结合全卷积神经网络与条件随机场的资源3号遥感影像云检测
|
摘要
提出一种结合全卷积神经网络与条件随机场的资源3号卫星遥感影像云检测方法。优化了全卷积神经网络(FCN)模型,对3次上采样后的全卷积神经网络(FCN-8s)进行上采样,采用自适应+动量算法调整参数学习率加速收敛;将全卷积神经网络与条件随机场结合,以全卷积输出影像作为前端一阶势,高斯核函数作为后端二阶势;加入mean-shift区域约束0条件保护影像的局部特征信息,运用平均场算法推断条件随机场模型后验概率。实验结果表明,本研究提出的云检测方法可将影像云区识别准确率提高至97.38%,较FCN-8s算法提高13.42%。
A novel method for the cloud detection of ZY-3 satellite remote sensing images is proposed based on the fully convolutional neural network(FCN)combined with the conditional random field.The model of a fully convolutional neural network is optimized and the FCN after three times of upsampling(FCN-8 s)is upsampled.The momentum combined adaptive algorithm is used for the acceleration of convergence by adjusting the learning rate of parameters.The fully convolutional neural network is combined with the conditional random field,the fully convolutional output image is taken as the first-order potential of the front end,and the Gaussian kernel function is used as the second-order potential of the back end.The mean-shift regional constraints are added to protect the local feature information of images and the posterior probability of the conditional random field model is inferred by the mean field algorithm.The experimental results show that the proposed cloud detection method can increase the identification accuracy rate of an image cloud region to 97.38%,which is 13.42% higher than that from FCN-8 s.
A novel method for the cloud detection of ZY-3 satellite remote sensing images is proposed based on the fully convolutional neural network(FCN)combined with the conditional random field.The model of a fully convolutional neural network is optimized and the FCN after three times of upsampling(FCN-8 s)is upsampled.The momentum combined adaptive algorithm is used for the acceleration of convergence by adjusting the learning rate of parameters.The fully convolutional neural network is combined with the conditional random field,the fully convolutional output image is taken as the first-order potential of the front end,and the Gaussian kernel function is used as the second-order potential of the back end.The mean-shift regional constraints are added to protect the local feature information of images and the posterior probability of the conditional random field model is inferred by the mean field algorithm.The experimental results show that the proposed cloud detection method can increase the identification accuracy rate of an image cloud region to 97.38%,which is 13.42% higher than that from FCN-8 s.
引文
[1] Liu D W,Han L,Han X Y.High spatial resolution remote sensing image classification based on deep learning[J].Acta Optica Sinica,2016,36(4):0428001.刘大伟,韩玲,韩晓勇.基于深度学习的高分辨率遥感影像分类研究[J].光学学报,2016,36(4):0428001.
[2] Krizhevsky A,Sutskever I,Hinton G E.ImageNet classification with deep convolutional neural network[C/OL]//Advances in Neural Information Processing Systems,2012:1097-1105.[2018-08-28].https://papers.nips.cc/paper/4824-imagenet-classificationwith-deep-convolutional-neural-networks.pdf.
[3] Chen Y,Fan R S,Wang J X,et al.Cloud detection of ZY-3satellite remote sensing images based on deep learning[J].Acta Optica Sinica,2018,38(1):0128005.陈洋,范荣双,王竞雪,等.基于深度学习的资源三号卫星遥感影像云检测方法[J].光学学报,2018,38(1):0128005.
[4] Long J,Shelhamer E,Darrell T.Fully convolutional networks for semantic segmentation[C]//2015IEEE ConferenceonComputerVisionandPattern Recognition,June 7-12,2015,Boston,MA,USA.New York:IEEE,2015:3431-3440.
[5] Koller D,Friedman N.Probabilistic graphical models principles and techniques[M].Cambridge:MIT Press,2009.
[6] Geman S,Geman D.Stochastic relaxation,Gibbs distributions,and the Bayesian restoration of images[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,1984,6(6):721-741.
[7] Lafferty J,McCallum A,Pereira F.Conditional random fields:probabilistic models for segmenting andlabelingsequencedata[C/OL]//18th International Conference on Machine Learning,June28-July 01, 2001, San Francisco, CA, USA.Burlington, MA:Morgan Kaufmann Publishers,2001:282-289.[2018-07-20].http://portal.acm.org/citation.cfm?id=655813.
[8] Xia M,Cao G,Wang G Y,et al.Remote sensing image classification based on deep learning and conditional random fields[J].Journal of Image and Graphics,2017,22(9):1289-1301.夏梦,曹国,汪光亚,等.结合深度学习与条件随机场的遥感图像分类[J].中国图象图形学报,2017,22(9):1289-1301.
[9] Li Z Q,Zhu R F,Gao F,et al. Hyperspectral remote sensing image classification based on threedimensional convolution neural network combined with conditional random field optimization[J].Acta Optica Sinica,2018,38(8):0828001.李竺强,朱瑞飞,高放,等.三维卷积神经网络模型联合条件随机场优化的高光谱遥感影像分类[J].光学学报,2018,38(8):0828001.
[10] Kumar S,Hebert M.Discriminative random fields:a discriminative framework for contextual interaction in classification[C]//ProceedingsNinthIEEE International ConferenceonComputerVision,October 13-16,2003,Nice,France.New York:IEEE,2003:1150-1157.
[11] Farabet C,Couprie C,Najman L,et al.Learning hierarchical features for scene labeling[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2013,35(8):1915-1929.
