New VR technologies have been developed, which form the basis for a realistic, three dimensional, (real-time) simulations and visualisation of individualised garments put on by virtual counterparts of real customers, reveals O L Shanmugasundaram.

Clothing manufacturers require systems with the capacity to deal with garment design process as a whole, including possibilities to work directly within a 3D graphic environment. In such a context, the aim of the author's research work has been the development of a graphic and interactive environment to design men garments and simulate their behaviour according to fabric properties and manufacturing processes. The system should permit to predict the real garment behaviour acting on the parameters characterising its physical model in order to reduce the number and the role of the physical prototype.

This kind of tool is not to be confused with other research prototypes or systems currently available on the market (eg, MAYA? Alias/Wave front), since they are mainly oriented to animation and applications, such as movies, cartoons, or virtual catwalks. Figure 1 shows the reference architecture of a 3D system for garment design and simulation. One can distinguish two modi operandi according to garment types, design process, and strategies adopted within manufacturing companies.

The first one involves drawing 2D patterns by using 2D CAD systems. Modules have to be developed to generate, starting from 2D patterns, data necessary to define the 3D physical model of the garment and execute the simulation. The latter, more innovative, consists of the design of 3D garments around a digital mannequin. This technique is more natural and creative, because it allows direct transference of the garment concept.

The mode list uses 3D graphic tools, the 3D Modeler and the 3D Simulator, to both design and simulate the garment. The 2D patterns are automatically generated as well as data for assembling and positioning 2D patterns around the mannequin to define the physical model. The author followed the second approach, even if different solutions have to be adopted for men and women garments.

Virtual try-on

In the course of the project Virtual Try-On, innovative VR technologies have been developed, which form the basis for a realistic, three-dimensional, (real-time) simulation and visualisation of individual customers and garments. Utilising these VR techniques, an integrated virtual shop infrastructure is provided, which facilitates the presentation and trade of individualised garments at the point-of-sales and soon over the Internet. Instead of replacing the current shopping experience (eg, really touching garments and materials), Virtual Try-On, in fact, aims at enhancing customer support and decision making through extending corresponding customer services.

The following scenario describes a typical way a customer will experience the VTO shopping environment at the point of sales.

Types

There are basically two methods for the simulation of cloth: Geometrical models and physically-based models.

Physically-based Cloth Simulation in Virtual Reality (WSI/GRIS)

The physically-based cloth simulation is responsible for sewing the pre-positioned garments along the seam lines and for computing the drape of the clothes on the avatar. For Physical Model and Numerical Solution to compute realistic animations of clothes, the author developed an efficient model based on finite elements for viscoelastic, highly flexible surfaces. It is particularly designed for numerically stiff materials such as textiles because it yields linear equations in each time step and allows fast time stepping in an implicit integration method. This is achieved by reducing the nonlinear elasticity problem to the planar, linear case in each step.

With this model, one can assemble garments from CAD cloth patterns, seam these together, and animate the cloth in dynamic scenes with any chosen material properties (see Figure 10). This results in a physically accurate and also fast simulation. The basic idea in this approach is to use a linear strain formulation and to construct a rotated rest state for each element. The arising ordinary differential equations time are solved by an implicit Euler method.

Textiles show very different physical behaviour in weft and warp directions, so we model elastic and viscous material parameters for the two directions independently. Material measurements are carried out with the Kawabata evaluation system for the two Young moduli, the shear modulus and the Poisson number, which controls the transverse contraction. Additionally, the bending moduli describe the curvature elasticity in the weft and warp directions. In order to model the exact hysteresis effects of the corresponding tissue, dynamic material parameters are measured with the Kawabata and Zwick systems.

Geometric modeling of garments

As pointed out, the standard way of designing realistic virtual simulations begins with the definition of all the necessary 2D fabric patterns, thus requiring knowledge of tailoring that is not necessarily possessed by most computer artists. The users then need to specify sewing constrains for the patterns. This tedious process is made even more complex by the need to set adequate values for a set of physical parameters and run a simulation to obtain the final shape for the garment, even then the character is at rest.

The alternative is to model the garments directly in 3D, as proposed by, who developed a sketch based garment modeling tool. However, using direct modeling in 3D, the generated shape is typically not piecewise developable, and thus depicts a garment, which is not physically plausible. Moreover, the modeled garment is unlikely to depict fold patterns specific to real garments in any rest position. The author's work enhances this alternative approach by making it usable as a starting point for modeling the static shape of realistic garment. In particular this method approximates the garment surface mesh obtained from mask etch by a piecewise developable surface.

