Laser & Optoelectronics Progress, Volume. 56, Issue 2, 021001(2019)

Laser Range-Gated Imaging Target Recognition Based on Convolutional Neural Network

Shuyu Wang*, Shengxiang Tao, Fan Yang, and Lei Ai
Author Affiliations
  • Army Artillery Air Defense Academy of PLA, Hefei, Anhui 230031, China
  • show less
    Figures & Tables(8)
    Working principle of laser rang-gated imaging
    Structural diagram of laser rang-gated target recognition by KFCNN
    Partial atlas of laser rang-gated targets
    Flow chart of KFCNN training
    Feature histograms. (a) Body samples; (b) histograms corresponding to VGG16; (c) histograms corresponding to KFCNN
    Average recognition rates of different algorithms
    • Table 1. Average recognition rates of different algorithms for human images with different blur scales%

      View table

      Table 1. Average recognition rates of different algorithms for human images with different blur scales%

      Blur scale12345678
      Algorithm in Ref. [3]72.0575.7274.0173.5178.5079.1574.3374.58
      Algorithm in Ref. [4]83.5589.4382.6085.6787.2582.3386.7089.25
      VGG1690.2694.2292.3789.1686.4781.7177.8374.32
      KFCNN without fine-tuning92.3493.4195.0291.7894.3494.0291.9892.56
      KFCNN with fine-tuning98.9198.5397.6596.7595.1295.4791.5796.94
    • Table 2. Real-time computing time of different algorithms

      View table

      Table 2. Real-time computing time of different algorithms

      ModelAlgorithmin Ref. [3]VGG16KFCNN withfine-tuning
      Recognitiontime /s11.0579.63510.217
    Tools

    Get Citation

    Copy Citation Text

    Shuyu Wang, Shengxiang Tao, Fan Yang, Lei Ai. Laser Range-Gated Imaging Target Recognition Based on Convolutional Neural Network[J]. Laser & Optoelectronics Progress, 2019, 56(2): 021001

    Download Citation

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

    Category: Image Processing

    Received: Jun. 22, 2018

    Accepted: Jul. 26, 2018

    Published Online: Aug. 1, 2019

    The Author Email: Wang Shuyu (1512264822@qq.com)

    DOI:10.3788/LOP56.021001

    Topics