Study On Resizable Body Model In Three-Dimensional Virtual Fitting System

With rapid development of clothing sales on the network, people are not only satisfied with browsing color, style and size of clothes on line, but also wonder if those clothes fit them. Three-dimensional virtual fitting system comes into being which provides real shopping experience for the consumers and meet their needs of fit. Such vivid and virtual way of showing the clothes will give consumers more intuitive experience.This paper outlined three-dimensional fitting system research, pointed out difficulties of implementation and founded basis for future realization. . First of all, the speedy establishment of the human model, espeically fast reconstruction of partial human model, plays an important role in the user experience by simply changing some parameters such as bust, waist, and shoulder etc. However, it is hard to reconstruct the model when the parameters are changed. Secondly, it is difficult to figure out effects of different clothes on different human body model, especially those unfit clothes. This further disturbs consumers'judgment on the comfort of the clothes. Due to limitation of time and resources, only the first question was analyzed and implemented thoroughly. By studying related foreign and domestic researches which use complex algorithm implementation, we simplify the design ideas to meet the speedy requirements of model without affecting the real effect. Because the body builds are differ in thousands ways in the world, it is hard to store all the body data. Targeting this problem, we use standard human body model and the scaling model to achieve the model resizable when the parameters is changed. The scaling model is partial reconstruction based on the standard model to meet the new demand of the model.Analyzing the parametric modeling techniques and human body modeling, , we decide to use the NURBS surface to implement human body model. Firstly, the vertex data of the standard model is exported from Poser software. According to these vertex data, we sample the data and remove the noise, then use NURBS curve to fit the data. We get the control value according to the NURBS curves. Based on the control value, the model of standard human body is established. Secondly, using the standard human body model as the reference model and combining data like the height, weight, and waist, we achieve the scaling model of the human body. In this paper, we use reversed standard normal distribution curve to solve the problems of chest distortion. We consider using the distance of adjacent samples layer to determine the magnitude of pulling. With the anisotropy ellipse curve, we solve different resized ratio in different direction in weight resizable model. Using standard normal distribution curve as resized ratio, we optimize the smooth change of waist curve. Finally, comparing with the length- weight resizable model and standard model in same ratio, we get the accuracy of resizable model. Analyzing the data of experiment, we find the accuracy meet the requirement when the resized ratio is from 0.8 to 1.4.At last, we carry out HTML5 and WEBGL to implement the model in web style,which laid the foundation of three-dimensional fitting system.