Journal of Applied Optics, Volume. 46, Issue 2, 336(2025)

Hyperspectral image band selection method for camouflage target recognition

Beibei WANG1,2, Kaixin LIU1,2, and Ping CHEN1,2、*
Author Affiliations
  • 1State Key Laboratory for Electronic Testing Technology, North University of China, Taiyuan 030051, China
  • 2Department of Information and Communication Engineering, North University of China, Taiyuan 030051, China
  • show less
    Figures & Tables(14)
    Schematic diagram of principle of spectral gradient angle
    Real image and ground truth map of camouflaged vehicle dataset
    Real image and ground truth map of San Diego airport dataset
    Overall accuracy of each method at different bands in camouflage vehicle dataset
    Overall accuracy of each method at different bands in San Diego airport dataset
    Classification results of camouflage vehicle images of each method
    Classification results of San Diego airport images of each method
    • Table 1. Band selection algorithm based on SDI model

      View table
      View in Article

      Table 1. Band selection algorithm based on SDI model

      算法:SDI波段选择算法
      输入:高光谱图像X,真值图g,波段选择数量n
      输出:特征波段子集{b1,b2,,bn}
      1. 遍历真值图g,得到类别为伪装目标的所有像素坐标集合w以及类别为背景的所有像素坐标集合v
      2. 对图像X进行高斯滤波,计算X中对应坐标集合wv内所有像素位置上的光谱矢量的平均值xy
      3. 对所有波段进行均匀子空间划分,得到n个子空间{s1,s2,,sn}
      4. fori=1ton:
      5.   遍历si,计算si内的波段数量m
      6.  fork=1tom:
      7.   通过式(7)计算第k个波段的光谱梯度角θSGA(xk,yk)
      8.   通过式(8)计算第k个波段的Fréchet距离F(xk,yk)
      9.   通过式(9)计算第k个波段的皮尔逊相关系数r(xk,yk)
      10.   通过式(10)计算第k个波段的光谱差异度指数ISD(xk,yk)
      11.  endfor
      12.  将si内光谱差异度指数最大的波段序号值赋给bi
      13. endfor
    • Table 2. San Diego hyperspectral public dataset

      View table
      View in Article

      Table 2. San Diego hyperspectral public dataset

      采集地点传感器空间 大小/像素波段 数量/个波长 范围/nm空间 分辨率/m
      美国圣地 亚哥机场机载AVIRIS 传感器400×400224370~25103.5
    • Table 3. Results of ablation experiments in camouflage vehicle dataset

      View table
      View in Article

      Table 3. Results of ablation experiments in camouflage vehicle dataset

      θSGAFrOA/%
      90.51
      91.06
      91.07
      95.06
    • Table 4. Results of ablation experiments in San Diego airport dataset

      View table
      View in Article

      Table 4. Results of ablation experiments in San Diego airport dataset

      θSGAFrOA/%
      97.70
      97.75
      97.75
      98.65
    • Table 5. Performance of different methods on camouflage vehicle dataset

      View table
      View in Article

      Table 5. Performance of different methods on camouflage vehicle dataset

      方法ASNRAREAIE
      SDI43.2312.525.29
      OCF34.556.012.36
      BS_Net45.744.295.17
      GA36.5210.935.12
      SAM38.2112.495.08
      F38.3112.455.34
    • Table 6. Performance of different methods on San Diego airport dataset

      View table
      View in Article

      Table 6. Performance of different methods on San Diego airport dataset

      方法ASNRAREAIE
      SDI51.0512.694.95
      OCF50.1812.575.12
      BS_Net52.1310.254.55
      GA49.7910.134.28
      SAM51.6512.564.43
      F50.938.994.71
    • Table 7. Running time of different methods s

      View table
      View in Article

      Table 7. Running time of different methods s

      数据集SDIOCFBS_NetGASAMF
      Camouflage vehicle3.2611.57304.3265.494.684.35
      San Diego1.271.6823.2567.341.361.53
    Tools

    Get Citation

    Copy Citation Text

    Beibei WANG, Kaixin LIU, Ping CHEN. Hyperspectral image band selection method for camouflage target recognition[J]. Journal of Applied Optics, 2025, 46(2): 336

    Download Citation

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

    Category:

    Received: Aug. 28, 2024

    Accepted: --

    Published Online: May. 13, 2025

    The Author Email: Ping CHEN (陈平)

    DOI:10.5768/JAO202546.0202006

    Topics