The Design And Implementation Of Parametric Modeling Of Three-dimensional Human Body

With the rapid development of internet technologies and electronic commerce at present, humans’requirements of online virtual fitting and clothing customizationare growing, which boosts a number of business opportunities. The greatest challenge of online personalizedclothing customization and virtual fitting is how to establish3D body model that is consistent with humans’ physical shapes in real time.And the quality of body modeling algorithms directly affects the difficulty of following processing and final exhibition results. Thus, parametric3Dbody modeling has caused much attention in the clothing industry, as well as been a hot spot for the foreign and domesticresearchers. Currently, the typical3D human body modeling methods face one or more of following problems:complicated processing, a large scale of data, high demand of equipment and space, weak sense of reality, and so on.To address the above problems, we focus on the virtual fitting system and the study of parametric3D human body modeling, and a novel parametric3D human body modeling method is proposed, based on which we implemented a real-time system. It exploited theories of anthropometry and anatomy, and defined a group of feature points and scales, which can control topologiesof human body, integrating with points and lines that reflect body physical shape. Compared with other parametric modeling methods, the new approach not only defined the key parameters including bust, waist, hip and height, but also introduced auxiliary ones, such as milk spacing, breast height and thickness of hip, improving the reconstructed body model more natural and vivid.According to the properties of3D virtual fitting system and the feasibility of the system implementation, we proposed to obtain feature points and sizes of human body with the Microsoft Kinect depth camera. Based on the study of body sectional contour of each layer, we further combined feature points, feature size, and standard body template library to learn points of body surface.On this basis, the paper tried two methods to achieve a three-dimensional body surface reconstruction:first we applied NURBS curve fitting techniques to3D modeling of trunk and limbs. In this part, we analyzed the algorithm that reversely calculated NURBS curve grid points with surface points, and NURBS curve fitting method as well, and we take a further discussion about curve surface splicing and smoothing.For the second method, we further adopts Zhang Spline Surface Interpolation presented by Zhang to simulate body surface. This method has obvious advantages on calculating control points、curved stitching、parametric deformation and animationcomparing with NURBS fitting.Our system implemented the API to interact with Kinect camera,if humans do some actions in front of the Kinect camera, and the3D body model of the system will capture and display the animations. besides this, the system will create initial body model, and users just input personal body parameters or auxiliary parameters to get their new-personalized model.