Sage Journals: Discover world-class research

Abstract

This paper proposes a method for three-dimensional style modeling of loose sweaters. Through the correlation analysis of the three-dimensional human body and the classic sweater style, a style model was built on the three-dimensional human body model to realize efficient personalized sweater design and production. First of all, the design model was extracted from the human body model based on the characteristics of the ring-cutting algorithm. Secondly, the loose model of the sweater was established based on the chest, waist, and hip data of the human body. Subsequently, the feature line between the size information and style features was created, and curve interpolation values were combined with joint smoothing methods to generate a multi-faceted sweater style model. Finally, the mapping function was used to flatten the style model, the related styles were woven by operating the computer, and the suitability of the established sweater fabric was verified. The comparison results showed that the accuracy of the style construction of this model was improved. Through the analysis of experimental data, it can be proven that the method proposed in this paper can quickly and accurately establish a three-dimensional style model of a sweater, without the need for repeated measurements to make templates, thus saving development time.

Keywords

Sweater three-dimensional human body measuring characteristic parameter style model ring cutting algorithm

Introduction

When designing and producing garments, we must consider whether the size of garments suits the feature size and physiological characteristics of the human body so that people feel comfortable wearing these garments. Therefore, we should use the relevant theories of anthropometry for statistical analysis of the feature size of different human bodies to establish fit garment styles.^1
–3 There are two main applications of anthropometry. The first application is anthropometry-based parametric human body modeling, which is further applied in a virtual fitting context. On the basis of the parametric human body model, the garment pattern for constructing a feature silhouette is used by the feature design method. Moreover, the feature line of the parametric human body model is connected with the feature pattern of garments to realize individualized garments.⁴ In addition, artificial neural networks are used to test the predictive fit of 3D garments.⁵

The second application is clothing customization based on three-dimensional human body measurement. Since the structure of clothing and human features have commonalities, they can use the same set of feature definitions, that is, clothing style construction based on three-dimensional human body measurement data.^6
–8 Some scholars have compared the dressed human body model and the clean human body model through three-dimensional measurement to evaluate whether a garment fits or not.⁹ Others have recognized human feature points on the existing human body model through geometric modeling methods, obtained the human body profile curve, and established a fitness style model by adding related profile curves. The establishment of the target model is mainly based on the feature point method of three-dimensional human body measurement.¹⁰ In the latest research, three-dimensional scanning technology has been used to measure the lower limbs of the human body and build a digital style model of warp-knitted yoga pants.¹¹ In addition, some scholars have reconstructed the clothing model by controlling the key points in the clothing model.¹² The common feature of the above studies is that they mainly focus on model styles for tight-fitting clothing rather than loose sweaters. In the previous literature, there is a greater focus on style modeling for tight-fitting and body-hugging garments, with limited attention given to designing loose-fitting sweaters. Despite utilizing human body models for construction, the style characteristics of sweaters are not prominently addressed. This paper employs a 3D anthropometric measurement approach to construct sweater style models, which differs from the recent studies that utilize artificial intelligence methods to create clothing style patterns.

On this basis, a method based on 3D anthropometric characteristic parameters is proposed in this paper to build a sweater style model.

This paper first introduces the method for obtaining 3D human body data and producing 3D anthropometric characteristic parameters. Then, it further elaborates on the method for obtaining key points of the sweater style by the ring cutting machine method and generating the style model in combination with the surface interpolation method. This paper takes common sweater styles (e.g. round neck sweaters, sleeveless skirts, and half skirts) as examples to build the style model. Finally, it proposes the mapping algorithm to generate multi-style sweater patterns and perform machine weaving on the styles realized to verify the fitness of the styles established, as shown in Figure 1

Figure 1.

3D Sweater Garment Style Generation Method Based on 3D Anthropometric: (a) key points of sweater, (b) loose quantity distribution, (c) style area division of sweater, (d) area division of sweater, (e) sweater mapping, and (f) designed knitted sweaters using our method.

