高分影像场景分类的半监督深度卷积神经网络学习方法

doi:10.11947/j.AGCS.2021.20200017

摘要/Abstract

摘要： 传统基于深度卷积神经网络的场景分类方法往往需要大量标记样本用于模型的参数训练，在标记训练集数量有限的情况下，学习得到的特征泛化能力降低。针对这一问题，本文提出了高分影像分类的半监督深度卷积神经网络学习方法（3sCNN），采用自学习半监督策略，训练阶段不断增加训练样本：首先，通过有限的标记数据对深度网络进行初步训练；然后，利用经过初步训练的网络对未标记数据进行预测，得到未标记样本的预测标签及其对应的置信度；最后，将具有高置信度的未标记样本作为真实标记数据加入到训练集中，继续对网络进行训练并重复上述过程。为验证算法的有效性，本文在3个常用数据集上进行试验，试验结果证明本文算法可以有效提高有限样本下高分影像场景分类精度。

关键词: 卷积神经网络, 高分影像, 半监督, 分类

Abstract: The large amount of labeled dataset is always required to train the deep convolutional neural networks (CNNs) for high spatial resolution (HSR) imagery scene classification. However, the generalization of the learned deep features is decayed when limited labeled data is available. To solve this problem, the scene classification based on semi-supervised CNNs (3sCNN) is proposed. In the proposed method, the labeled data is first used to train the model and then the prediction label and confidence of the unlabeled data is obtained with the trained model. Finally, the unlabeled data with high confidence is considered as the labeled data to train the network again and the progress described above is repeated. To demonstrate the effectiveness of the proposed method, the experiments on three datasets are performed. The results show that the proposed method can effectively improve the classification.

Key words: convolutional neural networks, high spatial resolution imagery, semi-supervised, classification

中图分类号:

P227

杨秋莲, 刘艳飞, 丁乐乐, 孟凡效. 高分影像场景分类的半监督深度卷积神经网络学习方法[J]. 测绘学报, 2021, 50(7): 930-938.

YANG Qiulian, LIU Yanfei, DING Lele, MENG Fanxiao. High spatial resolution imagery scene classification based on semi-supervised CNNs[J]. Acta Geodaetica et Cartographica Sinica, 2021, 50(7): 930-938.