The developed system

1. The 3D Modeler that possesses all the necessary functionality to edit (Editor Module) men garment according to traditional designer's working method and to automatically generate (Export Module) data for 2D patterns representation, necessary for garment simulation and manufacturing.

2. The 2D CAD system used to produce, using above mentioned data, the 2D patterns of the 3D designed garment.

3. The 3D Simulator that consists of three main modules: The Module to generate physical model for defining the particle model of 2D panels according to fabric properties, the Module to assemble patterns for sewing and assembling 2D patterns over the mannequin (ie, the initial 3D configuration), and, finally the Solver for executing the dynamic simulation.

4. The User Interface that proposes a unique environment for both the Modeler and the Simulator. It has been designed to be as close as possible to traditional design process and based upon the designer's knowledge.

3D Modeler

The design process, currently followed within companies, has been analysed in order to extrapolate functionalities of the Modeler. As reference garment, we selected a jacket since it is considered the most representative men cloth. Typically, the designer creates a new style by modifying the shape of a physical prototype, eg, changing sleeves length or tightening the waist, according to fashion trends and stylist's sketches. To do this, the designer uses reference elements: Sewing lines, significant and structural elements, such as waist or shoulders.

Therefore, the Modeler had to permit the designer to operate in the same way using a digital prototype instead of a physical one. This required the acquisition of the mannequin and jacket geometry to be used as reference shape from which derives new ones by interactive changes. The Editor Module has been implemented using and combining MAYA Deformers [Maya00], which enable the user to change the shape of a geometric model. From the analysis of modelist's modus operandi, the author have identified and implemented a set of modifiers that permits the designer to perform traditional modifications. Some examples are: Shorten/lengthen sleeves, tighten/enlarge shoulders, tighten/enlarge waist, shorten/lengthen jacket. For each type of modifier, we have identified the interaction style and allowable range of values.

3D Simulator

As already said the 3D simulator is based upon the particle-based approach. As shown in Figure 2, it possesses all the necessary functionality to: ?

• Import the mannequin used by the modelist to design the 3D garment.

• Automatically generate the particle model of 2D panels according to material properties.

• Position and assemble 2D patterns over the mannequin using data generated by the 3D Modeler and acquired by digitalisation.

• Execute the simulation.

The prototype proceeds according to the following steps:

1. 2D garment patterns are discretised according to the particle-based approach and warp-weft directions.

2. Forces among particles and corresponding parameters value are established on the basis of fabric mechanical characteristics.

3. 2D patterns are sewed using reference lines and assembled over the mannequin in order to reach an initial configuration.

4. Garment's behaviour is simulated.

Real-time animation of garments

The real-time animation has been integrated into the garment design tool to provide a preview of the clothes. Imated body and a garment whose rest shape has been computed with the garment design tool, this module provides the real-time visualisation of the garments on the animated body. This module can be used at anytime during the making of the clothes helping the designers to judge the quality of their clothes.

Garment preprocessing

Simulating garments in real-time requires drastic simplifications of the simulation process to be carried out, possibly at the expense of mechanical and geometrical accuracy. The approach, described by Cordier et al, is based on a hybrid method where the cloth is segmented into various sections where different algorithms are applied.

When observing a garment worn on a moving character, the authors notice that the movement of the garment can be classified into several categories depending on how the garment is laid on, whether it sticks to, or flows on, the body surface. For instance, a tight pair of trousers will mainly follow the movement of the legs, whilst a skirt will flow around the legs. Thus the authors segment the cloth into three layers that we define as follows (Figure 11):

• Layer 1 (Stretch cloth): Garment regions that stick to the body with a constant offset. In this case, the cloth follows exactly the movement of the underlying skin surface.

• Layer 2 (Loose cloth): Garment regions that move within a certain distance to the body surface are placed in another category. The best examples are shirtsleeves. The assumption in this case is that the cloth surface always collides with the same skin surface and its movement is mainly perpendicular to the body surface.

• Layer 3 (Floating cloth): Garment regions that flow around the body. The movement of the cloth does not follow exactly the movement of the body. Collisions are not predictable; For a long skirt, for instance, the left side of the skirt may collide with the right leg during animation.

These three categories are animated using three different cloth layers. The idea behind the proposed method is to avoid the heavy calculation of physical deformation and of collision detection wherever possible, ie, where collision detection is not necessary. The main interest of the approach is to pre-process the target cloth and body model so that they are efficiently computable during runtime. The garments are divided into a set of segments and the associated simulation method is defined for each.