Related work

The features of the human body shape constitute the basis for style modeling. In accordance with the anthropometric theory, human body feature information is divided into three aspects of point, line, and plane. As an element of fundamental importance to garment parameterization, feature information is an important basis for sweater style modeling. The determined parameters were extracted from many sizes of the human body model. The corresponding styling features were modified after determining the style characteristic parameters to meet the requirements of the style model.¹³

Obtaining 3D human body data

Obtaining 3D human body data can be used to establish a realistic human body model with the data measured by the measuring system. However, there still is room for improvement in terms of rapidity and reusability due to huge data and complicated processes.^14
–17 The human body model applied to the sweater garment design system can also realize the parameterization of the human body model to affect the changes in garment styling of different sizes of the human body. Therefore, in this paper, the standard 3D female human body model is constructed using a 3D point cloud acquisition algorithm based on a depth camera¹⁸ and the 3D mesh editing software MeshLab.¹⁹ The initial mesh data of the 3D-scanned human body is obtained using a depth camera and a mesh reconstruction algorithm. The mesh issues, such as holes and overlaps in the initial mesh, are repaired using the 3D mesh processing software MeshLab to construct a usable 3D model.

Obtaining 3D human body feature points

3D human body feature points are important reference points for anthropometry and modeling. Human body feature points are automatically obtained by a method based on 2D images and depth map, or by a method based on 3D point clouds and 3D mesh.^20
–23 It is difficult for the method based on 2D images and depth maps to get accurate measurement data. However, for 3D mesh human body, human body feature points can be conveniently and rapidly obtained by the cutting ring method.^24
–26 In this paper, the 3D human body was evenly divided with a horizontal section, obtaining a 3D human body cutting ring and converting 3D body features to cutting ring features.²⁷ Therefore, 3D human body feature points were obtained by analyzing cutting ring features. In order to accelerate cutting ring generation and obtain even cutting rings, it is necessary to perform mesh uniformization on the 3D human body. The processed 3D human body underwent cutting ring processing from head to toe. Sections with an interval of 0.6 cm were selected according to the experiment to intersect with the human body to obtain the structural characteristics of the human body cutting ring, as shown in Figure 2(a).Based on our experience with three-dimensional human body measurements, a cutting distance of 0.6 cm can form sufficient cutting planes for the generation of feature points. This is within the allowable range of anthropometric errors for garment production and also ensures the computational efficiency of three-dimensional mesh cutting.

Figure 2.

Feature loops of the human model: (a) the human model slice loops, (b) feature points of the human model, and (c) girth sequence of the left arm.

Human body feature points were obtained through the following two steps after obtaining the cutting ring of the 3D human body:

The feature cutting ring where human body feature points were located was found using the number of cutting rings, girth and other cutting ring features.

Feature analysis was performed on the cutting ring where feature points were located to obtain the designated human body feature points as shown in Figure 2(b).

Figure 2(c) shows the bar chart for changes in girth of cutting rings at the left arm. The abscissa is the number of the girth of cutting rings from shoulder to finger, and the ordinate is the girth of the cutting ring. The minimum girth between two peaks is determined as the human body feature cutting ring according to the changes in the girth. The cutting rings at the left elbow and the left wrist can be found from left to right and set to $L_{e l b o w}$ and $L_{w r i s t}$ , respectively.

After obtaining the human body feature cutting rings by the two methods above, the human body feature points are extracted from the corresponding feature cutting rings. Table 1 illustrates the methods used to obtain human feature points.

Table 1.

Human body feature points and acquisition methods.

