局部聚类分析的FCN-CNN云图分割方法
|
摘要
空气中的尘埃、污染物及气溶胶粒子的存在严重影响了大气预测的有效性,毫米波雷达云图的有效分割成为解决这一问题的关键.提出了一种基于超像素分析的全卷积神经网路FCN和深度卷积神经网络CNN(FCNCNN)的云图分割方法.首先通过超像素分析对云图每个像素点的近邻域实现相应的聚类,同时将云图输入到不同步长的全卷积神经网络FCN 32s和FCN 8s中实现云图的预分割;FCN 32s预测结果中的"非云"区域一定是云图中的部分"非云"区域,FCN 8s预测结果中的"云"区域一定是云图中的部分"云"区域;余下的不确定的区域通过深度卷积神经网络CNN进行进一步分析.为提高效率,FCN-CNN选取了不确定区域中超像素的几个关键像素来代表超像素区域的特征,通过CNN网络来判断关键像素是"云"或者是"非云".实验结果表明,FCN-CNN的精度与MR-CNN、SP-CNN相当,但是速度相比于MR-CNN提高了880倍,相比于SP-CNN提高了1.657倍.
Dust, pollutant and the aerosol particles in the air bring significant challenge to the atmospheric prediction, and the segmentation of millimeter-wave radar cloud image has become a key to deal with the problem. This paper presents superpixel analysis based cloud image segmentation with fully convolutional networks(FCN) and convolutional neural networks(CNN), named FCN-CNN.Firstly, the superpixel analysis is performed to cluster the neighborhood of each pixel in the cloud image. Then the cloud image is given to the FCN with different steps, such as FCN 32 s and FCN 8 s. The "non-cloud" area in the FCN 32 s result must be a part of the "non-cloud"area in the cloud image. Meanwhile, the "cloud" area in the FCN 8 s result must be a part of the "cloud" area in the cloud image. The remaining uncertain region of the cloud image needs to be further estimated by CNN. For efficiency, it is necessary to select several key pixels in the superpixel to represent the characteristics of the superpixel region. The selected key pixels are classified by CNN as "cloud"or "non-cloud". The experimental results illustrate that while the accuracy of FCN-CNN is almost equivalent to MR-CNN and SP-CNN,the speed is 880 times higher than MR-CNN, and 1.657 times higher than SP-CNN.
Dust, pollutant and the aerosol particles in the air bring significant challenge to the atmospheric prediction, and the segmentation of millimeter-wave radar cloud image has become a key to deal with the problem. This paper presents superpixel analysis based cloud image segmentation with fully convolutional networks(FCN) and convolutional neural networks(CNN), named FCN-CNN.Firstly, the superpixel analysis is performed to cluster the neighborhood of each pixel in the cloud image. Then the cloud image is given to the FCN with different steps, such as FCN 32 s and FCN 8 s. The "non-cloud" area in the FCN 32 s result must be a part of the "non-cloud"area in the cloud image. Meanwhile, the "cloud" area in the FCN 8 s result must be a part of the "cloud" area in the cloud image. The remaining uncertain region of the cloud image needs to be further estimated by CNN. For efficiency, it is necessary to select several key pixels in the superpixel to represent the characteristics of the superpixel region. The selected key pixels are classified by CNN as "cloud"or "non-cloud". The experimental results illustrate that while the accuracy of FCN-CNN is almost equivalent to MR-CNN and SP-CNN,the speed is 880 times higher than MR-CNN, and 1.657 times higher than SP-CNN.
