基于深度学习特征和支持向量机的遥感图像分类
|
摘要
随着遥感图像采集技术的迅速发展,传统的遥感图像处理方法已经不能满足当前实际的生产需要。近年来,深度学习模型的流行为遥感图像分类问题的解决提供了新的途径。因此,为了进一步提升遥感图像的分类精度,笔者提出了一种基于深度学习特征和支持向量机(support vector machine,SVM)的遥感图像分类模型。首先,针对深度学习模型需要海量训练数据的特点,运用旋转、剪裁等方法对原始的遥感图像进行数据扩增;然后,将扩增数据按照种类随机地分为训练集和验证集两部分测试集,并使用训练集和验证集训练改进的针对遥感图像分类问题的卷积神经网络(convolutional neural network,CNN)中的参数,进而在训练好的CNN模型上提取第一部分测试集的深度学习特征;最后,使用第一部分测试集的深度特征训练多分类SVM,并对第二部分测试集图像进行分类验证。实验采用NWPU-RESISC45公共数据集对本研究模型精度进行验证,与现有的遥感图像分类方法相比较,实验结果表明,提出模型的总体分类精度有明显提升,从而验证了方法的有效性和实用性。
To detect and monitor the feature changes in a specific area,remote sensing can measure the reflected and emitted radiation signals at a long distance from that area. Remote sensing can be utilized in many fields,such as,natural resource detection,agricultural crops and forestry management,mineral exploration,buildings and bridges monitoring,land usage and coverage mapping,and global weather change research. Recently,the remote sensing has been widely applied in the forestry industry economy. With the remarkable progress of the remote sensing image acquisition technology,the traditional remote sensing image classification methods cannot meet the actual needs of the forestry industry economy. In recent years,the emergence and prevalence of the deep learning models have provided many novel ideas for the classification and recognition of remote sensing images. To further improve the classification accuracy of remote sensing images,in this paper,a method of the remote sensing image classification based on the combination of the deep learning feature and support vector machine( SVM) model was proposed. Firstly,because the characteristic of the deep learning models needs the massive training data,the rotation and clipping for augmenting the number of the original remote sensing images were operated. Then,the augmented data were divided into the training set,twofold testing sets and the validation set. The parameters of the convolutional neural network( CNN) were trained by the training set and the validation set,and the features of the first testing samples were extracted based on the trained CNN model. Finally,the deep features of the second testing set were classified using the trained SVM with the first testing data. The method was validated by the NWPU-RESISC45 public data set. Compared with the existing classification methods of remote sensing image,experimental results showed that the overall classification accuracy of the proposed method was improved,which validated the validity and practicability of the proposed method.
To detect and monitor the feature changes in a specific area,remote sensing can measure the reflected and emitted radiation signals at a long distance from that area. Remote sensing can be utilized in many fields,such as,natural resource detection,agricultural crops and forestry management,mineral exploration,buildings and bridges monitoring,land usage and coverage mapping,and global weather change research. Recently,the remote sensing has been widely applied in the forestry industry economy. With the remarkable progress of the remote sensing image acquisition technology,the traditional remote sensing image classification methods cannot meet the actual needs of the forestry industry economy. In recent years,the emergence and prevalence of the deep learning models have provided many novel ideas for the classification and recognition of remote sensing images. To further improve the classification accuracy of remote sensing images,in this paper,a method of the remote sensing image classification based on the combination of the deep learning feature and support vector machine( SVM) model was proposed. Firstly,because the characteristic of the deep learning models needs the massive training data,the rotation and clipping for augmenting the number of the original remote sensing images were operated. Then,the augmented data were divided into the training set,twofold testing sets and the validation set. The parameters of the convolutional neural network( CNN) were trained by the training set and the validation set,and the features of the first testing samples were extracted based on the trained CNN model. Finally,the deep features of the second testing set were classified using the trained SVM with the first testing data. The method was validated by the NWPU-RESISC45 public data set. Compared with the existing classification methods of remote sensing image,experimental results showed that the overall classification accuracy of the proposed method was improved,which validated the validity and practicability of the proposed method.
