• Chinese Journal of Lasers
  • Vol. 47, Issue 12, 1204005 (2020)
Wu Junlong1、2、3, Guo Zhenghua1、2、3, Chen Xianfeng1、2、3, Ma Shuai1、2、3, Yan Xu1、2、3, Zhu Licheng1、2、3, Wang Shuai1、3, and Yang Ping1、3、*
Author Affiliations
  • 1Key Laboratory on Adaptive Optics, Chinese Academy of Sciences, Chengdu, Sichuan 610209, China
  • 2Institute of Optics and Electronics, Chinese Academy of Sciences, Chengdu, Sichuan 610209, China
  • 3University of Chinese Academy of Sciences, Beijing 100049, China
  • show less
    DOI: 10.3788/CJL202047.1204005 Cite this Article Set citation alerts
    Wu Junlong, Guo Zhenghua, Chen Xianfeng, Ma Shuai, Yan Xu, Zhu Licheng, Wang Shuai, Yang Ping. Three-Dimensional Measurement Method of Light Field Imaging Based on Deep Learning[J]. Chinese Journal of Lasers, 2020, 47(12): 1204005 Copy Citation Text show less
    Projection model of light field camera
    Fig. 1. Projection model of light field camera
    Relationship between disparity and depth
    Fig. 2. Relationship between disparity and depth
    Illustration of equivalent baselines
    Fig. 3. Illustration of equivalent baselines
    Structure of network
    Fig. 4. Structure of network
    Illustration of network input
    Fig. 5. Illustration of network input
    Center view and GT disparity of synthetic datasets
    Fig. 6. Center view and GT disparity of synthetic datasets
    Test results of synthetic dataset
    Fig. 7. Test results of synthetic dataset
    Test results of real dataset
    Fig. 8. Test results of real dataset
    Three-dimensional reconstruction results of test scene
    Fig. 9. Three-dimensional reconstruction results of test scene
    Measurement of real scene scale. (a) Test scene; (b) disparity map; (c) three-dimensional reconstruction structure
    Fig. 10. Measurement of real scene scale. (a) Test scene; (b) disparity map; (c) three-dimensional reconstruction structure
    SceneLF_OCCEPI2LFFocalStackNetEPInetProposed
    Boxes26.5229.8023.0214.3312.2411.46
    Cotton6.2216.697.830.580.460.51
    Dino14.9115.6719.032.531.261.14
    Sideboard18.5018.9521.995.404.784.56
    Backgammon19.0122.085.524.343.281.87
    Pyramids3.171.0812.350.290.150.28
    Dots5.8246.532.901.021.983.49
    Stripes18.4123.8135.743.720.910.85
    Table 1. Badpixel(0.07) comparison of different algorithms
    SceneLF_OCCEPI2LFFocalStackNetEPInetProposed
    Boxes9.8510.9317.4311.826.014.81
    Cotton1.074.329.170.880.220.23
    Dino1.142.071.160.890.150.15
    Sideboard2.304.655.071.960.810.59
    Backgammon21.5920.7813.016.583.912.39
    Pyramids0.100.020.270.020.0070.01
    Dots3.306.665.681.871.984.29
    Stripes8.136.1017.451.790.910.85
    Table 2. MSE comparison of different algorithms
    SceneLF_OCCEPI2LFFocalStackNetEPInetProposed
    Boxes10408.268.91962.1885.042.031.03
    Cotton6325.519.07984.5384.902.031.04
    Dino10099.058.091130.5685.622.031.07
    Sideboard13531.308.74987.4784.772.021.11
    Backgammon5116.256.93979.8784.242.021.04
    Pyramids11688.426.88929.7292.092.021.04
    Dots10820.857.66979.8781.562.031.08
    Stripes19331.428.531093.9891.922.031.07
    Table 3. Runtime comparison of different algorithms unit: s
    Wu Junlong, Guo Zhenghua, Chen Xianfeng, Ma Shuai, Yan Xu, Zhu Licheng, Wang Shuai, Yang Ping. Three-Dimensional Measurement Method of Light Field Imaging Based on Deep Learning[J]. Chinese Journal of Lasers, 2020, 47(12): 1204005
    Download Citation