论文部分内容阅读
Abstract. This study shows a novel three-dimensional (3D) parametric body model development using cross-section control and control algorithm retrieved from anthropometric survey. The 3D parametric body model was shaped into the most common body shape of the young Hong Kong female and be able to automatically change its critical body shape according to the user’s critical body dimension inputs. The control algorithm controls the profile of the parametric model is retrieved from an anthropometric survey using 3D scanner to study the profile change of body and the relation between some critical body dimensions. Compared to the traditional anthropometric surveys, the 3D body scanner provides more accurate body dimension and information as well as new body shape measurements.
Key words: Parametric, body model, anthropometric.
1. Introduction
Realistic virtual human models are rapidly becoming common place in many fields of industry. These models, often adopted in various areas such as garment drape simulation [1], female figure identification [2] and virtual garment design [3], is an essential part in these applications, which provides garment designers with a virtual three-dimensional body model for them to refer their designs to, for try-on and to demonstrate their products in virtual environment. Nicola D’Apuzzo [4] reviewed the existing market usage of 3D body scanning technology for fashion and apparel industry, virtual try-on has became possible for the industry. Commercial systems like Gerber, Lectra, and Pattern aided design system are some of the widely-used garment design computer aided design(CAD) software which relies on the virtual human models to visualize the garment design. Another point is that, online visual merchandising is showing a very promising future. Ref. [5] shows that in the past few years, e-commerce has been taking place of the traditional shopping channel. In 2008, online buying has reached an estimated 63 million US households, while the reported apparel sales were higher than other products. But the online apparel shopping is still facing obstacles, 85 percent of the online women purchasers avoided buying apparel because of the inability to try on an item for size or fitting [6]. With the help of this parametric body model, online purchasers can indicate the robotic mannequin to change into their own body shape, then “try on” the interested item on the robotic mannequin and “see” the performance through a webcam. However, there are some problems in body model that the body scan data is not in a feasible format for any commercially available CAD system, and whenever a different size body is required, a real human body of that specific size has to be scanned and the whole body model building process has to start over again [1, 7]. It will be relatively costly. A further consideration is that the 3D body model data size can reach more than 10 MB, bringing inconvenience in data storing. Some researchers has been working on the parametric body model [1], for new approaches to build the 3D body model. First, the original shape of the parametric body model is generated from a 3D body scan data, with a set of parametric surface and deformation algorithm which can easily change into different body sizes. Moreover, it can integrate with CAD and computer aided design (CAM) software. However there are still limitations for the parametric body model. Firstly, how precisely the deformation algorithm can describe the body shape change. Although the original body model is built according to the real human or dummy scan data, the accuracy of the deformation algorithm is not approved against any anthropometric data, in other words, it is not guaranteed that the parametric body model is changed according to the real human body shape variety. Secondly, different races, different age groups of human has their unique way of body shape variety, which should be studied before setting the deformation algorithm.
This paper presents a new approach to build a 3D parametric body model. Firstly, study the various bodily profiles of the human body, obtained in the form of data clouds and avatars are extracted from body scanners database. Secondly, the morphological and topological profiles of the body parts that are used in the 3D model design are studied in the digital format. Thirdly, these digital body model is built using common CAD software for easy integrate into other software. The paper is organized as follows: Section 2 introduces the basic 3D body model profile development; section 3 presents the modification of the basic 3D body model profile based on the anthropometry survey results; section 4 discusses the deformation algorithm development; section 5 gives conclusions.
2. Basic 3D Body Model Profile Development
The shape of the 3D parametric body model was developed with the software Solidworks. It was controlled by 14 specific cross sections (See Fig. 1). There are three steps to construct the body model. First, a basic body shape is obtained using the point cloud data of a normal size 12 full body from using the body scanner. Then filter the noise in the point cloud date. Connect the remaining points using B-spline method to generate the basic curves of the cross sections (See Fig. 2). Then, the second step is modifying the basic body model according to the scan data and setting the driving points and driven points on every cross sections which determines the shape of curves and then forming the whole body model profile. Last step is the development of the deformation algorithm. There two steps will be discussed in the following sections.
