Accurate digital reconstruction of the human body is critical for a wide range of applications, including apparel design, ergonomic evaluation, and biomedical engineering. This study introduces a novel 3D human modeling framework that combines skeleton-driven deformation algorithms with high-resolution 3D body scan data to generate anatomically accurate virtual bodies. The proposed method was benchmarked against two widely adopted modeling systems—the Skinned Multi-Person Linear (SMPL) model and the CLO 3D avatar—using a representative female scan from the Size Korea anthropometric dataset as a reference. Comparative evaluations were conducted across multiple dimensions, including anthropometric measurement accuracy and geometric surface fidelity. The SMPL model consistently overestimated key body dimensions due to its reliance on global average parameters, while the CLO 3D model tended to underestimate volumes, particularly in anatomically complex regions. Both models exhibited limitations in reproducing local morphological features such as curvature transitions and subcutaneous contours. In contrast, the proposed model achieved superior alignment with the reference scan, demonstrating statistically non-significant deviations in most measurements and enhanced surface realism. Additionally, an automated measurement module was developed to extract standardized anthropometric values from the mesh with high precision. These results validate the effectiveness of the proposed system as a robust tool for human-centered applications requiring high-fidelity modeling and optimized fit in mass-customized product development environments.
扫码关注我们
求助内容:
应助结果提醒方式:
