A Study Of 3D Human Deformation Based On Parametric Estimation

With the continuous development of the Internet and computer technology,e-commerce for clothing has been evolving and changing people’s shopping habits.Online purchases of clothing have become mainstream,and virtual try-on technology has become a new form of the clothing industry.3D human body modeling and human body size measurement are the main research topics in the virtual try-on industry.Traditional virtual dressing rooms often require various equipment or can only display 2D wearing effects,and there are still shortcomings in convenience,effectiveness,and efficiency.Therefore,this study proposes a 3D human body deformation reconstruction technology based on parameter estimation,which uses 2D images as input,obtains human parameter information through images,and applies it to 3D human body deformation to obtain a human model that is similar to the user’s body shape.The work of this article is as follows:(1)Proposed a human body image feature point detection based on H-ASM algorithm.Based on the method of active shape model to locate human feature points,this article proposes the H-ASM algorithm,which detects feature points by establishing a global shape model and fusing the local texture model of Hermite features using a multi-resolution search algorithm.The experimental results show that this method improves the localization accuracy of human body image feature points.(2)Proposed a human body 3D size prediction based on GBWO-ENN.This article optimizes the position update of α-wolf in grey wolf optimization algorithm by fusing the movement method in the black widow optimization algorithm,optimizes the convergence factor,and proposes a strategy to update the population according to the position level.GBWO algorithm is proposed to improve the global optimization ability and reduce the probability of being trapped in local optima.Then,the optimal value output by the GBWO algorithm is used as the initial weight and threshold of the Elman neural network,and GBWO-ENN is applied to human body size prediction.The experimental results show that this model speeds up the convergence speed and reduces prediction errors.(3)Designed and implemented a parameterized 3D human body deformation system.Based on the parameterized 3D human body reconstruction technology,the human body image feature point detection technology based on H-ASM algorithm and the human body 3D size prediction method based on GBWO-ENN are used to develop a 3D human body deformation system,and system testing is carried out.