Unsupervised domain adaptive classification for hyperspectral remote sensing by adversary coupled with distillation
- Vol. 28, Issue 1, Pages: 231-246(2024)
Published: 07 January 2024
DOI: 10.11834/jrs.20232580
扫 描 看 全 文
浏览全部资源
扫码关注微信
Published: 07 January 2024 ,
扫 描 看 全 文
于纯妍,徐铭阳,宋梅萍,胡亚斌,张建祎.2024.对抗与蒸馏耦合的高光谱遥感域自适应分类方法.遥感学报,28(1): 231-246
Yu C Y,Xu M Y,Song M P,Hu Y B and Chang C I. 2024. Unsupervised domain adaptive classification for hyperspectral remote sensing by adversary coupled with distillation. National Remote Sensing Bulletin, 28(1):231-246
高光谱遥感域自适应分类旨在利用有标注样本的源域知识对无标注的目标域场景进行分类,是高光谱跨场景分类的重要方法之一。目前流行的域自适应分类方法利用对抗训练模式实现目标域与源域的特征对齐,但未考虑源域知识是否充分转移至目标域这一关键问题。为了有效提取并迁移源域知识,本文提出一种基于对抗与蒸馏耦合模式的高光谱遥感自适应分类方法UDAACD(Unsupervised Domain Adaptation by Adversary Coupled with Distillation)。该方法采用类内样本自蒸馏方式对源域信息进行提炼,提高自适应分类模型对源域监督知识的提取能力;同时,构建知识蒸馏与对抗耦合机制使目标域与源域特征在对抗与蒸馏中实现对齐,利用对抗与蒸馏耦合机制相互补充、相互促进,提升高光谱遥感知识从源域至目标域的迁移能力,进而完成目标域高光谱影像的无监督分类。本文选用Pavia University、Pavia Center、Houston 2013及Houston 2018高光谱遥感场景数据集进行了4组跨场景图像分类实验,结果表明所提出的模型优于其他高光谱域自适应方法,在相同样本条件下取得了较高的分类精度,准确率分别为91.75%(Pavia University->Pavia Center)、74.41%(Pavia Center->Pavia University)、70.68%(Houston 2013->Houston 2018)及67.76%(Houston 2018->Houston 2013),验证了方法的鲁棒性。
Unsupervised Domain Adaptive (UDA) classification aims to categorize the target domain scenes without labeled samples using knowledge from the source domain data with the labeled samples. Thus
UDA classification is one of the important cross-scene classification methods in the field of Hyperspectral Image Classification (HSIC). The existing domain adaptive classification methods for hyperspectral remote sensing data mainly utilize the adversarial training mode to achieve the feature alignment between the target and source domains. The popular UDA approach with the local alignment condition of dual domains generates acceptable classification accuracy. However
the key issue of the sufficient transfer of the source domain knowledge to the target domain is not considered. An unsupervised domain adaptation classification by adversary coupled with distillation is proposed in this paper for the unsupervised HSIC to effectively extract and transfer source domain knowledge. In the proposed framework
the dense-base network with convolutional block attention module is presented to extract abundant features for the representation of the source and target domain categories. In the source domain training
a self-distillation learning schema is adopted to reduce the class-wised difference by matching the predictive distribution of the same class samples. The self-distillation regularization constraint is increased between the samples of the same category in the source domain to reduce the intraclass difference of the classification subspace and improve the knowledge expression accuracy of the source domain classification model. Thus
the capability of the adaptive classification model to refine the source domain supervision knowledge is improved. In addition
a novel mechanism of adversarial training coupled with distillation knowledge is presented to guarantee the complete transfer of source domain knowledge to the target domain scene with feature alignment. Moreover
dual classifiers are employed in the adversarial training process to eliminate the prediction effect of the confused samples. The maximum and minimum discrepancies of the dual classifiers during the adversarial training rapidly promote the feature alignment without confusion. Thus
knowledge distillation is conducted to improve the recognition capability of the network in the domain while ensuring the complete transfer of hyperspectral source domain knowledge in the feature alignment process to improve the knowledge acquisition capability of the model in the target domain. Finally
the unsupervised classification of HSIs in the target domain is completed after the knowledge transfer. The experiments for HSI cross-scene image classification are conducted on four hyperspectral remote sensing scene datasets
including Pavia University
Pavia Center
Houston 2013
and Houston 2018. Results demonstrate that the proposed model is superior to other hyperspectral domain adaptive methods. Under the same sample conditions
the classification accuracy achieves 91.75% (Pavia University to Pavia Center)
74.41% (Pavia Center to Pavia University)
70.68% (Houston 2013 to Houston 2018)
and 67.76% (Houston 2018 to Houston 2013). In addition
the ablation study illustrates that the final classification accuracy of the unsupervised HSIC is improved with the self-distillation and the distillation loss in the adversarial training model. The parameters with different weights and temperatures are analyzed in the experiments with variations of values. The validity of the method is verified by all the mentioned experimental results and analyses.
