• Laser & Optoelectronics Progress
  • Vol. 58, Issue 8, 0810017 (2021)
Qingpeng Zhang and Yu Cao*
Author Affiliations
  • School of Automation, Harbin University of Science and Technology, Harbin, Heilongjiang 150080, China
  • show less
    DOI: 10.3788/LOP202158.0810017 Cite this Article Set citation alerts
    Qingpeng Zhang, Yu Cao. Research on Three-Dimensional Reconstruction Algorithm of Weak Textured Objects in Indoor Scenes[J]. Laser & Optoelectronics Progress, 2021, 58(8): 0810017 Copy Citation Text show less
    Image to be matched taken by mobile phone
    Fig. 1. Image to be matched taken by mobile phone
    Matching results of different algorithms. (a) Harris algorithm; (b) SIFT algorithm; (c) Harris-SIFT algorithm; (d) our algorithm
    Fig. 2. Matching results of different algorithms. (a) Harris algorithm; (b) SIFT algorithm; (c) Harris-SIFT algorithm; (d) our algorithm
    Experimental images at different l. (a) l=3; (b) l=2; (c) l=1
    Fig. 3. Experimental images at different l. (a) l=3; (b) l=2; (c) l=1
    Structure of the ResNet50
    Fig. 4. Structure of the ResNet50
    Some experimental images
    Fig. 5. Some experimental images
    Sparse three-dimensional point cloud. (a) SFM algorithm; (b) improved SFM algorithm
    Fig. 6. Sparse three-dimensional point cloud. (a) SFM algorithm; (b) improved SFM algorithm
    Dense point cloud of the object. (a) MVS algorithm; (b) our algorithm
    Fig. 7. Dense point cloud of the object. (a) MVS algorithm; (b) our algorithm
    Three-dimensional reconstruction results of the Poisson surface. (a) Original algorithm; (b) improved algorithm
    Fig. 8. Three-dimensional reconstruction results of the Poisson surface. (a) Original algorithm; (b) improved algorithm
    Matching algorithmHarrisSIFTHarris-SIFTImproved Harris-SIFT
    Left image feature point2057278614901450
    Right image feature point2172280116121597
    Matching point1600821506492
    Matching accuracy /%81.6193.5493.6794.92
    Matching time /s5.0211.643.943.61
    Table 1. Accuracies and running time of 4 matching algorithms
    lNumber of image matchesMatch time /sNumber of point clouds
    13271994101823
    2245896199745
    3155414265521
    Table 2. Image parameters at different l
    AlgorithmXRMSEYRELZlog10
    MVS2.2160.1010.075
    Ours1.1230.0810.038
    Table 3. Errors of different algorithms
    AlgorithmTime /minPoint cloudSSIM
    SFM+MVS8.212281660.53
    Improve algorithm6.482616170.81
    Table 4. Three-dimensional reconstruction results of different algorithms
    Qingpeng Zhang, Yu Cao. Research on Three-Dimensional Reconstruction Algorithm of Weak Textured Objects in Indoor Scenes[J]. Laser & Optoelectronics Progress, 2021, 58(8): 0810017
    Download Citation