参考文献

[1] 杜培军. 高分辨率卫星遥感的发展及其在矿山的应用[J]. 2001, 10(1): 5-7. DU Peijun. Development of high-resolution satellite remote sensing and its application in mines[J]. Coal, 2001, 10(1): 5-7.
[2] 曹福成. 高分系列遥感卫星布设中国太空“慧眼”: 我国高分专项建设回眸[J]. 中国军转民, 2015(1): 28-33. CAO Fucheng. Gaofen series remote sensing satellite China’s space “eye”:review of China’s special construction of high marks [J]. Defense Industry Conversion in China, 2015(1): 28-33.
[3] 明冬萍, 骆剑承, 沈占锋, 等. 高分辨率遥感影像信息提取与目标识别技术研究[J]. 测绘科学, 2005, 30(3): 18-20, 3. MING Dongping, LUO Jiancheng, SHEN Zhanfeng, et al. Research on information extraction and target recognition from high resolution remote sensing image[J]. Science of Surveying and Mapping, 2005, 30(3): 18-20, 3.
[4] 周成虎, 骆剑承. 高分辨率卫星遥感影像地学计算[M]. 北京: 科学出版社, 2009. ZHOU Chenghu, LUO Jiancheng.Geologic calculation of high resolution satellite remote sensing image[M]. Beijing: Science Press, 2009.
[5] 张乐飞. 遥感影像的张量表达与流形学习方法研究[D]. 武汉大学, 2013. ZHANG Lefei.Tensor representation and manifold learning methods for remote sensing images[D].Wuhan: Wuhan University, 2013.
[6] 赵济. 面向高分辨率遥感影像分类的条件随机场模型研究[D]. 武汉: 武汉大学, 2017. ZHAO Ji. Conditional random fields for high resolution remote sensing image classification[D]. Wuhan: Wuhan University, 2017.
[7] 孙显, 付琨, 王宏琦. 高分辨率遥感图像理解[M]. 北京: 科学出版社, 2011. SUN Xian, FU Kun, WANG Hongqi. High resolution remote sensing image understanding[M]. Beijing: Science Press, 2011.
[8] 张帆. 面向高分辨率遥感影像分析的深度学习方法研究[D]. 武汉: 武汉大学, 2017. ZHANG Fan. Deep learning for very high resolution remote sensing data analysis[D]. Wuhan: Wuhan University, 2017.
[10] 朱祺琪. 面向高分辨率遥感影像场景语义理解的概率主题模型研究[D]. 武汉: 武汉大学, 2018. ZHU Qiqi. Topic model for high resolution remote sensing imagery semantic scene understanding[D]. Wuhan: Wuhan University, 2018.
[11] ZHONG Y, WU S, ZHAO B. Scene semantic understanding based on the spatial context relations of multiple objects[J]. Remote Sensing, 2017, 9(10):1030.
[12] ZHANG Xiuyuan, DU Shihong, ZHANG Yuan. Semantic and spatial co-occurrence analysis on object pairs for urban scene classification[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2018, 11(8): 2630-2643.
[13] ZHU Qiqi, ZHONG Yanfei, ZHANG Liangpei, et al. Scene classification based on the fully sparse semantic topic model[J]. IEEE Transactions on Geoscience and Remote Sensing, 2017, 55(10): 5525-5538.
[14] ZHAO Bei, ZHONG Yanfei, XIA Guisong, et al. Dirichlet-derived multiple topic scene classification model for high spatial resolution remote sensing imagery[J]. IEEE Transactions on Geoscience and Remote Sensing, 2016, 54(4): 2108-2123.
[15] XIA Guisong, HU Jingwen, HU Fan, et al. AID: a benchmark data set for performance evaluation of aerial scene classification[J]. IEEE Transactions on Geoscience and Remote Sensing, 2017, 55(7): 3965-3981.
[16] GONG Zhiqiang, ZHONG Ping, YU Yang, et al. Diversity-promoting deep structural metric learning for remote sensing scene classification[J]. IEEE Transactions on Geoscience and Remote Sensing, 2018, 56(1): 371-390.
[17] ZHANG Wei, TANG Ping, ZHAO Lijun. Remote sensing image scene classification using CNN-CapsNet[J]. Remote Sensing, 2019, 11(5): 494.
[18] LIU Qingshan, HANG Renlong, SONG Huihui, et al. Learning multiscale deep features for high-resolution satellite image scene classification[J]. IEEE Transactions on Geoscience and Remote Sensing, 2018, 56(1): 117-126.
[19] CHENG Gong, HAN Junwei, LU Xiaoqiang. Remote sensing image scene classification: benchmark and state of the art[J]. Proceedings of the IEEE, 2017, 105(10): 1865-1883.
[20] NOGUEIRA K, PENATTI O A B, DOS SANTOS J A. Towards better exploiting convolutional neural networks for remote sensing scene classification[J]. Pattern Recognition, 2017, 61: 539-556.
[21] ZHAO Lijun, TANG Ping, HUO Lianzhi. A 2-D wavelet decomposition-based bag-of-visual-words model for land-use scene classification[J]. International Journal of Remote Sensing, 2014, 35(6): 2296-2310.
[22] CHEN Shizhi, TIAN Yingli. Pyramid of spatial relations for scene-level land use classification[J]. IEEE Transactions on Geoscience and Remote Sensing, 2015, 53(4): 1947-1957.
[23] ZHANG Fan, DU Bo, ZHANG Liangpei. Saliency-guided unsupervised feature learning for scene classification[J]. IEEE Transactions on Geoscience and Remote Sensing, 2015, 53(4): 2175-2184.
[24] HAN Xiaobing, ZHONG Yanfei, ZHAO Bei, et al. Scene classification based on a hierarchical convolutional sparse auto-encoder for high spatial resolution imagery[J]. International Journal of Remote Sensing, 2017, 38(2): 514-536.
[25] ZHONG Yanfei, FEI Feng, LIU Yanfei, et al. SatCNN: satellite image dataset classification using agile convolutional neural networks[J]. Remote Sensing Letters, 2017, 8(2): 136-145.
[26] ZHANG Fan, DU Bo, ZHANG Liangpei. Scene classification via a gradient boosting random convolutional network framework[J]. IEEE Transactions on Geoscience and Remote Sensing, 2016, 54(3): 1793-1802.
[27] WANG Qi, LIU Shaoteng, CHANUSSOT J, et al. Scene classification with recurrent attention of VHR remote sensing images[J]. IEEE Transactions on Geoscience and Remote Sensing, 2019, 57(2): 1155-1167.
[28] CHENG Gong, YANG Ceyuan, YAO Xiwen, et al. When deep learning meets metric learning: remote sensing image scene classification via learning discriminative CNNs[J]. IEEE Transactions on Geoscience and Remote Sensing, 2018, 56(5): 2811-2821.
[29] ZHONG Yanfei, FEI Feng, ZHANG Liangpei. Large patch convolutional neural networks for the scene classification of high spatial resolution imagery[J]. Journal of Applied Remote Sensing, 2016, 10(2):025006.
[30] GONG Zhiqiang, ZHONG Ping, HU Weidong, et al. An end-to-end joint unsupervised learning of deep model and pseudo-classes for remote sensing scene representation[C]//Proceedings of 2019 International Joint Conference on Neural Networks (IJCNN). Budapest, Hungary. IEEE, 2019: 1-7.
[31] ZHANG Jun, LIU Jiao, PAN Bin, et al. Domain adaptation based on correlation subspace dynamic distribution alignment for remote sensing image scene classification[J]. IEEE Transactions on Geoscience and Remote Sensing, 2020, 58(11): 7920-7930.
[32] SONG Shaoyue, YU Hongkai, MIAO Zhenjiang, et al. Domain adaptation for convolutional neural networks-based remote sensing scene classification[J]. IEEE Geoscience and Remote Sensing Letters, 2019, 16(8): 1324-1328.
[33] OTHMAN E, BAZI Y, MELGANI F, et al. Domain adaptation network for cross-scene classification[J]. IEEE Transactions on Geoscience and Remote Sensing, 2017, 55(8): 4441-4456.
[34] ZHANG Jun, ZHANG Min, PAN Bin, et al. Semisupervised center loss for remote sensing image scene classification[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2020, 13: 1362-1373.
[35] YAN Peiyao, HE Feng, YANG Yajie, et al. Semi-supervised representation learning for remote sensing image classification based on generative adversarial networks[J]. IEEE Access, 2020, 8: 54135-54144.
[36] DAI Xueyuan, WU Xiaofeng, WANG Bin, et al. Semisupervised scene classification for remote sensing images: a method based on convolutional neural networks and ensemble learning[J]. IEEE Geoscience and Remote Sensing Letters, 2019, 16(6): 869-873.
[37] HAN Wei, FENG Ruyi, WANG Lizhe, et al. A semi-supervised generative framework with deep learning features for high-resolution remote sensing image scene classification[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2018, 145: 23-43.
[38] HE Kaiming, ZHANG Xiangyu, REN Shaoqing, et al. Deep residual learning for image recognition[C]//Proceedings of 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Las Vegas, NV, USA. IEEE, 2016: 770-778.
[39] YANG Yi, NEWSAM S. Bag-of-visual-words and spatial extensions for land-use classification[C]//Proceedings of the 18th SIGSPATIAL International Conference on Advances in Geographic Information Systems. New York: ACM Press, 2010: 270-279.

