Complex scene classification of remote sensing images based on CNN
Kang ZHANG1,2,3, Baoqin HEI1,2, Shengyang LI1,2, Yuyang SHAO1,2
1. Technology and Engineering Center for Space Utilization, Chinese Academy of Sciences, Beijing 100094, China 2. Key Laboratory of Space Utilization, Chinese Academy of Sciences, Beijing 100094, China 3. University of Chinese Academy of Sciences, Beijing 100049, China
复杂场景分类对于挖掘遥感图像中的价值信息具有重要意义。针对于遥感图像的复杂场景分类,提出了一种基于卷积神经网络(convolutional neural network,CNN)模型的分类方法,在该方法中构建了8层CNN网络结构,并对输入图像进行预处理操作以进一步增强模型的适应性,且在模型分类器的选择问题上提供了Softmax和支持向量机2种分类器,使其能够自动化提取特征,避免了前期繁琐的图像处理和人工提取特征等过程。在UC Merced Land Use和Google of SIRI-WHU这2组数据集中进行实验,结果表明,相比于CNN with Overfeat feature和SRSCNN方法,该模型提高了2%以上的分类精度,且2种分类器的总体分类精度均能达到95%以上。
Complex scene classification has great significance for mining the value information in remote sensing images. The proposed convolutional neural networks (CNN) can improve the complex scene classification of remote sensing images. The CNN method extracts features automatically, avoiding problems in the image pretreatment and the feature extraction by manual labor. An eight-layer CNN model is constructed in this paper, and the pre-treatment module has enhanced the adaptability of this method. Given the problem in choosing classifiers, this paper provides the Softmax and support vector machine (SVM) in the presented CNN. The experiment results in two datasets, the UC Merced Land Use and the Google of SIRI-WHU indicate that the presented CNN method can increase the accuracy of classification by more than 2% compared with the CNN with Overfeat feature method and the SRSCNN method, and the total classification accuracy of the two classifiers is over 95%.
Liu Z Y . Remote sensing land use change monitoring based on special chart restriction[J]. Journal of Anhui Agricultural Sciences, 2011.
[2]
Voogt J A, Oke T R . Thermal remote sensing of urban climates[J]. Remote Sensing of Environment, 2003,86(3):370-384.
doi: 10.1016/S0034-4257(03)00079-8
Xu S H, Mu X D, Ke B , et al. Dynamic monitoring of military position based on remote sensing image[J]. Remote Sensing Technology and Application, 2014,29(3):511-516.
Wang Z Z, Zhang Q J, Lyu K . A brief analysis of challenges and opportunities of space remote sensing system in the big data age[J].Journal of Engineering Studies, 2014(3):244-258
[5]
Duda R O, Hart P E, Stork D G . Pattern Classification[M]. New York:Wiley, 1998: 55-88.
[6]
Li Q, Bao W, Li X, et al. High resolution remote sensing image classification based on SVM and FCM [C]// International Conference on Electrical,Computer Engineering and Electronics. 2015.
Deng Z, Li D, Ke Y H , et al. An improved SVM algorithm for high spatial resolution remote sensing image classification[J]. Remote Sensing for Land and Resources, 2016,28(3):12-18.doi: 10.6046/gtzyyg.2016.03.03.
Yan J N, Zhou K F, Wang J L , et al. Extraction of hyper-spectral remote sensing alteration information based on SAM and SVM[J]. Computer Engineering and Applications, 2013,49(19):141-146.
[9]
Wu M, Chen L. Image recognition based on deep learning [C]// Chinese Automation Congress.IEEE, 2015.
Cao L L, Li H T, Han Y S , et al. Application of convolutional neural networks in classification of high resolution remote sensing imagery[J]. Science of Surveying and Mapping, 2016,41(9):170-175.
[11]
Hinton G E, Salakhutdinov R R . Reducing the dimensionality of data with neural networks[J]. Science, 2006,313(5786):504.
doi: 10.1126/science.1127647
Rui T, Fei J C, Zhou Y , et al. Pedestrian detection based on deep convolutional neural network[J]. Computer Engineering and Applications, 2016,52(13):162-166.
[13]
Zhang Q, Xu J, Xu L , et al. Deep convolutional neural networks for forest fire detection[C]//2016 International Forum on Management,Education and Information Technology Application.Atlantis Press, 2016.
[14]
Zhang R, Yao J, Zhang K , et al.S-CNN ship detection from high-resolution remote sensing images[J].ISPRS, 2016, B7:423-430.
[15]
Krizhevsky A, Sutskever I, Hinton G E . ImageNet classification with deep convolutional neural networks[J]. Advances in Neural Information Processing Systems, 2012,25(2):2012.
doi: 10.1145/3065386
[16]
Girshick R, Donahue J, Darrell T , et al. Rich feature hierarchies for accurate object detection and semantic segmentation[J]. Computer Science, 2014: 580-587.
doi: 10.1109/CVPR.2014.81
[17]
Yosinski J, Clune J, Bengio Y , et al. How transferable are features in deep neural networks?[J]. Eprint Arxiv, 2014,27:3320-3328.
[18]
UC Merced Land Use Dataset[DB/OL].( 2010- 10- 28)[2017-16]. .
[19]
Zhao B, Zhong Y, Xia G S , et al. Dirichlet-derived multiple topic scene classification model fusing heterogeneous features for high resolution remote sensing imagery[J]. IEEE Transactions on Geoscience and Remote Sensing, 2015,54(4):2108-2123.
[20]
Liu Y, Zhong Y, Fei F, et al. Scene semantic classification based on random-scale stretched convolutional neural network for high-spatial resolution remote sensing imagery[C]// IGARSS 2016-2016 IEEE International Geoscience and Remote Sensing Symposium.IEEE, 2016: 763-766.
Xu S H, Mu X D, Zhao P , et al. Scene classification of remote sensing image based on multi-scale feature and deep neural network[J]. Acta Geodaetica et Cartographica Sinica, 2016,45(7):834-840.
[22]
Yang Y, Newsam S. Spatial pyramid co-occurrence for image classification [C]// International Conference on Computer Vision.IEEE, 2011: 1465-1472.
[23]
Marmanis D, Datcu M, Esch T , et al. Deep learning earth observation classification using imagenet pretrained networks[J]. IEEE Geoscience and Remote Sensing Letters, 2015,13(1):1-5.
doi: 10.1109/LGRS.2015.2499239
2018, Vol. 30 
