##
###
Parameterization and parametric design of mannequins

Charlie C.L. Wang

2005
*
Computer-Aided Design
*

This paper presents a novel feature based parameterization approach of human bodies from the unorganized cloud points and the parametric design method for generating new models based on the parameterization. The parameterization consists of two phases. Firstly, the semantic feature extraction technique is applied to construct the feature wireframe of a human body from laser scanned 3D unorganized points. Secondly, the symmetric detail mesh surface of the human body is modeled. Gregory patches
## more »

... e utilized to generate G 1 continuous mesh surface interpolating the curves on feature wireframe. After that, a voxel-based algorithm adds details on the smooth G 1 continuous surface by the cloud points. Finally, the mesh surface is adjusted to become symmetric. Compared to other template fitting based approaches, the parameterization approach introduced in this paper is more efficient. The parametric design approach synthesizes parameterized sample models to a new human body according to user input sizing dimensions. It is based on a numerical optimization process. The strategy of choosing samples for synthesis is also introduced. Human bodies according to a wide range of dimensions can be generated by our approach. Different from the mathematical interpolation function based human body synthesis methods, the models generated in our method have the approximation errors minimized. All mannequins constructed by our approach have consistent feature patches, which benefits the design automation of customized clothes around human bodies a lot. Keywords: human body, sizing dimensions, 3D scan data, feature-based modeling, and fashion industry. semantic features; however, recent developments find that only feature curves and points are not enough for modeling 3D digital clothes around human models -feature patches are actually required. In this paper, we begin the parameterization with the unorganized cloud points of human bodies obtained from 3D laser scanners. The entire human body is subdivided into a certain number of feature patches interpolating the given cloud points. The feature patches at the same location on different human bodies are correlated, where the detail geometric shape of a human model is given. The sizing parameters are given by feature curves. By the feature entities (feature patches, feature curves, and feature nodes), the geometry of a human body is fully registered and parameterized. Based on the parameterization of a human model, we can easily obtain its sizing dimensions through feature curves. The parameterized models are stored in a database using the sizing dimensions as searching criteria. After establishing the database, we can construct new models according to user input sizing parameters by 3 synthesizing selected models from the database. This leads to the problems of how to select example models and how to synthesize the models. In our approach, we select a number of the closest example models to the given sizes; then, a numerical optimization approach is applied to compute the weight coefficients of synthesizing the example models. Finally, the requested model is generated by interpolating the examples models with optimized weights. This is named as the parametric design of human bodies The target example data size of our digital human database is more than 5000 persons. Since human models born in different regions have different morphologic features, the samples from different regions are stored separately. The working flow of our approach is clearly illustrated by Fig.1. The parameterization has two phases: 1) the registration of feature wireframe -this is based on the feature extraction technique (the object under our consideration are of the same class, so the semantic feature extraction technique [1] is applied); 2) the modeling of surfaces according to feature wireframe and cloud points. After inputting sizing parameters to create a new human model, our synthesizer computes the weight coefficients for interpolation and interpolates selected examples by the weights -the parametric design result is then obtained. Optimization based synthesizer ... ... + ... ... Parameterization Parametric Design Fig. 1 The procedure of parameterization and parametric design of human models The major contributions of this paper are 1) an efficient feature-based parameterization technique for establishing a point-to-point corresponding among a set of human body surfaces with the same overall structure, 4 and 2) a numerical optimization based synthesis technique for constructing a new human body with the by specified sizing dimensions -the resultant model with approximation errors minimized. As the example models are parameterized, the synthetic result is certainly parameterized, which gives great benefits to the design automation of cloth products around human models. The rest of this paper is organized as follows: after reviewing related works in section 2, section 3 describes the necessary steps for registering a feature wireframe on the unorganized cloud points for a human body. In section 4, Gregory patch is adopted to construct G 1 continuous surface interpolating the feature wireframe, a voxel-based algorithm is applied to add details on the surface, and the human surface is made symmetric. The numerical optimization based synthesis algorithm for the parametric design of human bodies is given in section 5, where the strategy of choosing appropriate example models from database is also described. Finally, in section 6, the application for the design automation of customized clothes, which gains great benefit from the technique presented here, is demonstrated. Literature Review The human body modeling methodologies in literature can be classified into the creative approaches and the reconstructive approaches. Anatomically based modelers [4, 5] can simulate underlying muscles, bones, and generalized tissue. They fall into the creative category of human modeling approaches. The interactive design is allowed in the anatomy-based modelers; however, these modelers require a relatively slow production time. Recently, a lot of the reconstruction approaches has been investigated to build 3D geometry of human automatically by capturing existing shape [1-2, 6-9]. As mentioned by Seo and Magnenat-Thalmann [10], the disadvantage of these techniques is that it is very difficult to automatically modify the reconstructed models to different shapes following the user intends. Example-based shape modeling technique [10-13] is a good alternative to overcome this disadvantage. Our parametric design algorithm borrows some idea from the example based shape modeling. In the example-based shape modeling, all examples must have the same parameterization. Thus, our approach begins from the parameterization of a human model. Related to the parameterization of unorganized cloud points, Ma and He [14] presented an approach to shape a single B-spline surface with a cloud of points, their work is further enhanced on fitting a hybrid mathematical model of B-spline surfaces and Catmull-Clark subdivision surfaces to represent objects with general quadrilateral topology [15]; Barhak and Fischer [16] also presented a PDE based method about the parameterization for reconstruction of 3D freeform objects from laser-scanned data. Sienz et al. [17] developed a fitting technique to generate computational geometric models of 3D objects defined in the form of a point min 2 )) ( ( D X H s M − Ψ (7) Based on the above equation, 2 )) ( ( ] [ D X H X J s M 0 1 X J X J X J i i or the iteration number is greater than max N , where ] [ i X J is the value of the objective function in the ith iteration (current value), ] [ 0 X J is the value of the algorithm. If the optimization gives inaccurate result according to D (i.e., the iteration stops at the N max criterion and the returned value of ] [ X J exceeds some threshold), add m more models as the examples. The algorithm repeatedly selects models, and synthesizes models until the new model with accurate dimensions to D is obtained. The same as other example-based approaches, this approach also relies on the number of models store in Π . Thus, building a 3D digital human model database with a large number of models is quite an important work. To let the synthesized human model have the morphology features of geography, the human models born in different regions should be stored separately.

doi:10.1016/j.cad.2004.05.001
fatcat:u2ug2drvcbegtgjuimelo3xiya