[1]	白坤, 慕晓冬, 陈雪冰, 朱永清, 尤轩昂. 融合半监督学习的无监督遥感影像场景分类[J]. 测绘学报, 2022, 51(5): 691-702.
[2]	王丹菂, 邢帅, 徐青, 林雨准, 李鹏程. 单频机载激光测深海陆回波自动分类方法[J]. 测绘学报, 2022, 51(5): 750-761.
[3]	李凡, 金绍华, 边刚, 崔杨, 汤寓麟, 张永厚. 多波束测深误差改进模型构建与验证[J]. 测绘学报, 2022, 51(5): 762-771.
[4]	张玉鑫, 颜青松, 邓非. 高分辨率遥感影像建筑物提取多路径RSU网络法[J]. 测绘学报, 2022, 51(1): 135-144.
[5]	方莉娜, 王爽, 赵志远, 付化胜, 陈崇成. 车载激光点云中交通标线自动分类与矢量化[J]. 测绘学报, 2021, 50(9): 1251-1265.
[6]	纪雪, 唐秋华, 陈义兰, 李杰, 丁德秋. 联合支持向量机和增强学习算法的多波束声学底质分类[J]. 测绘学报, 2021, 50(7): 972-981.
[7]	董卫华, 王圣凯, 王雪元, 杨天宇. 地图线状要素眼动识别的朴素贝叶斯方法[J]. 测绘学报, 2021, 50(6): 749-756.
[8]	王思, 王光霞, 田江鹏. 面向位置聚合的泛在地图信息分类模型[J]. 测绘学报, 2021, 50(6): 789-799.
[9]	叶发茂, 孟祥龙, 董萌, 聂运菊, 葛芸, 陈晓勇. 遥感图像蚁群算法和加权图像到类距离检索法[J]. 测绘学报, 2021, 50(5): 612-620.
[10]	张传辉, 姚国标, 张力, 艾海滨, 满孝成, 黄鹏飞. 多分支网络联合的倾斜影像仿射不变特征匹配[J]. 测绘学报, 2021, 50(5): 641-651.
[11]	施慧慧, 徐雁南, 滕文秀, 王妮. 高分辨率遥感影像深度迁移可变形卷积的场景分类法[J]. 测绘学报, 2021, 50(5): 652-663.
[12]	张红, 吴智伟, 王继成, 高培超. 高光谱图像分类的Wasserstein配置熵非监督波段选择方法[J]. 测绘学报, 2021, 50(3): 405-415.
[13]	蓝朝桢, 卢万杰, 于君明, 徐青. 异源遥感影像特征匹配的深度学习算法[J]. 测绘学报, 2021, 50(2): 189-202.
[14]	汤寓麟, 金绍华, 边刚, 张永厚, 李凡. 侧扫声呐识别沉船影像的迁移学习卷积神经网络法[J]. 测绘学报, 2021, 50(2): 260-269.
[15]	方莉娜, 沈贵熙, 游志龙, 郭迎亚, 付化胜, 赵志远, 陈崇成. 融合点云和多视图的车载激光点云路侧多目标识别[J]. 测绘学报, 2021, 50(11): 1558-1573.