引文
[1]Martínez-Chico M,Batlles FJ,Bosch JL.Cloud classification in a Mediterranean location using radiation data and sky images.Energy,2011,36(7):4055-4062.[doi:10.1016/j.energy.2011.04.043]
[2]Long CN,Sabburg JM,CalbóJ,Pagès D.Retrieving cloud characteristics from ground-based daytime color all-sky images.Journal of Atmospheric&Oceanic Technology,2006,23(5):633-652.[doi:10.1175/JTECH1875.1]
[3]Shields JE,Johnson RW,Karr ME,Wertz JL.Automated day/night whole sky imagers for field assessment of cloud cover distributions and radiance distributions.In:Proc.of the 10th Symp.on Meteorological Observations and Instrumentation.1998.http://xueshu.baidu.com/s?wd=paperuri:(c7eaa7569237eec16316553da79c38bd)&filter=sc_long_sign&sc_ks_para=q%3DAutomat ed+day%2Fnight+whole+sky+imagers+for+field+assessment+of+cloud+cover+distributions+and+radiance+distributions&tn=SE_b aiduxueshu_c1gjeupa&ie=utf-8&sc_us=11998081953391704638
[4]Shields JE,Karr ME,Johnson RW,Burden AR.Day/Night whole sky imagers for 24-h cloud and sky assessment:History and overview.Applied Optics,2013,52(8):1605-1616.[doi:10.1364/A0.52.001605]
[5]Long C,Slater D,Tooman T.Total sky imager model 880 status and testing results.Office of Scientific&Technical Information Technical Reports,2001.https://xs.glgoo.net/scholar?hl=zh-CN&q=Total+sky+imager+model+880+status+and+testing+results&btnG=&lr=
[6]Shaw JA,Thurairajah B.Short-Term arctic cloud statistics at NSA from the infrared cloud imager.2003.https://www.researchgate.net/publication/228878636_Short-term_arctic_cloud_statistics_at_NSA_from_the_infrared_cloud_imager
[7]Racette P,Adler RF,Wang JR,Gasiewski AJ,Jakson DM,Zacharias DS.An airborne millimeter-wave imaging radiometer for cloud,precipitation,and atmospheric water vapor studies.Journal of Atmospheric&Oceanic Technology,1992,13(3):1426-1428.[doi:10.1175/1520-0426(1996)013<0610:AAMWIR>2.0.CO;2]
[8]Kreuter A,Zangerl M,Schwarzmann M,Blumthaler M.All-Sky imaging:A simple,versatile system for atmospheric research.Change,2009,48(6):1091-1097.[doi:10.1364/A0.48.001091]
[9]Mantelli Neto SL,Von Wangenheim A,Pereira EB,Comunello E.The use of euclidean geometric distance on RGB color space for the classification of sky and cloud patterns.Journal of Atmospheric&Oceanic Technology,2010,27(9):1504-1517.[doi:10.1175/2010JTECHA1353.1]
[10]Souzaecher MP,Pereira EB,Bins LS,Andrade MAR.A simple method for the assessment of the cloud cover state in high-latitude regions by a ground-based digital camera.Journal of Atmospheric&Oceanic Technology,2006,23(3):437.[doi:10.1175/JTECH1833.1]
[11]Yang J,LüWT,Ma Y,Yao W.An automatic ground-based cloud detection method based on adaptive threshold.Journal of Applied Meteorological Science,2009,20(6):713-721(in Chinese with English absrtact).
[12]Liu S,Zhang L,Zhang Z,Wang C,Xiao B.Automatic cloud detection for all-sky images using superpixel segmentation.IEEE Geoscience&Remote Sensing Letters,2015,12(2):354-358.[doi:10.1109/LGRS.2014.2341291]
[13]Buch KAJ,Sun CH.Cloud classification using whole-sky imager data.In:Proc.of the 9th Symp.on Meteorological Observations&Instrumentation.1995,16(3):353-358.http://xueshu.baidu.com/s?wd=paperuri%3A(9cebe35398fab368b4f3d5382a8e4d6b)&filter=sc_long_sign&tn=SE_baiduxueshu_c1gjeupa&ie=utf-8&sc_ks_para=q%3DCloud%20classification%20using%20whole-sky%20imager%20data
[14]Taravat A,Frate FD,Cornaro C,Vergari S.Neural networks and support vector machine algorithms for automatic cloud classification of whole-sky ground-based images.IEEE Geoscience&Remote Sensing Letters,2015,12(3):666-670.[doi:10.1109/LGRS.2014.2356616]
[15]Zhang JN,Wu LF,He JY.Millimeter-Wave cloud image segmentation based on multi-resolution convolutional neural network.In:Proc.of the National Conf.on Multimedia Technology.2016(in Chinese with English abstract).hppt://hhme.cs.tsinghua.edu.cn
[16]Wu Z,Huang Y,Yu Y,Wang L,Tan T.Early hierarchical contexts learned by convolutional networks for image segmentation.In:Proc.of the Int'l Conf.on Pattern Recognition.IEEE,2014.1538-1543.[doi:10.1109/ICPR.2014.273]
