• Laser & Optoelectronics Progress
  • Vol. 58, Issue 6, 615005 (2021)
Guo Keyou1、*, Yang Min1, Zhang Mo2, Guo Xiaoli1, and Li Xue1
Author Affiliations
  • 1School of Artificial Intelligence, Beijing Technology and Business University, Beijing 100048, China
  • 2Research Institute of Highway Ministry of Transport, Beijing 100088, China
  • show less
    DOI: 10.3788/LOP202158.0615005 Cite this Article Set citation alerts
    Guo Keyou, Yang Min, Zhang Mo, Guo Xiaoli, Li Xue. Real-Time Monocular Depth Estimation Method Based on Perspective N-Point model[J]. Laser & Optoelectronics Progress, 2021, 58(6): 615005 Copy Citation Text show less
    Specific implementation process of our method
    Fig. 1. Specific implementation process of our method
    Flow chart of the camera calibration
    Fig. 2. Flow chart of the camera calibration
    Flow chart of the license plate location algorithm
    Fig. 3. Flow chart of the license plate location algorithm
    License plate images obtained through preliminary screening
    Fig. 4. License plate images obtained through preliminary screening
    Schematic diagram of the homography model
    Fig. 5. Schematic diagram of the homography model
    Simplified geometry model of road scene. (a) Depth geometry model; (b) road model of the vehicle-mounted camera and imaging diagram; (c) principle of the vertical imaging ; (d) principle of the horizontal imaging
    Fig. 6. Simplified geometry model of road scene. (a) Depth geometry model; (b) road model of the vehicle-mounted camera and imaging diagram; (c) principle of the vertical imaging ; (d) principle of the horizontal imaging
    Model of the quadrilateral space object point coordinate system
    Fig. 7. Model of the quadrilateral space object point coordinate system
    Depth measurement result
    Fig. 8. Depth measurement result
    Application effect diagram of our method on video sequence. (a) Video sequence effect chart1; (b) video sequence effect chart2
    Fig. 9. Application effect diagram of our method on video sequence. (a) Video sequence effect chart1; (b) video sequence effect chart2
    Experiment methodBest parameter set(Cp,g)Experience risk/%Confidence risk/%Structured risk/%Training time/s
    PSO(1.3204,0.01)00.47570.475712364.9101
    GA(1.0053,0.04673)0004521.0216
    Table 1. Combination characteristic license plate data selected by SVM
    Object pointTrue depth information/mImage coordinates
    Object point15.5(1220,751)
    Object point26.0(1260,753)
    Object point36.5(1289,759)
    Table 2. Basic information of the objects
    Guo Keyou, Yang Min, Zhang Mo, Guo Xiaoli, Li Xue. Real-Time Monocular Depth Estimation Method Based on Perspective N-Point model[J]. Laser & Optoelectronics Progress, 2021, 58(6): 615005
    Download Citation