• Laser & Optoelectronics Progress
  • Vol. 60, Issue 10, 1010011 (2023)
Wen Yang, Mingquan Zhou, Guohua Geng*, and Xiaoning Liu
Author Affiliations
  • College of Information Science and Technology, Northwest University, Xi'an 710127, Shaanxi, China
  • show less
    DOI: 10.3788/LOP213393 Cite this Article Set citation alerts
    Wen Yang, Mingquan Zhou, Guohua Geng, Xiaoning Liu. Skull Identification Method Based on Fusion of View and Shape Features[J]. Laser & Optoelectronics Progress, 2023, 60(10): 1010011 Copy Citation Text show less
    Depth projection image of skull model
    Fig. 1. Depth projection image of skull model
    Depth projection image of facel model
    Fig. 2. Depth projection image of facel model
    Multi-view neural network architecture
    Fig. 3. Multi-view neural network architecture
    Examples of partial skull and face samples
    Fig. 4. Examples of partial skull and face samples
    Number of sampling points200040006000800010000
    SkullSampling time29.71957.79286.441115.869176.357
    Geometric distance23.167102.374237.072415.626574.989
    Biharmonic distance1.0642.39511.83323.25443.152
    FaceSampling time34.22467.556100.859127.433204.613
    Geometric distance24.866108.918252.316445.771611.438
    Biharmonic distance1.1182.50212.36024.43545.079
    Table 1. Time consuming of sampling operation and distance matrix calculation of skull and facial model under different sampling points
    MethodCorrect recognition rate
    Geodesic distance+Mean-based MVCNN91.4
    Geodesic distance+Max-based MVCNN91.8
    Geodesic distance+LMPF-based MVCNN93.6
    Biharmonic distance+Mean-based MVCNN92.1
    Biharmonic distance+Max-based MVCNN93.2
    Biharmonic distance+LMPF-based MVCNN95.4
    Table 2. Correct recognition rate results of six groups of experiments
    FeatureWave kernel featureScale invariant wave kernel feature
    Correct recognition rate /%93.995.4
    Table 3. Correct recognition rate results of wave kernel features and scale invariant wave kernel features
    MethodMulti-view feature3D shape featureRBF kernel functionPolynomial kernel functionExponential kernel function
    Correct recognition rate /%92.891.487.889.395.4
    Table 4. Recognition results of different kernel functions
    Number of views4812161820
    Correct recognition rate /%63.878.486.792.693.495.4
    Table 5. Recognition results of the number of different views
    MethodMethod of reference[30Method of reference[31Method of reference[15Method of reference[32

    method

    Proposed

    Correct recognition rate /%88.691.493.895.295.4
    Table 6. Recognition results of different methods
    Wen Yang, Mingquan Zhou, Guohua Geng, Xiaoning Liu. Skull Identification Method Based on Fusion of View and Shape Features[J]. Laser & Optoelectronics Progress, 2023, 60(10): 1010011
    Download Citation