3. Basic 3D Body Model Profile Modified according to the Anthropometry Survey
The parametric body model is designed to be able to change itself into the desired body shape configuration accurately, so it is essential to develop a comprehensive understanding of body shape variations. 130 Hong Kong females aged 17-22 were scanned using the TC2 body scanner. Participants were recruited for the anthropometric study launched by the Institute of Textiles and Clothing of The Hong Kong Polytechnic University. In the recruitment process, participants were chosen with reasonably symmetrical body proportions without any obvious deformities or postural problems caused by neurological or musculoskeletal diseases. Participants were scanned in a light color close-fitting undergarment, which allows 3D body scanner to capture the body shape while avoiding deforming it. Participants were positioned in an erect but relaxed posture, with arms and legs abducted slightly so that the cameras in the scanner could capture the full torso. The scan process finished in 10 seconds, and then the raw data was processed into the file format .obj and .rbd by the 3D body scanner software as the output. Then the critical measurements are extracted and analyzed to determine how to modify the profile of the basic body model.
The parametric body model needs a standard shape which allows all customized shapes to transform to. The standard body shape is set according to the average critical body measurement resulted from the 130 subjects. The critical body shape dimensions were selected according to the garment construction dimensions, following that of ISO 8559, and statistical analysis was applied. Because these dimensions are critical dimensions in garment design and the major dimensions change of the parametric body model will take place at these areas, these dimensions and the cup size are set to be the driving dimensions. The selected dimensions include: neck girth, bust girth, waist girth, abdomen girth, hip girth, bust width, neck to bust vertical length, bust to waist vertical length, waist to hip vertical length and shoulder length. Fig. 3 shows the distribution of the selected dimensions.
Bust size is an important and essential dimension for the women’s foundation garments, anthropometric and figure classification. Bust and bra classification follows the ISO DIS 4416 standard, and the bust size distribution analyses results are shown in Table 1.
4. Deformation Algorithm Development
The deformation algorithm is used to control the model profile changing. It was observed several body parts show variation and retrieving the relationship between different body parts to ensure the body model changes more realistically to the human form and shape.
As the body size grows, both the front and the back part of the girths are increasing, but at different rates. With the help of the 3D body scanner, the point cloud body model was generated. Then it was split into the front part and back part using a plane which is from the two side waist points intersecting the crotch point. Then the front half waist and hip girth are compared to the back part respectively. Yf = -0.0201Xh + 1.1716 (1) Yb = 0.0013Xw + 0.8356 (2) where the Xh is the full hip girth; Xw is the full waist girth; Yf is the value which is front hip girth divided by back hip girth and Yb is the value which is front waist girth divided by back waist girth.
The result indicate that as body size increases, the front part of the waist is increasing faster than the back part, while the front part of the hip is increasing slower than the back part. When the parametric body model is changing its size, the cross section shapes change should follow these rules.
The relationship between different body dimensions is also critical to the simulation of the body shape variation. When a dimension changes, it normally will affect the other body dimension as well. For example, when the hip girth becomes larger, the abdomen girth will increase as well. To gain a further understanding of the relationship between these critical dimensions during size changing, the correlation relationships between several critical dimensions are figured out. A sample correlation coefficient (r), sometimes referred to as a Pearson Product-moment correlation, is a measure of the strength of the linear relationship between two variables. If we have a series of n measurements of x and y written as xi and yi where i = 1, 2, ..., n, r is rewritten as (1),
=???∑∑∑∑∑∑∑(3)
Its value ranges from -1.00 to +1.00, with -1.00 representing a perfect negative linear relationship (as x increases, y decreases), 0.00 representing a total lack of linear relationship, and +1.0 representing a perfect positive linear relationship (as x increases, y increases) between the variables. From Table 2, the results are similar to the correlation coefficient analyses results of anthropometric survey of US Army (female) in 1988[8], indicating these relationships are not only true of Hong Kong young females but also American young female. With the correlation relationship found, it is able to determine how the driven dimensions change
following the driving dimensions of the parametric model. There are two driven dimensions; the scye circumference and under bust girth.