引文
[1]赵静.基于匹配技术的遥感影像目标识别研究[D].北京:中国科学院大学,2012.ZHAO J.Research on target recognition of remote sensing images with matching technology[D].Beijing:University of the Chinese Academy of Sciences,2012.
[2]任冲.中高分辨率遥感影像森林类型精细分类与森林资源变化监测技术研究[D].北京:中国林业科学研究院,2016.REN C.Forest types precise classification and forest resources change monitoring based on medium and high spatial resolution remote sensing images[D].Beijing:Chinese Academy of Forestry,2016.
[3]丁磊,张保明,郭海涛,等.矢量数据辅助的高分辨率遥感影像道路自动提取[J].遥感学报,2017,21(1):84-95.DOI:10.11834/jrs.20175306.DING L,ZHANG B M,GUO H T,et al.Automatic road extraction from high-resolution remote sensing images assisted by vector data[J].Journal of Remote Sensing,2017,21(1):84-95.
[4]王登峰,杨志刚,魏安世.纹理信息在遥感影像分类中的应用[J].南京林业大学学报(自然科学版),2010,34(3):97-100.DOI:10.3969/j.issn.1000-2006.2010.03.020.WANG D F,YANG Z G,WEI A S.Application of texture information on classification of remote sensing imagery[J].Journal of Nanjing ForestryUniversity(Natural Science Edition),2010,34(3):97-100.
[5]TUIA D,RATLE F,PACIFICI F,et al.Active learning methods for remote sensing image classification[J].IEEE Transactions on Geoscience and Remote Sensing,2009,47(7):2218-2232.DOI:10.1109/tgrs.2008.2010404.
[6]张康,黑保琴,周壮,等.变异系数降维的CNN高光谱遥感图像分类[J].遥感学报,2018,22(1):87-96.ZHANG K,HEI B Q,ZHOU Z,et al.CNN with coefficient of variation-based dimensionality reduction for hyperspectral remote sensing images classification[J].Journal of Remote Sensing,2018,22(1):87-96.
[7]MAGGIORI E,TARABALKA Y,CHARPIAT G,et al.Convolutional neural networks for large-scale remote-sensing image classification[J].IEEE Transactions on Geoscience and Remote Sensing,2017,55(2):645-657.DOI:10.1109/tgrs.2016.2612821.
[8]刘振波,张丽丽,葛云健,等.基于环境卫星数据的森林叶面积指数遥感反演与验证:以大兴安岭加格达奇林区为例[J].林业科技开发,2015,29(4):126-130.DOI:10.13360/j.issn.1000-8101.2015.04.031.LIU Z B,ZHANG L L,GE Y J,et al.Retrieval and validation of forest leaf area index based on HJ satellite data:taking Jiagadaqi District as a case study[J].China Forestry Science and Technology,2015,29(4):126-130.
[9]CHEN Y S,ZHAO X,JIA X P.Spectral-Spatial classification of hyperspectral data based on deep belief network[J].IEEEJournal of Selected Topics in Applied Earth Observations and Remote Sensing,2015,8(6):2381-2392.DOI:10.1109/jstars.2015.2388577.
[10]XING C,MA L,YANG X Q.Stacked denoise autoencoder based feature extraction and classification for hyperspectral images[J].Journal of Sensors,2016,2016:1-10.DOI:10.1155/2016/3632943.
[11]CHEN Y S,JIANG H L,LI C Y,et al.Deep feature extraction and classification of hyperspectral images based on convolutional neural networks[J].IEEE Transactions on Geoscience and Remote Sensing,2016,54(10):6232-6251.DOI:10.1109/tgrs.2016.2584107.
[12]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.DOI:10.3788/aos201636.0428001.LIU D W,HAN L,HAN X Y.High spatial resolution remote sensing image classification based on deep learning[J].Acta OpticaSinica,2016,36(4):0428001.
[13]刘扬,付征叶,郑逢斌.高分辨率遥感影像目标分类与识别研究进展[J].地球信息科学学报,2015,17(9):1080-1091.DOI:10.3724/SP.J.1047.2015.01080.LIU Y,FU Z Y,ZHENG F B.Review on high resolution remote sensing image classification and recognition[J].Journal of GeoInformation Science,2015,17(9):1080-1091.
