Research On Key Technologies Of Virtual 3D Garment Display

Unlike physical try-on,virtual try-on system allow shoppers to visually visualize the effect of virtual garment before purchasing.The current virtual fitting system can achieve better results in the interaction of simulation environment,style design,automatic cutting and cloth simulation,but for the virtual garment fitting evaluation,garment pattern adaptive adjustment,realistic small wrinkle generation,the display of virtual character garment effect and the display of garment deformation effect,there are still many defects in the key technologies of virtual 3D garment display,which have become a bottleneck in the application of 3D virtual try-on system.This thesis focuses on the research of key technologies of virtual 3D garment display.And has achieved the following research results.1)Based on the correlation between fitting evaluation and garment pattern modification,a novel virtual try-on method is proposed.In the phase of fitting evaluation,the shape variable or L2-tensor is selected as the fitting evaluation metric.In the phase of mode modification,a prior condition based on the maximum entropy coordinate and an improved mean value coordinate interpolation algorithm are proposed.Compared with traditional virtual try-on system,it can easily locate the position where needs to be modified.The modification of any garment mode can be realized through smooth mesh update,which improves the efficiency and accuracy of the garment design.2)Based on the characteristics of key parts of the human body,a dual adaptive adjustment method of garment pattern is proposed.Firstly,the rough automatic adjustment of the garment pattern is created,and the size of the garment is controlled by the length of the bounding box generated by five key parts of the human body,which not only ensures the accuracy of the virtual try-on but also improves the efficiency of it.Then,the relocation of the boundary nodes is constructed using the garment pattern boundary rule,and adapt the virtual human body using a fine automatic adjustment method.This method can automatically adjust on a virtual human body with a large difference body proportions and maintain garment style.3)A virtual garment wrinkle synthesis method based on single image is proposed.The deep learning image retrieval method is applied to the retrieval of human body shape.And the human body reshaping,image segmentation and shape recovery are introduced into the garment modeling process.At the same time,the unmatched wrinkles are removed by energy optimization method.The experimental results show that,compared with other existing garment modeling methods,this method can quickly capture the real wrinkles of garment,reducing manual operation and bringing more realistic wrinkle effect.4)A realistic garment display method for customized virtual characters is proposed.For oscillate effect caused by elastic force in the process of garment simulation,a linear damping force model is introduced on the basis of the finite element model.This model can accurately and stably simulate the anisotropic tensile behavior of various materials.Using the tensile measurement data and parameter optimization method,the stiffness parameters of the elastic model are obtained.The shape and posture of the human body are recovered from the single-view image,and the human body with the 3D face are synthesized for detail display.The experimental results show that this method can quickly restore a reasonable virtual character,wearing different fabrics could significantly increases the fidelity of the virtual garment.5)An optimized flattening method for energy-based fabric deformation model is proposed.Considering the energy associated with tensile or shear deformation in real woven fabrics,an energy-based fabric deformation model is proposed to update the energy distribution for determining the best 2D pattern.A unified axis coordinate system is created to obtain deformation energy.And local flattening deformation is released by the energy relaxation method.The experimental results show that a complex surface such as garment could achieve better flattening effect.Compared with other flattening methods,this method is more flexible and achieves a better flattening results.