5. Conclusions
This paper presents a parametric body model development which is based on the anthropometric survey information of the profile and human body variation. It can automatically change its size by inputting critical body dimensions and providing a realistic profile and variation supported by the anthropometric survey result. The outcome of the study provides valuable anthropometric data and a 3D paramedic body model that can be used in garment drape simulation, female figure identification and virtual garment design.
References
[1] S. Kim, C. Park, Parametric body model generation for garment drape simulation, Fibers and Polymers 5 (2004) 12-18.
[2] K. Simmons, C.L. Istook, P. Devarajan, Female figure identification technique (FFIT) for apparel: Part I. Describing female shapes, Journal of Textile and Apparel, Technology and Management 4 (2004).
[3] C.H.M. Hardaker, G.J.W. Fozzard, Towards the virtual garment: three-dimensional computer environments for garment design, International Journal of Clothing Science and Technology 10 (1997) 114-127.
[4] N. D’Apuzzo, H. Consulting, 3D body scanning technology for fashion and apparel industry, in: Proceedings of the SPIE, 2007, Vol. 6491, p. 64910O.
[5] L. Khakimdjanova, J. Park, Online visual merchandising practice of apparel e-merchants, Journal of Retailing and Consumer Services 12 (2005) 307-318.
[6] R. Greenspan, Apparel Sales Apparent Online, CyberAtlas, available online at: www.cyberatlas.internet.com/markets/retailing/print/0,60 61_2171361,00.html, 2003.
[7] C. Carrere, C. Isstook, T. Little, H. Hong, T. Plumlee, Automated garment development from body scan data, National Textile Center Annual Report I, 2000.
[8] N.U.E. Il, J. Cheverud, C. Gordon, R. Walker, C. Jacquish, L. Kohn, Anthropometric Survey of US Army Personnel(1988): Correlation Coefficients and Regression Equations, 1990.
Key words: Parametric, body model, anthropometric.
1. Introduction
Realistic virtual human models are rapidly becoming common place in many fields of industry. These models, often adopted in various areas such as garment drape simulation [1], female figure identification [2] and virtual garment design [3], is an essential part in these applications, which provides garment designers with a virtual three-dimensional body model for them to refer their designs to, for try-on and to demonstrate their products in virtual environment. Nicola D’Apuzzo [4] reviewed the existing market usage of 3D body scanning technology for fashion and apparel industry, virtual try-on has became possible for the industry. Commercial systems like Gerber, Lectra, and Pattern aided design system are some of the widely-used garment design computer aided design(CAD) software which relies on the virtual human models to visualize the garment design. Another point is that, online visual merchandising is showing a very promising future. Ref. [5] shows that in the past few years, e-commerce has been taking place of the traditional shopping channel. In 2008, online buying has reached an estimated 63 million US households, while the reported apparel sales were higher than other products. But the online apparel shopping is still facing obstacles, 85 percent of the online women purchasers avoided buying apparel because of the inability to try on an item for size or fitting [6]. With the help of this parametric body model, online purchasers can indicate the robotic mannequin to change into their own body shape, then “try on” the interested item on the robotic mannequin and “see” the performance through a webcam. However, there are some problems in body model that the body scan data is not in a feasible format for any commercially available CAD system, and whenever a different size body is required, a real human body of that specific size has to be scanned and the whole body model building process has to start over again [1, 7]. It will be relatively costly. A further consideration is that the 3D body model data size can reach more than 10 MB, bringing inconvenience in data storing. Some researchers has been working on the parametric body model [1], for new approaches to build the 3D body model. First, the original shape of the parametric body model is generated from a 3D body scan data, with a set of parametric surface and deformation algorithm which can easily change into different body sizes. Moreover, it can integrate with CAD and computer aided design (CAM) software. However there are still limitations for the parametric body model. Firstly, how precisely the deformation algorithm can describe the body shape change. Although the original body model is built according to the real human or dummy scan data, the accuracy of the deformation algorithm is not approved against any anthropometric data, in other words, it is not guaranteed that the parametric body model is changed according to the real human body shape variety. Secondly, different races, different age groups of human has their unique way of body shape variety, which should be studied before setting the deformation algorithm.