高光谱遥感图像分类域自适应知识蒸馏生成对抗网络
hyperspectral remote sensingimage classificationdomain adaptationknowledge distillationgenerative adversarial networks
Cao X H, Ren M R, Zhao J, Li H and Jiao L C. 2020. Hyperspectral imagery classification based on compressed convolutional neural network. IEEE Geoscience and Remote Sensing Letters, 17(9): 1583-1587 [DOI: 10.1109/LGRS.2019.2951372http://dx.doi.org/10.1109/LGRS.2019.2951372]
Du P J, Xia J S, Xue Z H, Tan K, Su H J and Bao R. 2016. Review of hyperspectral remote sensing image classification. Journal of Remote Sensing, 20(2): 236-256
杜培军, 夏俊士, 薛朝辉, 谭琨, 苏红军, 鲍蕊. 2016. 高光谱遥感影像分类研究进展. 遥感学报, 20(2): 236-256 [DOI: 10.11834/jrs.20165022http://dx.doi.org/10.11834/jrs.20165022]
Fang Z Q, Yang Y X, Li Z K, Li W, Chen Y S, Ma L and Du Q. 2022. Confident learning-based domain adaptation for hyperspectral image classification. IEEE Transactions on Geoscience and Remote Sensing, 60: 5527116 [DOI: 10.1109/TGRS.2022.3166817http://dx.doi.org/10.1109/TGRS.2022.3166817]
Ganin Y, Ustinova E, Ajakan H, Germain P, Larochelle H, Laviolette F, Marchand M and Lempitsky V. 2016. Domain-adversarial training of neural networks. The Journal of Machine Learning Research, 17(1): 2096-2030
Ghamisi P, Dalla Mura M and Benediktsson J A. 2015. A survey on spectral-spatial classification techniques based on attribute profiles. IEEE Transactions on Geoscience and Remote Sensing, 53(5): 2335-2353 [DOI: 10.1109/TGRS.2014.2358934http://dx.doi.org/10.1109/TGRS.2014.2358934]
Gong T F, Zheng X T and Lu X Q. 2022. Meta self-supervised learning for distribution shifted few-shot scene classification. IEEE Geoscience and Remote Sensing Letters, 19: 6510005 [DOI: 10.1109/LGRS.2022.3174277http://dx.doi.org/10.1109/LGRS.2022.3174277]
Goodfellow I J, Pouget-Abadie J, Mirza M, Xu B, Warde-Farley D, Ozair S, Courville A and Bengio Y. 2014. Generative adversarial nets//Proceedings of the 27th International Conference on Neural Information Processing Systems. Montreal: MIT Press: 2672-2680
Hinton G, Vinyals O and Dean J. 2015. Distilling the knowledge in a neural network. arXiv:1503.02531
Hu L Q, Kan M N, Shan S G and Chen X L. 2020. Unsupervised domain adaptation with hierarchical gradient synchronization//Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle: IEEE: 4042-4051 [DOI: 10.1109/CVPR42600.2020.00410http://dx.doi.org/10.1109/CVPR42600.2020.00410]
Liu X F, Liu J M and Fu M. 2022. Adversarial network samples were generated for hyperspectral image classification. Electronic Measurement Technology, 45(3): 146-152
刘雪峰, 刘佳明, 付民. 2022. 生成对抗网络扩充样本用于高光谱图像分类. 电子测量技术, 45(3): 146-152 [DOI: 10.19651/j.cnki.emt.2108021http://dx.doi.org/10.19651/j.cnki.emt.2108021]
