Lightweight deep global-local knowledge distillation network for hyperspectral image scene classification

被引：0
|
作者：
Liu Y. [1 ]
Pu C. [1 ]
Xu D. [2 ]
Yang Y. [1 ]
Huang H. [1 ]
机构：
[1] Key Laboratory of Optoelectronic Technology and Systems, the Education Ministry of China, Chongqing University, Chongqing
[2] Measurement and Control Technology and Instrument major, College of Optoelectronic Engineering, Chongqing University, Chongqing
来源：
Guangxue Jingmi Gongcheng/Optics and Precision Engineering | 2023年 / 31卷 / 17期
关键词：
benchmark dataset; feature extraction; hyperspectral scene classification; knowledge distillation; vision transformer;
D O I：
10.37188/OPE.20233117.2598
中图分类号：
学科分类号：
摘要：
To address the challenges of the complex spatial layouts of target scenes and inherent spatial-spectral information redundancy of HSIs， an end-to-end lightweight deep global-local knowledge distillation （LDGLKD） method is proposed herein. To explore the global sequence properties of spatial-spectral features， the vision transformer （ViT） is used as the teacher to guide the lightweight student model for HSI scene classification. In LDGLKD， pre-trained VGG16 is selected as the student model to extract local detail information. After collaborative training of ViT and VGG16 through knowledge distillation， the teacher model transmits the learned long-range contextual information to the small-scale student model. By combining the advantages of the two models through knowledge distillation， the optimal classification accuracy of LDGLKD on the Orbita HSI scene classification dataset （OHID-SC） and hyperspectral remote sensing dataset for scene classification reached 91.62% and 97.96%， respectively. The experimental results revealed that the proposed LDGLKD method presented good classification performance. In addition， the OHID-SC based on the remote sensing data obtained by the Orbita Zhuhai-1 satellite could reflect the detailed information of land cover and provide data support for HSI scene classification. © 2023 Chinese Academy of Sciences. All rights reserved.
引用
收藏
页码：2598 / 2610
页数：12
相关论文
共 24 条
  • [1] YAN J W, CHEN H D, LIU L., Overview of hyperspectral image classification ［J］, Opt. Precision Eng, 27, 3, pp. 680-693, (2019)
  • [2] DING G S, QIAO Y L, YI W N, Et al., Improved SIFT feature extraction and matching technology based on hyperspectral image ［J］, Opt. Precision Eng, 28, 4, pp. 954-962, (2020)
  • [3] CHEN J, ZHANG J., Spectral-spatial joint classification of hyperspectral image algorithm based on improved Gaussian process regression ［J］, Opt. Precision Eng, 27, 7, pp. 1649-1660, (2019)
  • [4] CAI Y M, LIU X B, CAI Z H., BS-nets： an end-to-end framework for band selection of hyperspectral image ［J］, IEEE Transactions on Geoscience and Remote Sensing, 58, 3, pp. 1969-1984, (2020)
  • [5] HU F, XIA G S, YANG W, Et al., Mining deep semantic representations for scene classification of high-resolution remote sensing imagery ［J］, IEEE Transactions on Big Data, 6, 3, pp. 522-536, (2019)
  • [6] LUO F L, DU B, ZHANG L P, Et al., Feature learning using spatial-spectral hypergraph discriminant analysis for hyperspectral image ［J］, IEEE Transactions on Cybernetics, 49, 7, pp. 2406-2419, (2019)
  • [7] TONG X Y, XIA G S, LU Q, Et al., Land-cover classification with high-resolution remote sensing images using transferable deep models ［J］, Remote Sensing of Environment, 237, (2020)
  • [8] YU Y T, LI Y Y, WANG J, Et al., C²-CapsViT： cross-context and cross-scale capsule vision transformers for remote sensing image scene classification ［J］, IEEE Geoscience and Remote Sensing Letters, 19, pp. 1-5, (2022)
  • [9] ZOU Q, NI L, ZHANG T, Et al., Deep learning based feature selection for remote sensing scene classification ［J］, IEEE Geoscience and Remote Sensing Letters, 12, 11, pp. 2321-2325, (2015)
  • [10] Bi Q, QIN K, LI Z, Et al., A multiple-instance densely-connected ConvNet for aerial scene classification ［J］, IEEE Transactions on Image Processing, 9, pp. 4911-4926, (2020)
123