英譯中2--有關(guān)服裝的,
Figure 1:Virtual human with feature points andbody segment hulls for neck,arms andlegs.A binary relationisLyingOn cp × bsh.Informally this relation is expressing that(the projection of) a certain pattern ismainly lying on a certain body segmenthull.The formal interpretation requires thealgorithmic methods that are described be-low.The necessary information to build these rela-tions can be extracted algorithmically from theCAD information of the cloth patterns and theannotations that we give below.In the followingsubsection we also show how the precise “geo-metric” interpretation of these properties has tobe obtained.2.2.1 Information on cloth patternsWithin a CAD system garments are representedby their two-dimensional cloth patterns.Thebounding curves of the patterns are given by asequence of two-dimensional vertices.In addi-tion to the bounding curves,there must be givensome sewing information.This is done by a setof seams that indicate,which part of one pat-tern has to be sewn together with correspondingparts of other patterns.Note that the two corre-sponding curves of a seam need not have neces-sarily the same length.In Figure 2 the bound-ing curves and sewing information of a femalejacket are shown.So the relation isDirectlySewedWith isan obvious coarsening of this CAD informationthat can easily be extracted algorithmically fromthese CAD information.Figure 2:Cloth patterns of a jacket and thesewing information.For interaction-free pre-positioning of clothpatterns additional information is needed.Forexample,a skirt can be rotated around the hipand you may not be able to determine the in-tended positions.Thus you need the positionand orientation in relation to the human body ofat least one pattern to determine the right posi-tions of the cloth patterns.This is done by giv-ing a two-dimensional position of a nearby fea-ture point in relation to this pattern and an up-vector which points in the direction of the prin-cipal axis of the appropriate body segment.For our pre-positioning algorithm,we requirethe position of one feature point for one patternper body segment.Clearly,this information isindependent from the size of the human figureand thus must be stored only once.Figure 3shows the structure of a piece of clothing con-taining all the information that is required forthe pre-positioning.From the garment data structure the prop-erties liesOn and isLyingOn can be ex-tracted.The pre-positioning algorithm is de-scribed in more detail in [14].It gives a rea-sonable geometric interpretation to this prop-erties:the pre-positioning and a physics-based“end positioning” yield the geometric shape ofthe collection of virtual garments with respectto a specific 3D virtual human.3