[17]Long J,Shelhamer E,Darrell T.Fully convolutional networks for semantic segmentation.In:Proc.of the IEEE Conf.on Computer Vision and Pattern Recognition.2015.3431-3440.[doi:10.1109/CVPR.2015.7298965]
[18]Comaniciu D,Meer P.Mean shift:A robust approach toward feature space analysis.IEEE Trans.on Pattern Analysis and Machine Intelligence,2002,24(5):603-619.[doi:10.1109/34.1000236]
[19]Tao W,Jin H,Zhang Y.Color image segmentation based on mean shift and normalized cuts.IEEE Trans.on Systems,Man,and Cybernetics,Part B(Cybernetics),2007,37(5):1382-1389.[doi:10.1109/TSMCB.2007.902249]
[20]Krizhevsky A,Sutskever I,Hinton GE.ImageNet classification with deep convolutional neural networks.In:Proc.of the Int'l Conf.on Neural Information Processing Systems.Curran Associates Inc.,2012.1097-1105.http://xueshu.baidu.com/s?wd=paperuri:(bfdf67dfdf8cea0c47038f63e91b9df1)&filter=sc_long_sign&sc_ks_para=q%3DImageNet+classification+with+deep+convolutional+neural+networks&tn=SE_baiduxueshu_c1gjeupa&ie=utf-8&sc_us=7479570315911023847
[21]Russakovsky O,Deng J,Su H,Krause J,Satheesh S,Mal S,Huang ZH,Andrej K,Aditya K,Michael B,Alexander C,Li FF.ImageNet large scale visual recognition challenge.Int'l Journal of Computer Vision,2014,115(3):211-252.
[11]杨俊,吕伟涛,马颖,姚雯,李清勇.基于自适应阈值的地基云自动检测方法.应用气象学报,2009,20(6):713-721.
[15]张加楠,毋立芳,贺娇瑜.基于多分辨率卷积神经网络的毫米波雷达云图像分割.见:第25届全国多媒体技术学术会议.2016.hppt://hhme.cs.tsinghua.edu.cn
[2]Long CN,Sabburg JM,CalbóJ,Pagès D.Retrieving cloud characteristics from ground-based daytime color all-sky images.Journal of Atmospheric&Oceanic Technology,2006,23(5):633-652.[doi:10.1175/JTECH1875.1]
[3]Shields JE,Johnson RW,Karr ME,Wertz JL.Automated day/night whole sky imagers for field assessment of cloud cover distributions and radiance distributions.In:Proc.of the 10th Symp.on Meteorological Observations and Instrumentation.1998.http://xueshu.baidu.com/s?wd=paperuri:(c7eaa7569237eec16316553da79c38bd)&filter=sc_long_sign&sc_ks_para=q%3DAutomat ed+day%2Fnight+whole+sky+imagers+for+field+assessment+of+cloud+cover+distributions+and+radiance+distributions&tn=SE_b aiduxueshu_c1gjeupa&ie=utf-8&sc_us=11998081953391704638
[4]Shields JE,Karr ME,Johnson RW,Burden AR.Day/Night whole sky imagers for 24-h cloud and sky assessment:History and overview.Applied Optics,2013,52(8):1605-1616.[doi:10.1364/A0.52.001605]
[5]Long C,Slater D,Tooman T.Total sky imager model 880 status and testing results.Office of Scientific&Technical Information Technical Reports,2001.https://xs.glgoo.net/scholar?hl=zh-CN&q=Total+sky+imager+model+880+status+and+testing+results&btnG=&lr=
[6]Shaw JA,Thurairajah B.Short-Term arctic cloud statistics at NSA from the infrared cloud imager.2003.https://www.researchgate.net/publication/228878636_Short-term_arctic_cloud_statistics_at_NSA_from_the_infrared_cloud_imager
[7]Racette P,Adler RF,Wang JR,Gasiewski AJ,Jakson DM,Zacharias DS.An airborne millimeter-wave imaging radiometer for cloud,precipitation,and atmospheric water vapor studies.Journal of Atmospheric&Oceanic Technology,1992,13(3):1426-1428.[doi:10.1175/1520-0426(1996)013<0610:AAMWIR>2.0.CO;2]
[8]Kreuter A,Zangerl M,Schwarzmann M,Blumthaler M.All-Sky imaging:A simple,versatile system for atmospheric research.Change,2009,48(6):1091-1097.[doi:10.1364/A0.48.001091]
[9]Mantelli Neto SL,Von Wangenheim A,Pereira EB,Comunello E.The use of euclidean geometric distance on RGB color space for the classification of sky and cloud patterns.Journal of Atmospheric&Oceanic Technology,2010,27(9):1504-1517.[doi:10.1175/2010JTECHA1353.1]
[10]Souzaecher MP,Pereira EB,Bins LS,Andrade MAR.A simple method for the assessment of the cloud cover state in high-latitude regions by a ground-based digital camera.Journal of Atmospheric&Oceanic Technology,2006,23(3):437.[doi:10.1175/JTECH1833.1]
[11]Yang J,LüWT,Ma Y,Yao W.An automatic ground-based cloud detection method based on adaptive threshold.Journal of Applied Meteorological Science,2009,20(6):713-721(in Chinese with English absrtact).
