Human body construction based on combination of parametric and nonparametric reconstruction methods

Nowadays, 3D human models are widely used in the garment industry where it is important to reconstruct compliant human model from scan of the body under partial dress for privacy reasons. A new 3D human construction method based on the combination of parametric and nonparametric reconstruction is proposed here. Inputs to the method include the raw scan and partial anthropometric parameters. The scan is divided into exposed area and clothing occluded area for modeling separately. The information of the clothing occluded area is restored by semantic parametric human modeling and pose fitting. The information of the exposed area is restored by the method of non-rigid registration. Then, the two types of information are fused to reconstruct the final human model. The experiment divided the human scan into three common situations: naked, partially clothed and highly clothed body scanning. The results of the qualitative and quantitative analyses show that the method is able to fit the parametric human model to the exposed area scan while matching the user input on the anthropometric parameters in the clothing occluded area. It is worth pointing out that the connection between the two areas is smooth. The performance is also better than previous related methods. The proposed method reduces the dressing requirements for reconstructing the human body from 3D scan and also demonstrates the validity, accuracy and versatility of the method for reconstructing human model.