[12] Tompson J,Jain A,Lecun Y,et al.Joint training of a convolutional network and a graphical model for human pose estimation[EB/OL].(2014-09-17)[2018-07-20].https://arxiv.org/abs/1406.2984v2.
[13] Liu F,Shen C,Lin G,et al.Learning depth from single monocular images using deep convolutional neural fields[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2016,38(10):2024-2039.
[14] Liu X, Wang G H, Yang H C,et al. Road extraction from remote sensing image based on fully convolutionalnetworks[J]. RemoteSensing Information,2018,33(1):69-75.刘笑,王光辉,杨化超,等.全卷积神经网络遥感影像道路提取方法[J].遥感信息,2018,33(1):69-75.
[15] Simonyan K,Zisse R, Man A. Very deep convolutionalnetworksforlarge-scaleimage recognition[EB/OL].(2015-04-10)[2018-01-25].https://arxiv.org/abs/1409.1556v6.
[16] Krahenbühl P,Koltun V.Efficient inference in fully connected CRFs with Gaussian edge potentials[J].Advances in Neural Information Processing Systems,2011(24):109-117.
[17] Zhao X.Study on cloud detection method for high resolution satellite remote sensing images[D].Harbin:Harbin Institute of Technology,2013:40-47.赵晓.高分辨率卫星遥感图像云检测方法研究[D].哈尔滨:哈尔滨工业大学,2013:40-47.
[2] Krizhevsky A,Sutskever I,Hinton G E.ImageNet classification with deep convolutional neural network[C/OL]//Advances in Neural Information Processing Systems,2012:1097-1105.[2018-08-28].https://papers.nips.cc/paper/4824-imagenet-classificationwith-deep-convolutional-neural-networks.pdf.
[3] Chen Y,Fan R S,Wang J X,et al.Cloud detection of ZY-3satellite remote sensing images based on deep learning[J].Acta Optica Sinica,2018,38(1):0128005.陈洋,范荣双,王竞雪,等.基于深度学习的资源三号卫星遥感影像云检测方法[J].光学学报,2018,38(1):0128005.
[4] Long J,Shelhamer E,Darrell T.Fully convolutional networks for semantic segmentation[C]//2015IEEE ConferenceonComputerVisionandPattern Recognition,June 7-12,2015,Boston,MA,USA.New York:IEEE,2015:3431-3440.
[5] Koller D,Friedman N.Probabilistic graphical models principles and techniques[M].Cambridge:MIT Press,2009.
[6] Geman S,Geman D.Stochastic relaxation,Gibbs distributions,and the Bayesian restoration of images[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,1984,6(6):721-741.
[7] Lafferty J,McCallum A,Pereira F.Conditional random fields:probabilistic models for segmenting andlabelingsequencedata[C/OL]//18th International Conference on Machine Learning,June28-July 01, 2001, San Francisco, CA, USA.Burlington, MA:Morgan Kaufmann Publishers,2001:282-289.[2018-07-20].http://portal.acm.org/citation.cfm?id=655813.
[8] Xia M,Cao G,Wang G Y,et al.Remote sensing image classification based on deep learning and conditional random fields[J].Journal of Image and Graphics,2017,22(9):1289-1301.夏梦,曹国,汪光亚,等.结合深度学习与条件随机场的遥感图像分类[J].中国图象图形学报,2017,22(9):1289-1301.
[9] Li Z Q,Zhu R F,Gao F,et al. Hyperspectral remote sensing image classification based on threedimensional convolution neural network combined with conditional random field optimization[J].Acta Optica Sinica,2018,38(8):0828001.李竺强,朱瑞飞,高放,等.三维卷积神经网络模型联合条件随机场优化的高光谱遥感影像分类[J].光学学报,2018,38(8):0828001.
[10] Kumar S,Hebert M.Discriminative random fields:a discriminative framework for contextual interaction in classification[C]//ProceedingsNinthIEEE International ConferenceonComputerVision,October 13-16,2003,Nice,France.New York:IEEE,2003:1150-1157.
[11] Farabet C,Couprie C,Najman L,et al.Learning hierarchical features for scene labeling[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2013,35(8):1915-1929.
[12] Tompson J,Jain A,Lecun Y,et al.Joint training of a convolutional network and a graphical model for human pose estimation[EB/OL].(2014-09-17)[2018-07-20].https://arxiv.org/abs/1406.2984v2.
[13] Liu F,Shen C,Lin G,et al.Learning depth from single monocular images using deep convolutional neural fields[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2016,38(10):2024-2039.
[14] Liu X, Wang G H, Yang H C,et al. Road extraction from remote sensing image based on fully convolutionalnetworks[J]. RemoteSensing Information,2018,33(1):69-75.刘笑,王光辉,杨化超,等.全卷积神经网络遥感影像道路提取方法[J].遥感信息,2018,33(1):69-75.
[15] Simonyan K,Zisse R, Man A. Very deep convolutionalnetworksforlarge-scaleimage recognition[EB/OL].(2015-04-10)[2018-01-25].https://arxiv.org/abs/1409.1556v6.
[16] Krahenbühl P,Koltun V.Efficient inference in fully connected CRFs with Gaussian edge potentials[J].Advances in Neural Information Processing Systems,2011(24):109-117.
[17] Zhao X.Study on cloud detection method for high resolution satellite remote sensing images[D].Harbin:Harbin Institute of Technology,2013:40-47.赵晓.高分辨率卫星遥感图像云检测方法研究[D].哈尔滨:哈尔滨工业大学,2013:40-47.