alignment of multi-temporal cross-sensor images with automated kernel canonical correlation analysis[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2015, 107: 50-63. doi: 10.1016/j.isprsjprs.2015.02.005
    [22] ANDREW G, ARORA R, BILMES J, et al. Deep canonical correlation analysis[C]//International Conference on International Conference on Machine Learning, 2013, 28: 1247-1255.
    [23] WANG W, ARORA R, LIVESCU K, et al. On deep multi-view representation learning[C]//International Conference on Machine Learning, 2015: 1083-1092.
    [24] ZAGORUYKO S, KOMODAKIS N. Learning to compare image patches via convolutional neural networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE Press, 2015: 4353-4361.
  • 加载中
图(13) / 表(5)
计量
  • 文章访问数:  207
  • HTML全文浏览量:  11
  • PDF下载量:  106
  • 被引次数: 0
出版历程
  • 收稿日期:  2020-08-24
  • 录用日期:  2020-09-25
  • 刊出日期:  2021-03-20

目录

    /

    返回文章
    返回
    常见问答