The segmentation process dispatches each garment region to its adequate layer. From the garment in its rest shape on the initial body, the distance between the garment and the skin surface is used to determine to which category the cloth triangles belong. Associated with each segment are distances from the skin surface that are used to determine the category. Each segment falls into one of three categories: Tight, loose and floating clothes. Cloth vertices that are located closely to the skin surface belong to the first or the second layer. Cloth vertices that do not collide with any skin surface belong to the third layer.

Garment animation

The techniques for real-time clothes simulation are proposed. Garment vertices are animated with three different methods, depending on which layer they belong that is defined during the pre-processing stage. Tight clothes in Layer 1 follow the deformation of the underlying skin. These deformations are calculated thanks to the mapping of the attachment data of the skin to the garment surface.

For Layer 2 that is composed of loose clothes, the relative movements of clothes to the skin remain relatively small, keeping a certain distance from the skin surface. Consider the movement of sleeve in relation with the arm: For a certain region of the garment, the collision area falls within a fixed region of the skin surface during simulation. With this in mind, the scope of the collision detection can be severely limited.

A basic assumption made is that the movement of the garment largely depends on that of the underlying skin and yet it should not follow the skin surface rigidly. It is necessary to simulate the local displacement of the garment from the skin surface. Two different methods have been developed; One for cloth deformation on the limbs (trousers and sleeves), and the other one for the deformation of cloth on the trunk.

Cloth vertices on the limbs are enclosed in half spheres that are attached to the skin surface. Vertices inside these spheres are displaced with the equation of the rigid body motion. A function defines the diameter of the spheres depending on the relative position of the cloth vertex to the normal of the skin surface. Cloth vertices located on the trunk are animated with a rough mesh. This rough mesh is animated with a physic-based method. The cloth mesh is deformed with the FFD method using the position of the vertices on the rough mesh (Figure 13).

Layer 3 is composed of vertices that freely float around the body. This will take care of cases, such as a large skirt floating around the legs. Any part on this skirt can collide with any part of the leg. Downloaded from the Web server, the Web Client application contains the body/garment sizing module and the real-time garment simulation module. The body/garment sizing module provides functionalities to deform the 3D mannequin from the customer's input body size and resize the selected garment accordingly. Once the garments and the mannequin have been adapted to the customer's measurements, the animation of the dressed mannequin is taken care of by the real-time garment simulation module. The simulation of this layer uses a classical approach with particle system and collision avoidance.

Visual results

Some visual results obtained using the system are presented in this section. Figure 6.1 shows a woven cloth emphasising the difference between the front and back sides. Figure 6.2 shows the deformation of the loops of a knitted cloth. Limitations and future work

Dynamic animation of garments was outside the scope of this paper. However, Virtual Garments: A Full Method could be used in two different ways toward this goal. The 2D patterns that is generated can be used by a cloth animation system to compute the rest lengths of the springs that model the cloth material. Thus, the garments designed can be animated using standard techniques.

Another option would be to take advantage of the procedural modeling of fabric folds during animation. Then, only the control mesh would need to be animated using physically-based simulation, while fine details such as folds, are costly to simulate. They are caused by stiff, buckling phenomena would be added procedurally at no cost prior to rendering. The author plan To explore this approach in the near future.

Another possible application of the work is the prototyping of real garments. A fashion designer could use the sketching system to quickly sketch some clothing, automatically get different 3D views of the garment, edit the model as necessary, and finally print the corresponding 2D patterns. This would require a number of enhancements to the method, such as introducing gussets and enabling axial symmetry specification.

Conclusion

An efficient technique for cloth simulation has been presented. It uses a mass-spring model with velocity and force modification methods for coping with the super-elasticity and for resolving responses to collisions. As a result, the speed is good enough for almost real time simulation of draping virtual garments on animated virtual humans. The body surface is acquired through a 3D scanner, which makes the approach very accurate and allows users to try on different sizes of clothing and see how they fit them. So far the author conducted experiments on static and very crudely animated human bodies. Future plans are to implement dynamic simulation of garments on naturally animated virtual actors using the presented approach. Animation will be done by applying motion tracker data and skin deformation techniques to the static body. The author will evaluate the efficiency and compare it to existing techniques.

O L Shanmugasundaram. Lecturer with the Department of Textile Technology, K S R College of Technology, Tiruchengode, Tamil Nadu 637 215.