Real-time Human Body Modeling And The System Implementation For Virtual Fitting

Along with the development of computer science and technology,augmented real-ity technology is gradually mature,and e-commerce has increasingly become an indis-pensable part of our daily life.As an important part of e-commerce,online shopping for clothing has a great demand.However,people are not just only satisfied with browsing pictures of dressed models on the webpage.So various virtual fitting systems appear constantly,to provide technical support for the online virtual fitting room.There are various implementation ways of virtual fitting according to different goals.In recent years,many works are either based on image processing to generate virtual fitting effects or based on the user input to construct 3D models and then simulate clothes in a virtual scene.All these works have produced impressive results,but these effects still lack immersion and reality since augmented reality has not been brought in.Virtual fitting with augmented reality effect not only requires fast generation of human body model that fits the gesture and shape of people,but also needs to process cloth simulation based on the generated human body model.Compared with the former implementation ways,its technology demand and computational cost are higher,so this kind of system is still in the research stage.This thesis presents a real-time 3D human body modeling method,and on the basis of this,a virtual fitting system with augmented reality effect is proposed.To build the 3D human body model,we divide the human body into static and dynamic parts in terms of deformation,and then process the building tasks of those parts separately according to the divide and conquer thoughts,in order to quickly generate a 3D human body model which fits the human body shape and the posture of current people.The bottom part is fitted to the human body through mesh deformation in particular.Also,we provide anti-twist technology and mesh smoothing technique to guarantee the smoothness of the generated model.As for the implementation of virtual fitting system,the human body model is first generated from the measurements,and then cloth simulation technology is used for simulating clothes,after that the simulation result is synthesized with the real scene image to obtain the fitting effect.Experiments demonstrate that the real-time 3D human modeling proposed in this thesis can generate 3D human models quickly andeffectively that fits to the shape and posture of the human body.The fitting results show that our virtual fitting system can bring a relatively realistic experience to users.The virtual fitting system is based on Kinect,which has a certain versatility and practicality in research and application.