[12]Liu S,Zhang L,Zhang Z,Wang C,Xiao B.Automatic cloud detection for all-sky images using superpixel segmentation.IEEE Geoscience&Remote Sensing Letters,2015,12(2):354-358.[doi:10.1109/LGRS.2014.2341291]
[13]Buch KAJ,Sun CH.Cloud classification using whole-sky imager data.In:Proc.of the 9th Symp.on Meteorological Observations&Instrumentation.1995,16(3):353-358.http://xueshu.baidu.com/s?wd=paperuri%3A(9cebe35398fab368b4f3d5382a8e4d6b)&filter=sc_long_sign&tn=SE_baiduxueshu_c1gjeupa&ie=utf-8&sc_ks_para=q%3DCloud%20classification%20using%20whole-sky%20imager%20data
[14]Taravat A,Frate FD,Cornaro C,Vergari S.Neural networks and support vector machine algorithms for automatic cloud classification of whole-sky ground-based images.IEEE Geoscience&Remote Sensing Letters,2015,12(3):666-670.[doi:10.1109/LGRS.2014.2356616]
[15]Zhang JN,Wu LF,He JY.Millimeter-Wave cloud image segmentation based on multi-resolution convolutional neural network.In:Proc.of the National Conf.on Multimedia Technology.2016(in Chinese with English abstract).hppt://hhme.cs.tsinghua.edu.cn
[16]Wu Z,Huang Y,Yu Y,Wang L,Tan T.Early hierarchical contexts learned by convolutional networks for image segmentation.In:Proc.of the Int'l Conf.on Pattern Recognition.IEEE,2014.1538-1543.[doi:10.1109/ICPR.2014.273]
[17]Long J,Shelhamer E,Darrell T.Fully convolutional networks for semantic segmentation.In:Proc.of the IEEE Conf.on Computer Vision and Pattern Recognition.2015.3431-3440.[doi:10.1109/CVPR.2015.7298965]
[18]Comaniciu D,Meer P.Mean shift:A robust approach toward feature space analysis.IEEE Trans.on Pattern Analysis and Machine Intelligence,2002,24(5):603-619.[doi:10.1109/34.1000236]
[19]Tao W,Jin H,Zhang Y.Color image segmentation based on mean shift and normalized cuts.IEEE Trans.on Systems,Man,and Cybernetics,Part B(Cybernetics),2007,37(5):1382-1389.[doi:10.1109/TSMCB.2007.902249]
[20]Krizhevsky A,Sutskever I,Hinton GE.ImageNet classification with deep convolutional neural networks.In:Proc.of the Int'l Conf.on Neural Information Processing Systems.Curran Associates Inc.,2012.1097-1105.http://xueshu.baidu.com/s?wd=paperuri:(bfdf67dfdf8cea0c47038f63e91b9df1)&filter=sc_long_sign&sc_ks_para=q%3DImageNet+classification+with+deep+convolutional+neural+networks&tn=SE_baiduxueshu_c1gjeupa&ie=utf-8&sc_us=7479570315911023847
[21]Russakovsky O,Deng J,Su H,Krause J,Satheesh S,Mal S,Huang ZH,Andrej K,Aditya K,Michael B,Alexander C,Li FF.ImageNet large scale visual recognition challenge.Int'l Journal of Computer Vision,2014,115(3):211-252.
[11]杨俊,吕伟涛,马颖,姚雯,李清勇.基于自适应阈值的地基云自动检测方法.应用气象学报,2009,20(6):713-721.
[15]张加楠,毋立芳,贺娇瑜.基于多分辨率卷积神经网络的毫米波雷达云图像分割.见:第25届全国多媒体技术学术会议.2016.hppt://hhme.cs.tsinghua.edu.cn