Liu Z X, Ma L and Du Q. 2021. Class-wise distribution adaptation for unsupervised classification of hyperspectral remote sensing images. IEEE Transactions on Geoscience and Remote Sensing, 59(1): 508-521 [DOI: 10.1109/TGRS.2020.2997863http://dx.doi.org/10.1109/TGRS.2020.2997863]
Lu X Q, Gong T F and Zheng X T. 2020. Multisource compensation network for remote sensing cross-domain scene classification. IEEE Transactions on Geoscience and Remote Sensing, 58(4): 2504-2515 [DOI: 10.1109/TGRS.2019.2951779http://dx.doi.org/10.1109/TGRS.2019.2951779]
Lu X Q, Zheng X T and Yuan Y. 2017. Remote sensing scene classification by unsupervised representation learning. IEEE Transactions on Geoscience and Remote Sensing, 55(9): 5148-5157 [DOI: 10.1109/TGRS.2017.2702596http://dx.doi.org/10.1109/TGRS.2017.2702596]
Ma X R, Mou X R, Wang J, Liu X K, Geng J and Wang H Y. 2021. Cross-dataset hyperspectral image classification based on adversarial domain adaptation. IEEE Transactions on Geoscience and Remote Sensing, 59(5): 4179-4190 [DOI: 10.1109/TGRS.2020.3015357http://dx.doi.org/10.1109/TGRS.2020.3015357]
Pan S J, Tsang I W, Kwok J T and Yang Q. 2011. Domain adaptation via transfer component analysis. IEEE Transactions on Neural Networks, 22(2): 199-210 [DOI: 10.1109/TNN.2010.2091281http://dx.doi.org/10.1109/TNN.2010.2091281]
Saito K, Watanabe K, Ushiku Y and Harada T. 2018. Maximum classifier discrepancy for unsupervised domain adaptation//Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City: IEEE: 3723-3732 [DOI: 10.1109/CVPR.2018.00392http://dx.doi.org/10.1109/CVPR.2018.00392]
Shen D and Ma L. 2019. Cross-domain extreme learning machine for classification of hyperspectral images//Proceedings of the 2019 IEEE International Geoscience and Remote Sensing Symposium. Yokohama: IEEE: 3305-3308 [DOI: 10.1109/IGARSS.2019.8898437http://dx.doi.org/10.1109/IGARSS.2019.8898437]
Shen J, Qu Y R, Zhang W N and Yu Y. 2018. Wasserstein distance guided representation learning for domain adaptation//Proceedings of the 32nd AAAI Conference on Artificial Intelligence and Thirtieth Innovative Applications of Artificial Intelligence Conference and Eighth AAAI Symposium on Educational Advances in Artificial Intelligence. New Orleans: AAAI Press: 4058-4065
Tzeng E, Hoffman J, Saenko K and Darrell T. 2017. Adversarial discriminative domain adaptation//Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. Honolulu: IEEE: 2962-2971 [DOI: 10.1109/CVPR.2017.316http://dx.doi.org/10.1109/CVPR.2017.316]
Wang D, Gao F, Dong J Y, Yang Y and Wang S K. 2018. Sea ice classification from hyperspectral images based on self-paced boost learning//Proceedings of the 2018 IEEE International Geoscience and Remote Sensing Symposium. Valencia: IEEE: 7324-7327 [DOI: 10.1109/IGARSS.2018.8518452http://dx.doi.org/10.1109/IGARSS.2018.8518452]
