Acta Photonica Sinica, Volume. 50, Issue 4, 254(2021)

Hyperspectral Anomaly Detection Based on 3D Convolutional Autoencoder and Low Rank Representation

Bangyong SUN1...2, Zhe ZHAO1, Bingliang HU2 and Tao YU2,* |Show fewer author(s)
Author Affiliations
  • 1College of Printing, Packaging and Digital Media, Xi'an University of Technology, Xi'an70048, China
  • 2Key Laboratory of Spectral Imaging Technology CAS, Xi'an Institute of Optics and Precision Mechanics, Chinese Academy of Sciences, Xi'an710119, China
  • show less
    Figures & Tables(17)
    Framework of the proposed method
    The AVIRIS-1 dataset
    The AVIRIS-2 dataset
    The AUC value of two datasets under
    The AUC value of two datasets under
    The AUC value of two datasets under different eps and MinSample when p equal 8
    The AUC value of two datasets under different eps and MinSample when p equal 9
    The AUC value of two datasets under different eps and MinSample when p equal 10
    The AUC value of two datasets under different eps and MinSample when p equal 11
    The AUC value of two datasets under different eps and MinSample when p equal 12
    The AUC value of two datasets under different p when eps equal 0.012 and MinSample equal 10
    Detection results of various detection algorithms in AVIRIS-1 dataset
    Detection results of various detection algorithms in AVIRIS-2 dataset
    ROC curves of two datasets
    • Table 1. The Architecture of 3D-CAE

      View table
      View in Article

      Table 1. The Architecture of 3D-CAE

      BlockLayerInput sizeKernel sizeStridesOutput size
      Encoder

      Conv1

      BN+LReLu

      189×5×5×1

      -

      1×3×3

      -

      1×1×1

      -

      189×3×3×12

      189×3×3×12

      Conv2

      BN+LReLu

      189×3×3×12

      -

      3×1×1

      -

      3×1×1

      -

      63×3×3×24

      63×3×3×24

      Conv3

      BN+LReLu

      63×3×3×24

      -

      1×3×3

      -

      1×1×1

      -

      63×1×1×36

      63×1×1×36

      Conv4

      BN+Sigmoid

      63×1×1×36

      -

      3×1×1

      -

      3×1×1

      -

      21×1×1×48

      21×1×1×48

      Decoder

      Deconv1

      BN+LReLu

      21×1×1×48

      -

      3×1×1

      -

      3×1×1

      -

      63×1×1×36

      63×1×1×36

      Deconv2

      BN+LReLu

      63×1×1×36

      -

      1×3×3

      -

      1×1×1

      -

      63×3×3×24

      63×3×3×24

      Deconv3

      BN+LReLu

      63×3×3×24

      -

      3×1×1

      -

      3×1×1

      -

      189×3×3×12

      189×3×3×12

      Deconv4

      BN

      189×3×3×12

      -

      1×3×3

      -

      1×1×1

      -

      189×5×5×1

      -

    • Table 2. The AUC value of different methods in two datasets

      View table
      View in Article

      Table 2. The AUC value of different methods in two datasets

      DatasetRXWWLRX4CRDLRASRDAEADRC-LRaSMDProposed

      AVIRIS-1

      AVIRIS-2

      0.911 10.944 50.972 50.989 60.977 40.989 90.993 2
      0.940 30.967 50.951 20.909 60.957 30.990 10.991 4
    • Table 3. Analysis of the results of different steps of the Proposed algorithm

      View table
      View in Article

      Table 3. Analysis of the results of different steps of the Proposed algorithm

      MethodAVIRIS-1(AUC)AVIRIS-2(AUC)
      3D-Conv+RX0.954 00.962 2
      3D-Conv+LRR0.991 60.988 9
      3D-Conv+LRR(K-means)+RE0.990 80.990 1
      3D-Conv(MSE) +LRR+RE0.977 20.977 4
      3D-Conv +LRR+RE0.993 20.991 4
    Tools

    Get Citation

    Copy Citation Text

    Bangyong SUN, Zhe ZHAO, Bingliang HU, Tao YU. Hyperspectral Anomaly Detection Based on 3D Convolutional Autoencoder and Low Rank Representation[J]. Acta Photonica Sinica, 2021, 50(4): 254

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category: Image Processing

    Received: Dec. 16, 2020

    Accepted: Jan. 4, 2021

    Published Online: May. 11, 2021

    The Author Email: YU Tao (yutao@opt.ac.cn)

    DOI:10.3788/gzxb20215004.0410003

    Topics