Laser & Optoelectronics Progress, Volume. 58, Issue 22, 2215002(2021)

Binocular Ranging Method Based on Improved ORB-RANSAC

Chunjian Hua1,2、*, Rui Pan1,2, and Ying Chen3
Author Affiliations
  • 1School of Mechanical Engineering, Jiangnan University, Wuxi, Jiangsu 214122, China
  • 2Jiangsu Key Laboratory of Advanced Food Manufacturing Equipment & Technology, Wuxi, Jiangsu 214122, China;
  • 3School of Internet of Things Engineering, Jiangnan University, Wuxi, Jiangsu 214122, China
  • show less
    Figures & Tables(13)
    Structure of the epipolar constraint. (a) Convergent structure; (b) parallel structure
    Results of the epipolar correction. (a) Original image; (b) corrected image
    Structure of k-dimensional tree
    Flow chart of the improve RANSAC method
    Feature matching results of different methods. (a) ORB; (b) ORB-RANSAC; (c) our method
    • Table 1. Binocular calibration results

      View table

      Table 1. Binocular calibration results

      ParameterLeft cameraRight camera
      Internal parameter2178.270610.9602175.49506.770012181.220640.7402186.80546.99001
      Distortion-0.1340.8210.0020.0009-4.576-0.2191.4750.0040.007-6.443
      Rotation matrix0.9976-0.0084-0.06870.00140.9949-0.10070.06920.10030.9925
      Translation matrix-187.67911.3603-12.5848T
    • Table 2. Screening results of matching points in each step of our method

      View table

      Table 2. Screening results of matching points in each step of our method

      ImageOriginal numberAfter epipolar constraintAfter order constraintAfter RANSAC
      Vehicle232217189169
      Human193181145126
    • Table 3. Matching results of different methods on vehicle obstacle images

      View table

      Table 3. Matching results of different methods on vehicle obstacle images

      MethodMatched pairCorrect matchCorrect rate /%Time /ms
      ORB23218579.770.2
      ORB-RANSAC18918497.377.8
      Ours16916798.876.3
    • Table 4. Matching results of different methods on passerby obstacle images

      View table

      Table 4. Matching results of different methods on passerby obstacle images

      MethodMatched pairCorrect matchCorrect rate /%Time /ms
      ORB19315982.469.6
      ORB-RANSAC14013797.872.2
      Ours12612498.471.2
    • Table 5. Sub-pixel coordinate experiment

      View table

      Table 5. Sub-pixel coordinate experiment

      NumberLeft image coordinateRight image coordinateActual coordinates /m
      1(296.507, 321.013)(296.285, 300.480)(2.2855, 1.9014, -25.3138)
      2(300.761, 261.101)(300.409, 240.930)(2.2877, 2.4962, -25.7667)
      3(296.948, 368.129)(296.460, 347.887)(2.3194, 1.4921, -25.6768)
      4(216.032, 719.462)(216.366, 701.969)(3.5133, -2.0448, -29.7122)
      5(250.254, 641.423)(250.399, 624.064)(3.2014,-1.2149, -29.9411)
      6(320.500, 732.100)(315.594, 714.438)(2.4056, -2.1744, 29.5954)
    • Table 6. Ranging results of Ref. [17]

      View table

      Table 6. Ranging results of Ref. [17]

      Ref. [17]Actual distance /mMeasuring distance /mError /mRelative error /%Time /s
      154.8080.1923.840.10825
      2109.6890.3113.110.11358
      31516.0851.0156.760.10433
      42021.2261.2266.130.10957
      Average--0.6864.960.10893
    • Table 7. Ranging results of Ref. [18]

      View table

      Table 7. Ranging results of Ref. [18]

      Ref. [18]Actual distance /mMeasuring distance /mError /mRelative error/%Time /s
      154.7860.2144.280.08125
      2109.5680.4324.320.08756
      31516.1241.1247.490.07812
      42021.6261.6268.130.07628
      Average--0.8496.060.08080
    • Table 8. Ranging results of our method

      View table

      Table 8. Ranging results of our method

      OursActual distance /mMeasuring distance /mError /mRelative error /%Time /s
      154.8740.1262.520.09097
      2109.7810.2192.190.09228
      31515.6590.6594.390.08814
      42020.9980.9984.990.08557
      Average--0.50053.520.08924
    Tools

    Get Citation

    Copy Citation Text

    Chunjian Hua, Rui Pan, Ying Chen. Binocular Ranging Method Based on Improved ORB-RANSAC[J]. Laser & Optoelectronics Progress, 2021, 58(22): 2215002

    Download Citation

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

    Category: Machine Vision

    Received: Dec. 8, 2020

    Accepted: Jan. 21, 2021

    Published Online: Nov. 5, 2021

    The Author Email: Chunjian Hua (cjhua@jiangnan.edu.cn)

    DOI:10.3788/LOP202158.2215002

    Topics