[14]赵爽.基于卷积神经网络的遥感图像分类方法研究[D].北京:中国地质大学(北京),2015.ZHAO S.Remote sensing image classification method based on convolutional neural networks[D].Beijing:China University of Geosciences,2015.
[15]CHENG G,HAN J,LU X.Remote sensing image scene classification:benchmark and state of the art[J].Proceedings of the IEEE,2017,105(10):1865-1883.DOI:10.1109/JPROC.2017.2675998.
[16]CHENG G,HAN J W,LU X Q.Remote sensing image scene classification:benchmark and state of the art[J].Proceedings of the IEEE,2017,105(10):1865-1883.DOI:10.1109/jproc.2017.2675998.
[17]YU Y L,LIU F X.A two-stream deep fusion framework for highresolution aerial scene classification[J].Computational Intelligence and Neuroscience,2018,2018:1-13.DOI:10.1155/2018/8639367.
[18]WONG S C,GATT A,STAMATESCU V,et al.Understanding data augmentation for classification:when to warp?[C]//International Conference on Digital Image Computing:Techniques and Applications(DICTA),2016.Gold Coast,QLD,Australia,2016:1-6.DOI:10.1109/DICTA.2016.7797091.
[19]SRIVASTAVA N,HINTON G,KRIZHEVSKY A,et al.Dropout:a simple way to prevent neural networks from overfitting[J].The Journal of Machine Learning Research,2014,15(1):1929-1958.
[20]KRIZHEVSKY A,SUTSKEVER I,HINTON G E.ImageNet classification with deep convolutional neural networks[J].Communications of the ACM,2017,60(6):84-90.DOI:10.1145/3065386.
[21]CHEN L C,PAPANDREOU G,KOKKINOS I,et al.Deep lab:Semantic image segmentation with deep convolutional nets,atrous convolution,and fully connected CRFs[J].IEEE transactions on pattern analysis and machine intelligence,2017,40(4):834-848.DOI:10.1109/TPAMI.2017.2699184.
[22]YE Q L,ZHAO C X,YE N,et al.Iterative support vector machine with guaranteed accuracy and run time[J].Expert Systems,2010,27(5):338-348.DOI:10.1111/j.1468-0394.2010.00550.x.
[23]YE Q L,ZHAO C X,YE N,et al.Multi-weight vector projection support vector machines[J].Pattern Recognition Letters,2010,31(13):2006-2011.DOI:10.1016/j.patrec.2010.06.005.
[24]NIE F,HUANG H,DING C,et al.Robust principal component analysis with non-greedy l1-norm maximization[C]//International Joint Conference on Artificial Intelligence,2011.
[25]HE K M,ZHANG X Y,REN S Q,et al.Deep residual learning for image recognition[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition(CVPR).Las Vegas,NV,USA,2016:770-778.DOI:10.1109/CVPR.2016.90.
[26]张冬,严锐,云挺,等.基于地面激光雷达的单木枝干建模方法[J].林业工程学报,2016,1(5):107-114.DOI:10.13360/j.issn.2096-1359.2016.05.019.ZHANG D,YAN R,YUN T,et al.The 3D reconstruction of tree branches from point cloud based on terrestrial laser scanner[J].Journal of Forestry Engineering,2016,1(5):107-114.
[27]ZHANG D,ZHANG L Y,YE Q L,et al.Robust learning-based prediction for timber-volume of living trees[J].Computers and Electronics in Agriculture,2017,136:97-110.DOI:10.1016/j.compag.2017.02.025.
[28]刘柯珍,舒立福,赵凤君,等.基于卫星监测热点的林火分布及发生预报模型[J].林业工程学报,2017,2(4):128-133.DOI:10.13360/j.issn.2096-1359.2017.04.021.LIU K Z,SHU L F,ZHAO F J,et al.Research on spatial distribution of forest fire based on satellite hotspots data and forecasting model[J].China Forestry Science and Technology,2017,2(4):128-133.
