With the advance in 3D body scanning technology, it opens opportunities for virtual try-on and automatic made-to-measure in apparel products domain. This paper proposed a novel feature-based parametric method of human body shape from the cloud points of 3D body scanner [T2. Firstly, we improved the skeleton construction through adding and adjusting the position of joints. Secondly, automatic extraction approach of semantic feature cross-sections is developed based on the hierarchy. According to the unique distribution of cloud points of each cross-section of each body part, the extraction method of key points on the cross-section is described. Thirdly, we presented an interpolation approach of key points which fit cardinal spline to cross-section for each body part, in which tension parameter is used to represent the simple deformation of body shape. Finally, a connection approach of body part is proposed by sharing a boundary curve. The proposed method has been tested with our virtual human model (VHM) system which is robust and easier to use. The process generally requires about five minutes for generating a full body model that represents the body shape captured by 3D body scanner. The model can be imported in a CAD environment for application to a wide variety of ergonomic analyses.
Robust Pose Estimation for Human Body Hidden with a Cover Using Body Shape Feature Obtained by Normal Vector