Woo S, Park J, Lee J Y and Kweon I S. 2018. CBAM: convolutional block attention module//Proceedings of the 15th European Conference on Computer Vision. Munich: Springer: 3-19 [DOI: 10.1007/978-3-030-01234-2_1http://dx.doi.org/10.1007/978-3-030-01234-2_1]
Xia J S, Yokoya N and Iwasaki A. 2017. Ensemble of transfer component analysis for domain adaptation in hyperspectral remote sensing image classification//Proceedings of the 2017 IEEE International Geoscience and Remote Sensing Symposium. Fort Worth: IEEE: 4762-4765 [DOI: 10.1109/IGARSS.2017.8128066http://dx.doi.org/10.1109/IGARSS.2017.8128066]
Xue Z H and Zhang Y J. 2022. Hyperspectral image classification method based on supervised hashing with RBF kernel and convolution. National Remote Sensing Bulletin, 26(4): 722-738
薛朝辉, 张瑜娟. 2022. 基于卷积核哈希学习的高光谱图像分类. 遥感学报, 26(4): 722-738 [DOI: 10.11834/jrs.20220359http://dx.doi.org/10.11834/jrs.20220359]
Yu C Y, Liu C Y, Song M P and Chang C I. 2022. Unsupervised domain adaptation with content-wise alignment for hyperspectral imagery classification. IEEE Geoscience and Remote Sensing Letters, 19: 5511705 [DOI: 10.1109/LGRS.2021.3126594http://dx.doi.org/10.1109/LGRS.2021.3126594]
Yue J, Fang L Y, Rahmani H and Ghamisi P. 2022. Self-supervised learning with adaptive distillation for hyperspectral image classification. IEEE Transactions on Geoscience and Remote Sensing, 60: 5501813 [DOI: 10.1109/TGRS.2021.3057768http://dx.doi.org/10.1109/TGRS.2021.3057768]
Yun S, Park J, Lee K and Shin J. 2020. Regularizing class-wise predictions via self-knowledge distillation//Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle: IEEE: 13873-13882 [DOI: 10.1109/CVPR42600.2020.01389http://dx.doi.org/10.1109/CVPR42600.2020.01389]
Zhan Y, Hu D, Wang Y T and Yu X C. 2018. Semisupervised hyperspectral image classification based on generative adversarial networks. IEEE Geoscience and Remote Sensing Letters, 15(2): 212-216 [DOI: 10.1109/LGRS.2017.2780890http://dx.doi.org/10.1109/LGRS.2017.2780890]
Zhang J and Bao W X. 2022. Research on classification method of hyperspectral remote sensing image based on Generative Adversarial Network. National Remote Sensing Bulletin, 26(2): 416-430
张健, 保文星. 2022. 生成式对抗网络的高光谱遥感图像分类方法研究. 遥感学报, 26(2): 416-430 [DOI: 10.11834/jrs.20219192http://dx.doi.org/10.11834/jrs.20219192]
Zhang Y, Xiang T, Hospedales T M and Lu H C. 2018. Deep mutual learning//Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City: IEEE: 4320-4328 [DOI: 10.1109/CVPR.2018.00454http://dx.doi.org/10.1109/CVPR.2018.00454]
Zhao Q Z, Jiang P, Wang X W, Zhang L H and Zhang J X. 2021. Classification of protection forest tree species based on UAV hyperspectral data. Transactions of the Chinese Society for Agricultural Machinery, 52(11): 190-199
赵庆展, 江萍, 王学文, 张丽红, 张建新. 2021. 基于无人机高光谱遥感影像的防护林树种分类. 农业机械学报, 52(11): 190-199 [DOI: 10.6041/j.issn.1000-1298.2021.11.020http://dx.doi.org/10.6041/j.issn.1000-1298.2021.11.020]
相关作者
相关机构