Laser & Optoelectronics Progress, Volume. 62, Issue 6, 0615014(2025)

Three-Dimensional Reconstruction Model for Unmanned Aerial Vehicle Images Combining Spatial Geometric Information and Global Features

Qiming Jin1,2、*, Feng Wang3,4, Juanjuan Yang3,5, Yang Pang3,4, and Jianwu Dang1,2
Author Affiliations
  • 1School of Electronic and Information Engineering, Lanzhou Jiaotong University, Lanzhou 730070, Gansu , China
  • 2National Virtual Simulation Experiment Teaching Center for Rail Transit Information and Control, Lanzhou 730070, Gansu , China
  • 3Gansu Luqiao Feiyu Transportation Facilities Co. Ltd., Lanzhou 730070, Gansu , China
  • 4Gansu Xinnetcom Technology Information Co., Ltd., Lanzhou 730070, Gansu , China
  • 5Gansu Provincial Highway Transportation Construction Group Co., Ltd., Lanzhou 730070, Gansu , China
  • show less
    Figures & Tables(14)
    Main framework
    Incremental SFM sparse reconstruction model framework
    Descriptor enhancement mode
    AFT model
    The SwiGLU model
    Datesets. (a) dataset 1; (b) dataset 2; (c) NPU_Central
    Point cloud reconstruction visualization results
    Comparison of point cloud results from self-collected datasets
    Comparison of public dataset point cloud results
    • Table 1. Experimental environment configuration

      View table

      Table 1. Experimental environment configuration

      NameParameters and information
      GPUNvidia GeForce RTX4080
      CPUIntel Core i7-13700kf@5.2 GHz
      RAM32 GB
      Operating systemUbuntu 20.04 LTS; Windows 11
    • Table 2. Algorithm performance comparison

      View table

      Table 2. Algorithm performance comparison

      DatasetAlgorithmSparse point↑Observation point↑Track length↑Reproj error↓
      dataset 1SIFT1578805369883.401.39
      ORB34383803552.331.49
      SuperPoint1697225940483.501.32
      SOSNet1689385710103.381.30
      SIFT+Boost1701176108783.591.15
      ORB+Boost39519889902.251.40
      dataset 2SIFT1137764582704.021.52
      ORB470281183162.511.55
      SuperPoint1345125835774.341.39
      SOSNet1316525634704.281.38
      SIFT+Boost1398176023354.311.35
      ORB+Boost496171385182.791.38
      NPU_CentralSIFT676254825657.130.79
      ORB2037837006183.430.88
      SuperPoint395552636806.661.25
      SOSNet505163288596.510.98
      SIFT+Boost764395748387.520.68
      ORB+Boost2102427236773.840.69
    • Table 3. Comparison of ablation experimental data

      View table

      Table 3. Comparison of ablation experimental data

      MethodSelf-mapping layerCross-mapping layerSparse pointReproj error
      SIFT1578801.39
      1623051.38
      1675491.35
      1701171.35
      ORB343831.49
      354921.49
      387511.47
      395191.46
    • Table 4. Feature extraction running time

      View table

      Table 4. Feature extraction running time

      Methoddataset 1dataset 2NPU_Central
      SIFT25.72811.86610.351
      ORB26.33813.09615.791
      SuperPoint27.57819.78814.559
      SOSNet26.35418.35413.854
      SIFT+Boost27.39913.80912.345
      ORB+Boost28.58912.71720.957
    • Table 5. Sparse reconstruction running time

      View table

      Table 5. Sparse reconstruction running time

      Methoddataset 1dataset 2NPU_Central
      SIFT382.538839.2992372.982
      ORB292.046521.1943637.606
      SuperPoint539.1471047.8623885.129
      SOSNet482.858853.3213186.541
      SIFT+Boost401.469867.4802836.873
      ORB+Boost289.590414.1353563.741
    Tools

    Get Citation

    Copy Citation Text

    Qiming Jin, Feng Wang, Juanjuan Yang, Yang Pang, Jianwu Dang. Three-Dimensional Reconstruction Model for Unmanned Aerial Vehicle Images Combining Spatial Geometric Information and Global Features[J]. Laser & Optoelectronics Progress, 2025, 62(6): 0615014

    Download Citation

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

    Category: Machine Vision

    Received: Jul. 25, 2024

    Accepted: Sep. 10, 2024

    Published Online: Mar. 13, 2025

    The Author Email:

    DOI:10.3788/LOP241743

    Topics