邻域信息嵌入的半监督判别字典对学习的遥感场景图像分类方法

Semi-supervised Discrimination Dictionary Pair Learning Based on Neighborhood Information Embedding for Remote Sensing Scene Image Classification

摘要

随着遥感成像技术的发展，雷达或卫星等地球观测仪器生成大量遥感场景图像. 但在训练样本标签不充足的情况下，遥感场景图像的分类的准确率仍有待提高. 本文提出了一种邻域信息嵌入的半监督判别字典对学习方法. 利用标记和未标记的样本进行综合字典和分析字典对的训练，通过最小化基于稀疏 ℓ_2,1范数的重构误差项来得到稳健的字典对. 对结构化分析字典施加稀疏约束，以确保编码的稀疏性. 同时，基于稀疏编码的邻域结构信息建立局部边界项，使得同类近邻稀疏编码间的距离较小，而不同类近邻稀疏编码间的距离大，提高稀疏编码的判别性. 在 RSSCN7 和 UCMerced 数据集上的半监督遥感场景图像分类的实验表明，本文方法在分类准确率上具有一定优势.

Abstract

With the development of remote sensing imaging technology, a large number of remote sensing scene images are generated by earth observation instruments such as radar or satellite. However, the classification of remote sensing scene images is still an open problem when the labels of training samples are insufficient. To solve this problem, a semi-supervised discrimination dictionary pair learning based on neighborhood information embedding is proposed for remote sensing scene image classification. The labeled and unlabeled samples are used to train the comprehensive dictionary and the analytical dictionary pair, and the robust dictionary pair is obtained by minimizing the reconstruction error term based on the sparse ℓ_2,1 norm. A sparse constraint is imposed on the structured analysis dictionary to ensure the sparsity of the encoding. At the same time, the local boundary term is established based on the neighborhood structure information of sparse coding, so that the distance between similar neighbor sparse codes is small, and the distance between different neighbor sparse codes is large, which improves the discrimination of sparse coding. Experiments on semi-supervised remote sensing scene image classification in RSSCN7 and UCMerced datasets show that the method proposed in this paper has certain advantages in classification accuracy.

关键词

半监督学习 / 图像分类 / 遥感场景 / 字典对学习 / 稀疏编码

Key words

semi-supervised learning / image classification / remote sensing scene / dictionary pair learning / sparse coding

参考文献

[1] 韩文军，孙小虎，吉根林，等. 基于卷积神经网络的多光谱与全色遥感图像融合算法[J].南京师大学报(自然科学版)，2021，44(3):123-130.

[2] 丁凯孟，朱长青，罗文，等. 基于自适应 PCNN 与 PCA 的遥感影像感知哈希认证算法[J].南京师大学报(自然科学版)，2019，42(2):17-29.

[3] Yang Y, Newsam S. Bag—of—visual—words and spatial extensions for land—use classification[C] //Proceedings of 18th SIGSPATIAL International Conference on Advances in Geographic Information systems. San Jose, CA, USA, 2010:270-279.

[4] Bachmann C M, Ainsworth T L, Fusina R A. Exploiting manifold geometry in hyperspectral imagery[J].IEEE Trans Geoscience & Remote Sensing, 2005, 43(3):441-454.

[5] Zhou W, Newsam S, Li C, et al. Pattern Net: A benchmark dataset for performance evaluation of remote sensing image retrieval[J].ISPRS Journal of Photogrammetry and Remote Sensing, 2018, 145(11):197-209.

[6] Liu Q, Cheng Y, Cao K, et al. Radar forward—looking imaging for complex targets based on sparse representation with dictionary learning[J].IEEE Geoscience and Remote Sensing Letters, 2022, 19:4026605.

[7] Xing L, Zhao L, Cao W, et al. Class shared dictionary learning for few—shot remote sensing scene classification[J].IEEE Geoscience and Remote Sensing Letters, 2022, 19:6512805.

[8] Liu T, Gu Y, Jia X. Class—guided coupled dictionary learning for multispectral—hyperspectral remote sensing image collaborative classification[J]. Science China Technological Sciences, 2022, 65:744—758.

[9] Huang Z, Chen H X, Zhou T, et al. Contrast—weighted dictionary learning based saliency detection for VHR optical remote sensing images[J].Pattern Recognition, 2021, 113(5):107757.

[10] Wang J, Jiang Y, Qi Y, et al. Remote sensing image fusion based on nonnegative dictionary learning[J].IEEE Access, 2021, 9(11):158908-158916.

[11] 白坤，慕晓冬，陈雪冰，等. 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.

[12] Xia J, Chanussot J, Du P, et al. (Semi—)supervised probabilistic principal component analysis for hyperspectral remote sensing image classification[J].IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2014, 7(6):2224-2236.

[13] Zheng C, Zhang Y, Wang L. Semantic segmentation of remote sensing imagery using an object—based Markov random field model with auxiliary label fields[J].IEEE Transactions on Geoscience and Remote Sensing, 2017, 55(5):3015-3028.

[14] Wang S, Yang S, Jiao L. Saliency—guided change detection for SAR imagery using a semi—supervised Laplacian SVM[J].Remote Sensing Letters, 2016, 7(11):1043-1052.

[15] Wang J, Zuo R, Xiong Y. Mapping mineral prospectivity via semi—supervised random forest[J].Natural Resources Research, 2020, 29(1):189-202.

[16] Shao S, Xu R, Wang Z, et al. DLDL: Dynamic label dictionary learning via hypergraph regularization[J].Neurocomputing, 2022, 475(2):80-88.

[17] Sun Y, Zhang Z, Jiang W, et al. Discriminative local sparse representation by robust adaptive dictionary pair learning[J].IEEE Transactions on Neural Networks and Learning Systems, 2020, 31(10):4303-4317.

[18] Dornaika F, El Traboulsi Y. Learning flexible graph—based semi—supervised embedding[J].IEEE Transactions on Cybernetics, 2015, 46(1):206-218.

[19] Belkin M, Niyogi P, Sindhwani V. Manifold regularization: a geometric framework for learning from labeled and unlabeled examples[J].Journal of Machine Learning Research, 2006, 7(1):2399-2434.

[20] Chen H, Ye M, Lei L, et al. Semisupervised dual—dictionary learning for heterogeneous transfer learning on cross—scene hyperspectral images[J].IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2020, 13:3164-3178.

[21] El Traboulsi Y, Dornaika F, Assoum A. Kernel flexible manifold embedding for pattern classification[J].Neurocomputing, 2015, 167(11):517-527.

[22] El Traboulsi Y, Dornaika F. Flexible semi—supervised embedding based on adaptive loss regression: Application to image categorization[J].Information Sciences, 2018, 444(5):1-19.

[23] Zhu R, Dornaika F, Ruichek Y. Joint graph based embedding and feature weighting for image classification[J].Pattern Recognition, 2019, 93(9):458-469.

[24] Duong V, Bui M, Wang J. Semi—supervised subspace learning via constrained matrix factorization[C] //2021 3rd International Conference on Sustainable Technologies for Industry 4.0(STI), 2021:1-4.

[25] Xia G S, Hu J, Hu F, et al. AID: a benchmark dataset for performance evaluation of aerial scene classification[J].IEEE Transactions on Geoscience and Remote Sensing. D.haka, Barglades h, 2017, 55(7):3965-3981.

基金资助

国家自然科学基金资助项目(61806026)

江苏省教育厅未来网络科研基金资助项目(FNSRFP-2021-YB-36)

常州市科技支撑社会发展资助项目(CE20215032)