Font Size: a A A

Applied Research Of The Hierarchy Sensing System With Deep Learning To The 3D Garment Design

Posted on:2019-04-23Degree:MasterType:Thesis
Country:ChinaCandidate:H L DunFull Text:PDF
GTID:2321330566952418Subject:Mechanical and electrical engineering
Abstract/Summary:PDF Full Text Request
Garment is an important part of human social culture and an important manifestation of human spiritual civilization.In recent years,with the rapid development of society,economy and culture,people's demands for garment has not only remained in the aspects of warmth retention and fit,but also the pursuit of individuation and artistry.There are limitations in large measurement error,long production cycle and low output of the traditional tailored garment manufacturing industry,which obviously can not meet the needs of apparel mass-customization.The development of computer graphics,artificial intelligence,virtual reality and other disciplines has promoted the transformation of traditional digital garment technology to three-dimensional technology,which is the inevitable trend and result of computer aided garment design technology.In view of the key technology and research difficulties in 3D digital garment design,this dissertation focuses on the feature recognition of the human body and the digital modeling of 3D garment.The neural network with deep learning is first employed to recognize the shape of 3D human body,which is followed by threshold discrimination to obtaine the feature line.The feature points are extracted from the feature line to fit the human body girth curve,and the human body size is accordingly measured.On this basis,the algorithm that is devoted to constructing 3D garment prototype based on the garment human body feature points is then proposed.The 3D garment prototype is divided into a number of simple subdivision surfaces.The subdivision surface is in the form of triangulated blocks,and the 2D expansion is subsequently carried out to form the garment pieces The local area editing of the 3Dgarment prototype is implemented through the superquadrics in order that the designated garment can well fit the human body figure.Finally,the collision detection and collision avoidance are realized using AABB bounding box,and the dressing simulation is realized by the mass-spring system model with Euler solutions.
Keywords/Search Tags:Convolutional Neural Network, Costume body features recognition, 3D garment prototype modeling, Superquadric surface, Digitally dressing
PDF Full Text Request
Related items