[29]LIN S H,CAI L,LIN X M,et al.Masked face detection via a modified LeNet[J].Neurocomputing,2016,218:197-202.DOI:10.1016/j.neucom.2016.08.056.
[2]任冲.中高分辨率遥感影像森林类型精细分类与森林资源变化监测技术研究[D].北京:中国林业科学研究院,2016.REN C.Forest types precise classification and forest resources change monitoring based on medium and high spatial resolution remote sensing images[D].Beijing:Chinese Academy of Forestry,2016.
[3]丁磊,张保明,郭海涛,等.矢量数据辅助的高分辨率遥感影像道路自动提取[J].遥感学报,2017,21(1):84-95.DOI:10.11834/jrs.20175306.DING L,ZHANG B M,GUO H T,et al.Automatic road extraction from high-resolution remote sensing images assisted by vector data[J].Journal of Remote Sensing,2017,21(1):84-95.
[4]王登峰,杨志刚,魏安世.纹理信息在遥感影像分类中的应用[J].南京林业大学学报(自然科学版),2010,34(3):97-100.DOI:10.3969/j.issn.1000-2006.2010.03.020.WANG D F,YANG Z G,WEI A S.Application of texture information on classification of remote sensing imagery[J].Journal of Nanjing ForestryUniversity(Natural Science Edition),2010,34(3):97-100.
[5]TUIA D,RATLE F,PACIFICI F,et al.Active learning methods for remote sensing image classification[J].IEEE Transactions on Geoscience and Remote Sensing,2009,47(7):2218-2232.DOI:10.1109/tgrs.2008.2010404.
[6]张康,黑保琴,周壮,等.变异系数降维的CNN高光谱遥感图像分类[J].遥感学报,2018,22(1):87-96.ZHANG K,HEI B Q,ZHOU Z,et al.CNN with coefficient of variation-based dimensionality reduction for hyperspectral remote sensing images classification[J].Journal of Remote Sensing,2018,22(1):87-96.
[7]MAGGIORI E,TARABALKA Y,CHARPIAT G,et al.Convolutional neural networks for large-scale remote-sensing image classification[J].IEEE Transactions on Geoscience and Remote Sensing,2017,55(2):645-657.DOI:10.1109/tgrs.2016.2612821.
[8]刘振波,张丽丽,葛云健,等.基于环境卫星数据的森林叶面积指数遥感反演与验证:以大兴安岭加格达奇林区为例[J].林业科技开发,2015,29(4):126-130.DOI:10.13360/j.issn.1000-8101.2015.04.031.LIU Z B,ZHANG L L,GE Y J,et al.Retrieval and validation of forest leaf area index based on HJ satellite data:taking Jiagadaqi District as a case study[J].China Forestry Science and Technology,2015,29(4):126-130.
[9]CHEN Y S,ZHAO X,JIA X P.Spectral-Spatial classification of hyperspectral data based on deep belief network[J].IEEEJournal of Selected Topics in Applied Earth Observations and Remote Sensing,2015,8(6):2381-2392.DOI:10.1109/jstars.2015.2388577.
[10]XING C,MA L,YANG X Q.Stacked denoise autoencoder based feature extraction and classification for hyperspectral images[J].Journal of Sensors,2016,2016:1-10.DOI:10.1155/2016/3632943.
[11]CHEN Y S,JIANG H L,LI C Y,et al.Deep feature extraction and classification of hyperspectral images based on convolutional neural networks[J].IEEE Transactions on Geoscience and Remote Sensing,2016,54(10):6232-6251.DOI:10.1109/tgrs.2016.2584107.
[12]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.DOI:10.3788/aos201636.0428001.LIU D W,HAN L,HAN X Y.High spatial resolution remote sensing image classification based on deep learning[J].Acta OpticaSinica,2016,36(4):0428001.
[13]刘扬,付征叶,郑逢斌.高分辨率遥感影像目标分类与识别研究进展[J].地球信息科学学报,2015,17(9):1080-1091.DOI:10.3724/SP.J.1047.2015.01080.LIU Y,FU Z Y,ZHENG F B.Review on high resolution remote sensing image classification and recognition[J].Journal of GeoInformation Science,2015,17(9):1080-1091.
