3D Human Body Modeling Based On Anthropometric Measurements

With the continuous development of the theory and application in Virtual Reality field,people pay more and more attention to the fidelity of virtual space.3D human modeling technology has become one of the most important research topics.The 3D human body model usually contains complex geometric information,which makes it difficult to construct human body models directly.Thus how to generate customized 3D human body models quickly and effectively has become the key research field of computer graphics,computer-aided design and so on.The existing customized human model generation methods include scanning model methods,creating model methods and reconstruction model methods.The reconstruction model method has gradually become the mainstream human modeling method for its realistic generated models,simple operation of users and diverse parameter forms.However,the reconstruction model method also has many disadvantages.For example,it costs longer time to generate the customized human body and the shape of generated model is dependent on the diversity of the example human shape space to a great extent In order to solve these problems,a new example-based method to generate customized human body model method called optimistic segmentation method is proposed in this thesis.And we simplify the number of input data,which improves the efficiency of model generation.In this thesis,the isoline of Morse function based on geodesic distance is used to extract the key measurements of example human body models after the registration of human point cloud model.According to the location of the key measurements,each example human model is segmented into 16 rigid regions.The relationship between the measurement pairs is analyzed by the linear correlation coefficient to recover the complete human measurement set from 7 input measurements,which simplifies the input data.Next,an algorithm called stochastic simulation algorithm is used to quickly establish the Multi-task Lasso regression model which shows the connection between 26 measurements and human shape parameters of 16 rigid regions.Then,the genetic algorithm is used to optimize the model,which makes the generated human body model fit the input data better and the optimization efficiency is high.Finally,experiment results show that the proposed optimistic segmentation method is able to accurately and quickly generate 3D customized model.The reconstructed human models are used as infantry battle formation in the damage simulations.