• Acta Optica Sinica
  • Vol. 38, Issue 8, 0815004 (2018)
Zhiyuan Zhang*, Wei Liu*, Yang Zhang, Yongkang Lu, Hongtu Di, Fan Ye, and Zhenyuan Jia
Author Affiliations
  • School of Mechanical Engineering, Dalian University of Technology, Dalian, Liaoning 116024, China
  • show less
    DOI: 10.3788/AOS201838.0815004 Cite this Article Set citation alerts
    Zhiyuan Zhang, Wei Liu, Yang Zhang, Yongkang Lu, Hongtu Di, Fan Ye, Zhenyuan Jia. Calibration Method for Large Field of View Image Matching Parameters Based on Non-Metric Correction[J]. Acta Optica Sinica, 2018, 38(8): 0815004 Copy Citation Text show less
    Layout strategy of feature points
    Fig. 1. Layout strategy of feature points
    Distribution of feature points in Hough space
    Fig. 2. Distribution of feature points in Hough space
    Schematic of single point distortion correction
    Fig. 3. Schematic of single point distortion correction
    Schematic of epipolar geometry
    Fig. 4. Schematic of epipolar geometry
    Principle of RANSAC method
    Fig. 5. Principle of RANSAC method
    Schematic of experimental system. (a) Binocular camera; (b) calibration target
    Fig. 6. Schematic of experimental system. (a) Binocular camera; (b) calibration target
    Matching results. (a) Small fluctuation area identification; (b) collinear feature point identification; (c) feature point matching results
    Fig. 7. Matching results. (a) Small fluctuation area identification; (b) collinear feature point identification; (c) feature point matching results
    Distance bias of corner point and epipolar line with different methods
    Fig. 8. Distance bias of corner point and epipolar line with different methods
    Point numberLeft image /pixelRight image /pixelCorrected left image /pixelCorrected right image /pixel
    1(1146.3296,491.7768)(840.5358,447.0849)(1142.8455,488.4470)(835.2091,442.9604)
    2(1149.5898,413.0614)(842.9693,373.3974)(1145.9285,409.1640)(837.2593,368.5852)
    3(1233.4019,497.7639)(917.1522,447.4341)(1230.5533,494.6613)(912.6476,443.4858)
    304(3278.3531,2642.6279)(2587.0324,2562.6324)(3284.9730,2649.5132)(2588.7491,2566.4980)
    Table 1. Results of feature points extraction andnonmetric correction
    MethodFundamental matrix
    Proposed methodF=-3.2069×10-94.0815×10-7-5.9270×10-43.7207×10-6-3.3833×10-8-0.0139-5.9749×10-40.01220.9998
    Classic RANSACF=-3.6104×10-93.2269×10-7-4.4663×10-44.2009×10-7-3.3961×10-8-0.0130-6.6870×10-40.01140.9998
    8-pointF=-1.7171×10-93.6731×10-7-5.2393×10-43.7925×10-7-2.9383×10-8-0.0130-6.0924×10-40.01140.9999
    Calibration methodF=3.7275×10-9-5.1012×10-77.4736×10-4-4.5940×10-74.1281×10-80.01727.3055×10-4-0.0152-0.9997
    Table 2. Comparison of fundamental matrices results
    MethodAverage /pixelStandarddeviation /pixel
    Proposed method0.17460.0861
    Classical RANSAC0.49170.4942
    8-point0.34190.2487
    Calibration method10.00412.0401
    Table 3. Comparison of matching error
    Zhiyuan Zhang, Wei Liu, Yang Zhang, Yongkang Lu, Hongtu Di, Fan Ye, Zhenyuan Jia. Calibration Method for Large Field of View Image Matching Parameters Based on Non-Metric Correction[J]. Acta Optica Sinica, 2018, 38(8): 0815004
    Download Citation