利用微调卷积神经网络的土地利用场景分类
|
摘要
针对场景类别之间的相同类内差异性与不同类间相似性所造成的遥感图像场景分类不够精确的问题,提出了将微调(fine-tuning)与卷积神经网络(convolutional neural network,CNN)模型相结合的方法,对土地利用遥感场景图像进行分类。该方法对CNN前层固定,调整分类层,保留了图像的泛性特征;通过卫星影像图获取土地利用场景图块作为训练样本,对训练样本图块进行预处理,然后对在ImageNet数据集上训练得到的AlexNet模型进行fine-tuning,利用得到的CNN模型即可自动提取土地利用遥感图像的图像特征并对其进行分类。为了验证本文方法,对实验区影像进行分割得到测试样本并进行同训练样本一致的预处理,将测试样本的分类结果与随机森林、支持向量机等经典方法的结果进行对比。结果表明,经过fine-tuning的CNN模型在土地利用分类中得到的结果要明显优于其他分类方法。
In view of the inaccuracy problem in classification of remote sensing images that are caused by differences within the same class and similarity among the different classes,a method combining fine-tuning with the convolution neural work to classify the remote sensing scene images of land use is proposed.In this method,the Convolutional Neural Network(CNN)front layers are fixed,the classification layer is adjusted,the features of the image are preserved,the land use scene blocks are obtained by satellite image map as the training sample,the training sample blocks are preprocessed,and then the AlexNet model trained on the ImageNet data set is entered into the Fine-tuning,and the modified model is used to classification of land-use scenarios.It can automatically extract and classify the land-use remote sensing images.In order to verify this method,the experimentation area images are segmented to get the test samples and are pretreated with the same training samples.The results of the test samples are compared with the classical methods such as the random forest and the support vector machine.The results show that the fine-tuning CNN model is superior to other classification methods in land use classification.
In view of the inaccuracy problem in classification of remote sensing images that are caused by differences within the same class and similarity among the different classes,a method combining fine-tuning with the convolution neural work to classify the remote sensing scene images of land use is proposed.In this method,the Convolutional Neural Network(CNN)front layers are fixed,the classification layer is adjusted,the features of the image are preserved,the land use scene blocks are obtained by satellite image map as the training sample,the training sample blocks are preprocessed,and then the AlexNet model trained on the ImageNet data set is entered into the Fine-tuning,and the modified model is used to classification of land-use scenarios.It can automatically extract and classify the land-use remote sensing images.In order to verify this method,the experimentation area images are segmented to get the test samples and are pretreated with the same training samples.The results of the test samples are compared with the classical methods such as the random forest and the support vector machine.The results show that the fine-tuning CNN model is superior to other classification methods in land use classification.
引文
[1]温一博,范文义.多时相遥感数据森林类型识别技术研究[J].森林工程,2013,29(2):14-20.
[2]张锦水,潘耀忠,韩立建,等.光谱与纹理信息复合的土地利用/覆盖变化动态监测研究[J].遥感学报,2007,11(4):500-510.
[3]王瑞,李杰平,卢志刚,等.基于人工神经网络遥感图像分类的应用研究[J].科技情报开发与经济,2011,21(3):108-110,135.
[4]卢柳叶,李光录,张莉,等.基于TM影像的半干旱区土地利用信息提取[J].干旱地区农业研究,2012,30(1):217-223.
[5]杨勇,任志远.基于GIS的关中地区土地利用/覆盖变化对比研究[J].干旱区资源与环境,2013,27(5):40-45.
[6]KRIZHEVSKY A,SUTSKEVER I,HINTON G E.ImageNet classification with deep convolutional neural networks[C]//International Conference on Neural Information Processing Systems.Curran Associates Inc.2012:1097-1105.
[7]ZEILER M D,FERGUS R.Visualizing and understanding convolutional networks[C]//European Conference on Computer Vision.Springer,Cham,2014:818-833.
[8]SIMONYAN K,ZISSERMAN A.Very deep convolutional networks for large-scale image recognition[J].Computer Science,2014.
