Journals Articles Conferences News About CLP
Submit a paper Email Alert
Search
Search
Articles
Journals
News
Advanced Search
Top Searches
2D MaterialsTransformation opticsQuantum PhotonicsSmall lasersSemiconductor UV PhotonicsSupersymmetric microring laser arrays

Chinese Journal of Lasers, Volume. 52, Issue 17, 1709001(2025)

Multi‐View 3D Reconstruction Based on Adaptive Feature Enhancement in Inspection Scenes

Yongkang Zhang1, Yi An1,2、*, Zhiyong Yang3, Yan Chen1, and Haochen Sun1
Author Affiliations
  • 1School of Electrical Engineering, Xinjiang University, Urumqi 830017, Xinjiang , China
  • 2School of Control Science and Engineering, Dalian University of Technology, Dalian 116024, Liaoning , China
  • 3Xinjiang Tianchi Energy Co., Ltd., Fukang831500, Xinjiang , China
    • show less
    AbstractGet PDF(in Chinese)
    Figures&Tables (14)
    Equations (0)
    References (31)
    Tools
    Paper Information
    Topics
    Figures & Tables(14)
    Multi-view 3D reconstruction framework for intelligent robot inspection scenes

    Fig. 1. Multi-view 3D reconstruction framework for intelligent robot inspection scenes

    Download full size
    Structure of AFE-MVSNet

    Fig. 2. Structure of AFE-MVSNet

    Download full size
    Structure of AFENet

    Fig. 3. Structure of AFENet

    Download full size
    Structure of SimAM and APMF. (a) SimAM; (b) APMF

    Fig. 4. Structure of SimAM and APMF. (a) SimAM; (b) APMF

    Download full size
    Example of DCN sampling offset

    Fig. 5. Example of DCN sampling offset

    Download full size
    Diagrams showing planar scanning method (left) and depth map refinement (right)

    Fig. 6. Diagrams showing planar scanning method (left) and depth map refinement (right)

    Download full size
    Fine-tuning process of multi-view 3D reconstruction network in inspection scenes

    Fig. 7. Fine-tuning process of multi-view 3D reconstruction network in inspection scenes

    Download full size
    Color images and depth maps of inspection scenes

    Fig. 8. Color images and depth maps of inspection scenes

    Download full size
    Comparison of the reconstruction effects of different methods on the DTU dataset

    Fig. 9. Comparison of the reconstruction effects of different methods on the DTU dataset

    Download full size
    Comparison of reconstruction effects between AFE-MVSNet before and after fine-tuning and CasMVSNet on the inspection scene dataset

    Fig. 10. Comparison of reconstruction effects between AFE-MVSNet before and after fine-tuning and CasMVSNet on the inspection scene dataset

    Download full size
    Comparison of reconstruction effects between AFE-MVSNet before and after fine-tuning and CasMVSNet on the large-scale scene

    Fig. 11. Comparison of reconstruction effects between AFE-MVSNet before and after fine-tuning and CasMVSNet on the large-scale scene

    Download full size

    • Table 1. Comparison of evaluation results of different methods on the DTU dataset

      View table

      Table 1. Comparison of evaluation results of different methods on the DTU dataset

      MethodAccuracyCompletenessOverall
      Gipuma0.2830.8730.578
      Colmap0.4000.6440.532
      MVSNet0.3960.5270.462
      R-MVSNet0.3830.4520.417
      CVP-MVSNet0.2960.4060.351
      CasMVSNet0.3250.3850.355
      PatchmatchNet0.4270.2770.352
      AA-RMVSNet0.3760.3390.357
      Effi-MVS0.3210.3130.317
      UniMVSNet0.3520.2780.315
      Ours0.3160.3020.309

    • Table 2. Comparison of reconstruction results between AFE-MVSNet before and after fine-tuning and CasMVSNet on the inspection scenes dataset

      View table

      Table 2. Comparison of reconstruction results between AFE-MVSNet before and after fine-tuning and CasMVSNet on the inspection scenes dataset

      ModelSceneEPE /me1 /%e3 /%
      CasMVSNet12.87316.1386.769
      22.65715.3266.116
      32.53417.7695.237
      41.71313.5884.256
      Mean2.44415.7055.595
      Before fine-tuning12.3598.9033.662
      21.9988.8543.931
      32.2259.3744.144
      40.9676.8952.458
      Mean1.8878.5073.549
      After fine-tuning11.2678.1542.927
      20.9087.7053.398
      31.1598.7713.883
      40.6896.5932.245
      Mean1.0067.8063.113

    • Table 3. AFE-MVSNet ablation experimental results

      View table

      Table 3. AFE-MVSNet ablation experimental results

      No.Model settingAccuracy /mmCompleteness /mmOverall /mmGPU memory usage /MBRuntime /s
      APMFSimAMF.L.
      10.3330.3950.36432280.270
      2✔0.3260.3660.34638580.687
      3✔✔0.3230.3090.31638580.695
      4✔✔✔0.3160.3020.30938580.695
    Tools

    Get Citation

    Copy Citation Text

    Yongkang Zhang, Yi An, Zhiyong Yang, Yan Chen, Haochen Sun. Multi‐View 3D Reconstruction Based on Adaptive Feature Enhancement in Inspection Scenes[J]. Chinese Journal of Lasers, 2025, 52(17): 1709001

    Download Citation

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

    Category: Imaging and Information Processing

    Received: Feb. 20, 2025

    Accepted: May. 8, 2025

    Published Online: Sep. 3, 2025

    The Author Email: Yi An (anyi@dlut.edu.cn)

    DOI:10.3788/CJL250543

    CSTR:32183.14.CJL250543

    Topics

    laser devices and laser physics
    Lasers and Laser Optics
    Laser physics
    laser manufacturing
    Instrumentation, Measurement and Metrology