[14]赵爽.基于卷积神经网络的遥感图像分类方法研究[D].北京:中国地质大学(北京),2015.ZHAO S.Remote sensing image classification method based on convolutional neural networks[D].Beijing:China University of Geosciences,2015.
[15]CHENG G,HAN J,LU X.Remote sensing image scene classification:benchmark and state of the art[J].Proceedings of the IEEE,2017,105(10):1865-1883.DOI:10.1109/JPROC.2017.2675998.
[16]CHENG G,HAN J W,LU X Q.Remote sensing image scene classification:benchmark and state of the art[J].Proceedings of the IEEE,2017,105(10):1865-1883.DOI:10.1109/jproc.2017.2675998.
[17]YU Y L,LIU F X.A two-stream deep fusion framework for highresolution aerial scene classification[J].Computational Intelligence and Neuroscience,2018,2018:1-13.DOI:10.1155/2018/8639367.
[18]WONG S C,GATT A,STAMATESCU V,et al.Understanding data augmentation for classification:when to warp?[C]//International Conference on Digital Image Computing:Techniques and Applications(DICTA),2016.Gold Coast,QLD,Australia,2016:1-6.DOI:10.1109/DICTA.2016.7797091.
[19]SRIVASTAVA N,HINTON G,KRIZHEVSKY A,et al.Dropout:a simple way to prevent neural networks from overfitting[J].The Journal of Machine Learning Research,2014,15(1):1929-1958.
[20]KRIZHEVSKY A,SUTSKEVER I,HINTON G E.ImageNet classification with deep convolutional neural networks[J].Communications of the ACM,2017,60(6):84-90.DOI:10.1145/3065386.
[21]CHEN L C,PAPANDREOU G,KOKKINOS I,et al.Deep lab:Semantic image segmentation with deep convolutional nets,atrous convolution,and fully connected CRFs[J].IEEE transactions on pattern analysis and machine intelligence,2017,40(4):834-848.DOI:10.1109/TPAMI.2017.2699184.
[22]YE Q L,ZHAO C X,YE N,et al.Iterative support vector machine with guaranteed accuracy and run time[J].Expert Systems,2010,27(5):338-348.DOI:10.1111/j.1468-0394.2010.00550.x.
[23]YE Q L,ZHAO C X,YE N,et al.Multi-weight vector projection support vector machines[J].Pattern Recognition Letters,2010,31(13):2006-2011.DOI:10.1016/j.patrec.2010.06.005.
[24]NIE F,HUANG H,DING C,et al.Robust principal component analysis with non-greedy l1-norm maximization[C]//International Joint Conference on Artificial Intelligence,2011.
[25]HE K M,ZHANG X Y,REN S Q,et al.Deep residual learning for image recognition[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition(CVPR).Las Vegas,NV,USA,2016:770-778.DOI:10.1109/CVPR.2016.90.
[26]张冬,严锐,云挺,等.基于地面激光雷达的单木枝干建模方法[J].林业工程学报,2016,1(5):107-114.DOI:10.13360/j.issn.2096-1359.2016.05.019.ZHANG D,YAN R,YUN T,et al.The 3D reconstruction of tree branches from point cloud based on terrestrial laser scanner[J].Journal of Forestry Engineering,2016,1(5):107-114.
[27]ZHANG D,ZHANG L Y,YE Q L,et al.Robust learning-based prediction for timber-volume of living trees[J].Computers and Electronics in Agriculture,2017,136:97-110.DOI:10.1016/j.compag.2017.02.025.
[28]刘柯珍,舒立福,赵凤君,等.基于卫星监测热点的林火分布及发生预报模型[J].林业工程学报,2017,2(4):128-133.DOI:10.13360/j.issn.2096-1359.2017.04.021.LIU K Z,SHU L F,ZHAO F J,et al.Research on spatial distribution of forest fire based on satellite hotspots data and forecasting model[J].China Forestry Science and Technology,2017,2(4):128-133.
[29]LIN S H,CAI L,LIN X M,et al.Masked face detection via a modified LeNet[J].Neurocomputing,2016,218:197-202.DOI:10.1016/j.neucom.2016.08.056.