[9]HINTON G E,SALAKHUTDINOV R R.Reducing the dimensionality of data with neural networks[J].Science,2006,313(5786):504-507.
[10]RUSSAKOVSKY O,DENG J,SU H,et al.ImageNet large scale visual recognition challenge[J].International Journal of Computer Vision,2014,115(3):211-252.
[11]KRIZHEVSKY A,SUTSKEVER I,HINTON G E.ImageNet classification with deep convolutional neural networks[C]//International Conference on Neural Information Processing Systems.Curran Associates Inc.2012:1097-1105.
[12]SERMANET P,EIGEN D,ZHANG X,et al.OverFeat:integrated recognition,localization and detection using convolutional networks[J].Eprint Arxiv,2013.
[13]SZEGEDY C,LIU W,JIA Y,et al.Going deeper with convolutions[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.USA:IEEE,2015:1-9.
[14]DAN C,MEIER U,SCHMIDHUBER J.Multi-column deep neural networks for image classification supplementary online material[J].Eprint Arxiv,2012,157(10):3642-3649.
[15]RAZAVIAN A S,AZIZPOUR H,SULLIVAN J,et al.CNN features off-the-shelf:an astounding baseline for recognition[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops.USA:IEEE,2014:806-813.
[16]何小飞,邹峥嵘,陶超,等.联合显著性和多层卷积神经网络的高分影像场景分类[J].测绘学报,2016,45(9):1073-1080.
[17]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,2015.
[2]张锦水,潘耀忠,韩立建,等.光谱与纹理信息复合的土地利用/覆盖变化动态监测研究[J].遥感学报,2007,11(4):500-510.
[3]王瑞,李杰平,卢志刚,等.基于人工神经网络遥感图像分类的应用研究[J].科技情报开发与经济,2011,21(3):108-110,135.
[4]卢柳叶,李光录,张莉,等.基于TM影像的半干旱区土地利用信息提取[J].干旱地区农业研究,2012,30(1):217-223.
[5]杨勇,任志远.基于GIS的关中地区土地利用/覆盖变化对比研究[J].干旱区资源与环境,2013,27(5):40-45.
[6]KRIZHEVSKY A,SUTSKEVER I,HINTON G E.ImageNet classification with deep convolutional neural networks[C]//International Conference on Neural Information Processing Systems.Curran Associates Inc.2012:1097-1105.
[7]ZEILER M D,FERGUS R.Visualizing and understanding convolutional networks[C]//European Conference on Computer Vision.Springer,Cham,2014:818-833.
[8]SIMONYAN K,ZISSERMAN A.Very deep convolutional networks for large-scale image recognition[J].Computer Science,2014.
[9]HINTON G E,SALAKHUTDINOV R R.Reducing the dimensionality of data with neural networks[J].Science,2006,313(5786):504-507.
[10]RUSSAKOVSKY O,DENG J,SU H,et al.ImageNet large scale visual recognition challenge[J].International Journal of Computer Vision,2014,115(3):211-252.
[11]KRIZHEVSKY A,SUTSKEVER I,HINTON G E.ImageNet classification with deep convolutional neural networks[C]//International Conference on Neural Information Processing Systems.Curran Associates Inc.2012:1097-1105.
[12]SERMANET P,EIGEN D,ZHANG X,et al.OverFeat:integrated recognition,localization and detection using convolutional networks[J].Eprint Arxiv,2013.
[13]SZEGEDY C,LIU W,JIA Y,et al.Going deeper with convolutions[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.USA:IEEE,2015:1-9.
[14]DAN C,MEIER U,SCHMIDHUBER J.Multi-column deep neural networks for image classification supplementary online material[J].Eprint Arxiv,2012,157(10):3642-3649.
[15]RAZAVIAN A S,AZIZPOUR H,SULLIVAN J,et al.CNN features off-the-shelf:an astounding baseline for recognition[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops.USA:IEEE,2014:806-813.
[16]何小飞,邹峥嵘,陶超,等.联合显著性和多层卷积神经网络的高分影像场景分类[J].测绘学报,2016,45(9):1073-1080.
[17]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,2015.