This paper presents a new approach to build a 3D parametric body model. Firstly, study the various bodily profiles of the human body, obtained in the form of data clouds and avatars are extracted from body scanners database. Secondly, the morphological and topological profiles of the body parts that are used in the 3D model design are studied in the digital format. Thirdly, these digital body model is built using common CAD software for easy integrate into other software. The paper is organized as follows: Section 2 introduces the basic 3D body model profile development; section 3 presents the modification of the basic 3D body model profile based on the anthropometry survey results; section 4 discusses the deformation algorithm development; section 5 gives conclusions.
2. Basic 3D Body Model Profile Development
The shape of the 3D parametric body model was developed with the software Solidworks. It was controlled by 14 specific cross sections (See Fig. 1). There are three steps to construct the body model. First, a basic body shape is obtained using the point cloud data of a normal size 12 full body from using the body scanner. Then filter the noise in the point cloud date. Connect the remaining points using B-spline method to generate the basic curves of the cross sections (See Fig. 2). Then, the second step is modifying the basic body model according to the scan data and setting the driving points and driven points on every cross sections which determines the shape of curves and then forming the whole body model profile. Last step is the development of the deformation algorithm. There two steps will be discussed in the following sections.
3. Basic 3D Body Model Profile Modified according to the Anthropometry Survey
The parametric body model is designed to be able to change itself into the desired body shape configuration accurately, so it is essential to develop a comprehensive understanding of body shape variations. 130 Hong Kong females aged 17-22 were scanned using the TC2 body scanner. Participants were recruited for the anthropometric study launched by the Institute of Textiles and Clothing of The Hong Kong Polytechnic University. In the recruitment process, participants were chosen with reasonably symmetrical body proportions without any obvious deformities or postural problems caused by neurological or musculoskeletal diseases. Participants were scanned in a light color close-fitting undergarment, which allows 3D body scanner to capture the body shape while avoiding deforming it. Participants were positioned in an erect but relaxed posture, with arms and legs abducted slightly so that the cameras in the scanner could capture the full torso. The scan process finished in 10 seconds, and then the raw data was processed into the file format .obj and .rbd by the 3D body scanner software as the output. Then the critical measurements are extracted and analyzed to determine how to modify the profile of the basic body model.
The parametric body model needs a standard shape which allows all customized shapes to transform to. The standard body shape is set according to the average critical body measurement resulted from the 130 subjects. The critical body shape dimensions were selected according to the garment construction dimensions, following that of ISO 8559, and statistical analysis was applied. Because these dimensions are critical dimensions in garment design and the major dimensions change of the parametric body model will take place at these areas, these dimensions and the cup size are set to be the driving dimensions. The selected dimensions include: neck girth, bust girth, waist girth, abdomen girth, hip girth, bust width, neck to bust vertical length, bust to waist vertical length, waist to hip vertical length and shoulder length. Fig. 3 shows the distribution of the selected dimensions.
Bust size is an important and essential dimension for the women’s foundation garments, anthropometric and figure classification. Bust and bra classification follows the ISO DIS 4416 standard, and the bust size distribution analyses results are shown in Table 1.
4. Deformation Algorithm Development
The deformation algorithm is used to control the model profile changing. It was observed several body parts show variation and retrieving the relationship between different body parts to ensure the body model changes more realistically to the human form and shape.
