Research on Intelligent Fitting Based on Parametric Human Modeling and 3D Garment Simulation

This study aims to design a highly realistic virtual fitting system using advanced 3D modeling and dynamic simulation techniques to achieve a seamless transition from 2D clothing design to 3D garment simulation, meeting the needs of modern fashion design and production. The system uses SMPL shape parameters to represent various human body shapes and employs CNN to extract human key points. Through the twist and swing decomposition technique, 3D joint rotations are converted into relative body part rotations, generating motion files for 3D human mesh reconstruction. Additionally, the study introduces methods to convert 2 D clothing design drawings into 3D stereoscopic images using parametric garment modeling techniques, which include sketching 2D clothing outlines and setting finite element segmentation particle spacing to impart physical properties to the fabric, enhancing the realism of virtual fitting technology. Intuitive explicit parametric garment modeling excellently simulates the dynamic effects and realism of clothing, demonstrating the relationship between various human parameters and garment parameters, making it convenient to implement.