Font Size: a A A

Hair Modeling And Simulation Based On A Single-view Picture

Posted on:2021-03-05Degree:MasterType:Thesis
Country:ChinaCandidate:B B ZhouFull Text:PDF
GTID:2428330614457272Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
The modeling of hair is too difficult to simulate because of its number and shape,as well as the texture of the hair itself.The traditional way of constructing hair based on physics and geometry requires complex calculations and various parameters.In recent years,hair modeling methods based on single images,based on multiple images,and based on videos have begun to develop,attaching some database.The advantage is that modeling is fast.At present,the geometry of the hair is mainly represented by lines of three-dimensional points.In this paper,the hair modeling and simulation from a single view technology of 3D virtual humans is studied.Aiming at the problems of manual intervention and long time during the modeling process,and elasticity during simulation,a hair modeling and simulation system based on a single image is designed.A three-dimensional strip is used to represent the geometry of the hair.The main research content of this paper includes four parts: hair modeling,hair fitting,dynamic simulation,collision detection.A similar hair model of the image is automatically constructed by inputting a picture of a human front,the selected hair model and head model are connected.Then we simulate dynamic hair by setting gravity,friction,collision detection,and other more,and the corresponding physical characteristics are given to the model.The model preserves the image appearance of the image as much as possible and can be used to simulate common hair geometry.The specific research contents include:(1)using a three-dimensional multi-strip to represent the geometry of the hair and build a small hair model database that is more suitable for real-time rendering and integration with existing game engines.(2)Hair type classification.Proposing a fully automatic classification framework for hair attributes that combines deep learning with traditional classification,which captures the shape of the hair from a single image and matches the corresponding model from the database,as well as the classification of the hair texture.(3)air model fitting.To ensure the output of high-quality hair model mesh,we propose a multi-strip fit optimization algorithm to prevent crossover and keep shape.(4)Hair dynamics simulation.using the particle-spring model,which mainly involves the problems of various force analysis,motion simulation,collision detection,and super-elasticity of the hair model.The experimental results show that the hair modeling and simulation based on a single-view method studied in this paper can more realistically build a character's hair model,and the hair movement conforms to the common sense of physics.It has good expansion application value.
Keywords/Search Tags:hair simulation, particle-spring model, image-based hair modeling, hair dynamics simulation
PDF Full Text Request
Related items