Laser & Optoelectronics Progress, Volume. 62, Issue 16, 1615009(2025)

Bridge Realistic Reconstruction Algorithm Based on Improved 3D Gaussian Splatting

Siqi Huang1, Nan Jiang2、*, Hong Liang1, Rong Wu1, and Haoyan Li1
Author Affiliations
  • 1School of Information Science & Engineering, Yunnan University, Kunming 650504, Yunnan , China
  • 2Yunnan Provincial Highway Science and Technology Research Institute, Kunming 650051, Yunnan , China
  • show less
    Figures & Tables(12)
    3DGS overall framework flowchart
    Specific process of adaptive density control
    Different pixel ranges selected by 3DGS and the proposed method. (a) 3DGS algorithm; (b) proposed method
    Different integration strategies used in 3DGS and the proposed method. (a) 3DGS algorithm (performing numerical integration in the integration area); (b) proposed method (performing window integration in the integration area)
    Rotation integration of pixels in the window area. (a) Integration in the original pixel area; (b) integration after rotating the pixel area
    Conditions for adaptive density control before and after improvement
    Improved overall framework flowchart
    Partial images in the dataset
    Comparison curves of PSNR, SSIM, and LPIPS indicators between 3DGS algorithm and the proposed method. (a) PSNR; (b) SSIM; (c) LPIPS
    Comparison of rendering effects before and after 3DGS algorithm improvement. (a) Original images; (b) 3DGS algorithm; (c) proposed method
    • Table 1. Results of ablation experiments

      View table

      Table 1. Results of ablation experiments

      MethodPSNR /dB↑SSIM↑LPIPS↓
      None27.380.9010.226
      3D filter27.540.9030.224
      Integration28.020.9100.219
      Pixel gradient scaling28.290.9240.193
      Tensor decomposition28.380.9110.221
      3D filter+integration28.380.9130.216
      3D filter+integration+pixel gradient scaling28.420.9240.189
      3D filter+integration+pixel gradient scaling+tensor decomposition29.290.9270.189
    • Table 2. Results of comparison experiments

      View table

      Table 2. Results of comparison experiments

      ModelIteration /103PSNR/dB↑SSIM↑LPIPS↓Model size /MBRendering time /min
      NeRF3018.450.4900.530709
      NeRF_pl3019.900.5480.5891655
      TensoRF3019.470.3390.49049
      3DGS3027.380.9010.226755.025
      CityGaussain193027.360.8990.225735.052
      SAGS233027.530.9010.224770.046
      Mip-splatting173028.960.9230.184969.034
      Optimal-GS243027.230.8980.227750.0117
      GaussianShader203025.970.8680.270291.0198
      Spec-Gaussian253029.150.9240.163300.0105
      Mini-splatting263026.630.8850.21997.533
      FIPT-GS3029.290.9270.189842.059
    Tools

    Get Citation

    Copy Citation Text

    Siqi Huang, Nan Jiang, Hong Liang, Rong Wu, Haoyan Li. Bridge Realistic Reconstruction Algorithm Based on Improved 3D Gaussian Splatting[J]. Laser & Optoelectronics Progress, 2025, 62(16): 1615009

    Download Citation

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

    Category: Machine Vision

    Received: Dec. 13, 2024

    Accepted: Mar. 18, 2025

    Published Online: Aug. 4, 2025

    The Author Email: Nan Jiang (1186285437@qq.com)

    DOI:10.3788/LOP242428

    CSTR:32186.14.LOP242428

    Topics