디지털 인체 모델의 자세 변화에 따른 체표 변형 가중치 산정 방법 개발

Abstract

Digital Human Modeling (DHM) is a technology that implements the shape of a human body in the form of 3D data in a virtual environment. It can be used for measuring dimensions of human body and analyzing shape changes. DHM used in ergonomic product design needs various forms because of the variety of product usage environment. Deformable DHM with skinning weight can be transformed from one DHM posture into multiple postures, so it can be effectively used in various product designs. DHM with appropriate skinning weight can deform its shape similar to the real shape of the reference postures. Present studies of calculating skinning weight method were conducted in two directions: (1) a mathematical model-based method that calculates skinning weights from one model data, and (2) an example poses-based method that calculates skinning weights using model data of various poses. The skinning weight derived in the previous study had a problem in that it was not possible to measure the accuracy of shape deformation, or that standardized measurement and comparative evaluation were not performed. Therefore, it was not possible to compare deformed shape by applying the skinning weight and actual human body shape. This study consists of three steps: (1) Construction of hand mesh data set contains joint Center of Rotation (CoR), registered hand mesh, (2) development registered hand mesh-based skinning weight derivation method, and (3) Evaluation of the performance of the skinning weight. First, the 3D hand model was constructed from CT scan data for 10 postures of 9 people. Using the MITK program (Nolden et al., 2013), the hand mesh and bone mesh for each subject's posture were separated and extracted. CoR location of each meshes was calculated by applying the Delonge-Kasa (Kasa, 1976) method. By applying template registration, hand mesh data of all postures had the same mesh structure (# of vertices, vertex index, face). Second, the skinning weight which is applicable to the registered hand mesh was derived. The optimal skinning weight for each posture was derived using the hand mesh data set constructed in the first step. Derived skinning weights for each posture include outliers causing defects when the mesh is deformed in different postures. A suitable skinning weight set for the registered hand mesh was derived by removing outliers using a statistical method. Third, the deformation error was measured by applying the skinning weight set derived in this study and compared with the previous studies. Along with (1) the skinning weight set derived from this study, (2) heat diffusion (Baran & Popovic, 2007), (3) distribution function (Flores & Sánchez, 2017), (4) distance ratio (Li et al, 2018) were simultaneously compared. The method of the previous study derives the skinning weight of each hand mesh of the basic (extension) posture. The skinning weight performance evaluation was quantified by measuring the root mean square deformation error (RMSE) of the deformed mesh and the actual mesh when the skinning weight was applied to the hand mesh of the basic posture and transformed into another posture. The RMSE for each method is (1) skinning weight set: 4.61 ± 0.02, (2) heat diffusion: 4.78 ± 0.02, (3) distribution function: 4.89 ± 0.02, (4) distance ratio: 4.74 ± 0.02. It has been shown that the method suggested in this study transforms the shape most similar to the real hand mesh. In the method of the previous study, there was no significant difference between heat diffusion and distance ratio, and the performance of the distribution function was found to be the worst.