As the body size grows, both the front and the back part of the girths are increasing, but at different rates. With the help of the 3D body scanner, the point cloud body model was generated. Then it was split into the front part and back part using a plane which is from the two side waist points intersecting the crotch point. Then the front half waist and hip girth are compared to the back part respectively. Yf = -0.0201Xh + 1.1716 (1) Yb = 0.0013Xw + 0.8356 (2) where the Xh is the full hip girth; Xw is the full waist girth; Yf is the value which is front hip girth divided by back hip girth and Yb is the value which is front waist girth divided by back waist girth.
The result indicate that as body size increases, the front part of the waist is increasing faster than the back part, while the front part of the hip is increasing slower than the back part. When the parametric body model is changing its size, the cross section shapes change should follow these rules.
The relationship between different body dimensions is also critical to the simulation of the body shape variation. When a dimension changes, it normally will affect the other body dimension as well. For example, when the hip girth becomes larger, the abdomen girth will increase as well. To gain a further understanding of the relationship between these critical dimensions during size changing, the correlation relationships between several critical dimensions are figured out. A sample correlation coefficient (r), sometimes referred to as a Pearson Product-moment correlation, is a measure of the strength of the linear relationship between two variables. If we have a series of n measurements of x and y written as xi and yi where i = 1, 2, ..., n, r is rewritten as (1),
=???∑∑∑∑∑∑∑(3)
Its value ranges from -1.00 to +1.00, with -1.00 representing a perfect negative linear relationship (as x increases, y decreases), 0.00 representing a total lack of linear relationship, and +1.0 representing a perfect positive linear relationship (as x increases, y increases) between the variables. From Table 2, the results are similar to the correlation coefficient analyses results of anthropometric survey of US Army (female) in 1988[8], indicating these relationships are not only true of Hong Kong young females but also American young female. With the correlation relationship found, it is able to determine how the driven dimensions change
following the driving dimensions of the parametric model. There are two driven dimensions; the scye circumference and under bust girth.
5. Conclusions
This paper presents a parametric body model development which is based on the anthropometric survey information of the profile and human body variation. It can automatically change its size by inputting critical body dimensions and providing a realistic profile and variation supported by the anthropometric survey result. The outcome of the study provides valuable anthropometric data and a 3D paramedic body model that can be used in garment drape simulation, female figure identification and virtual garment design.
References
[1] S. Kim, C. Park, Parametric body model generation for garment drape simulation, Fibers and Polymers 5 (2004) 12-18.
[2] K. Simmons, C.L. Istook, P. Devarajan, Female figure identification technique (FFIT) for apparel: Part I. Describing female shapes, Journal of Textile and Apparel, Technology and Management 4 (2004).
[3] C.H.M. Hardaker, G.J.W. Fozzard, Towards the virtual garment: three-dimensional computer environments for garment design, International Journal of Clothing Science and Technology 10 (1997) 114-127.
[4] N. D’Apuzzo, H. Consulting, 3D body scanning technology for fashion and apparel industry, in: Proceedings of the SPIE, 2007, Vol. 6491, p. 64910O.
[5] L. Khakimdjanova, J. Park, Online visual merchandising practice of apparel e-merchants, Journal of Retailing and Consumer Services 12 (2005) 307-318.
[6] R. Greenspan, Apparel Sales Apparent Online, CyberAtlas, available online at: www.cyberatlas.internet.com/markets/retailing/print/0,60 61_2171361,00.html, 2003.
[7] C. Carrere, C. Isstook, T. Little, H. Hong, T. Plumlee, Automated garment development from body scan data, National Textile Center Annual Report I, 2000.
[8] N.U.E. Il, J. Cheverud, C. Gordon, R. Walker, C. Jacquish, L. Kohn, Anthropometric Survey of US Army Personnel(1988): Correlation Coefficients and Regression Equations, 1990.