Feature Points	Human Body Position	Brief Description of Acquisition Methods
$p_{s h o u l d e r - l e f t}^{h u m a n}$ $p_{s h o u l d e r - r i g h t}^{h u m a n}$ $p_{a r m p i t - l e f t}^{h u m a n}$ $p_{a r m p i t - r i g h t}^{h u m a n}$	Shoulder Points, Underarm Points.	Starting from the top and moving downwards through the slicing rings, the commencement of the chest region was identified when three slicing rings appeared on the same plane. Shoulder and underarm points were derived based on these three slicing rings. As shown in Figure 3.
$p_{n e c k - b a c k}^{h u m a n}$ $p_{n e c k - f r o n t}^{h u m a n}$ $p_{n e c k - l e f t}^{h u m a n}$ $p_{n e c k - r i g h t}^{h u m a n}$	Neck Points	The slicing ring for the neck was determined using the method mentioned in Zhong’s article.²⁵ On the neck slicing plane, four neck points were obtained using horizontal rays that passed through the center of the neck slicing ring and aligned with the axis direction.As shown in Figure 4(c).
$p_{c h e s t - l e f t}^{h u m a n}$ $p_{c h e s t - r i g h t}^{h u m a n}$	Chest Points	In the slicing rings from the start of the chest to the waist, the point with the maximum value in the Z-direction was taken as the apex of the chest. Zhong’s article provides a detailed explanation.²⁵ As shown in Figure 4(a).
$p_{n a v e l}^{h u m a n}$	Navel Point	Among the slicing rings from the chest point to the hip, the one with the smallest circumference was chosen as the waist slicing ring. The navel point was derived from the waist slicing ring.As shown in Figure 2(b).
$p_{e l b o w - l e f t}^{h u m a n}$ $p_{e l b o w - r i g h t}^{h u m a n}$ $p_{w r i s t - l e f t}^{h u m a n}$ $p_{w r i s t - r i g h t}^{h u m a n}$	Arm Feature Points	The characteristic change in the circumference of the arm slicing rings was used to obtain these points, as shown in Figure 2(b).
$p_{l e g - l e f t}^{h u m a n}$ $p_{l e g - r i g h t}^{h u m a n}$ $p_{c r o t c h}^{h u m a n}$	Crotch Point Leg Points	Starting from the torso slicing rings, when two slicing rings appeared on the same plane, the initial slicing ring of the leg was identified. The crotch point and leg points were derived based on the initial leg slicing ring.As shown in Figure 2(b).
$p_{k n e e - l e f t}^{h u m a n}$ $p_{k n e e - r i g h t}^{h u m a n}$	Knee Points	Starting from the initial leg cross-section, the knee and ankle feature points were identified based on the changing characteristics of the leg cross-section circumference.As shown in Figure 2(b).

Figure 3.

Defining feature points based on slicing loops: (a) slicing loops of the armpit, (b) temporary points of the armpit, (c) slicing loops of the shoulder, and (d) armpit points and shoulder armpit.

Methodology

Obtaining key points of style

The conventional feature points of the human body obtained are key points at feature parts. The garment model established only by feature points cannot meet the requirements for style modeling. In order to facilitate the subsequent establishment of the sweater style model, key points of sweater style are established on the human body based on feature points, the requirements of style and structure and the principle of being able to fit the style characteristic curve. Key points are mainly divided into key points of the torso and key points of limbs. Combined with the classic sweater styles to select round neck long sleeve, knitted dresses and other styles to construct style key points.

Key points of the torso

According to style structure characteristics, after obtaining chest points $p_{c h e s t - l e f t}^{h u m a n}$ and $p_{c h e s t - r i g h t}^{h u m a n}$ , the center point between the breasts and the front and rear and left and right key points of the breast height are obtained, which helps establish garment chest. As shown in Figure 4(a), We obtain points $p_{c h e s t - r o o t L e f t}^{h u m a n}$ and $p_{c h e s t - r o o t R i g h t}^{h u m a n}$ by crossing the horizontal line of point $p_{m}$ with the contour line. Where $p_{m}$ is the center of point $p_{c h e s t - c e n t e r}^{h u m a n}$ and $p_{c h e s t - d e e p}^{h u m a n}$ . The point of maximum distance from $l_{0}$ on $c_{0}$ is found and used as $p_{c h e s t - s i d e L e f t}^{h u m a n}$ .Points $p_{c h e s t - s i d e R i g h t}^{h u m a n}$ , $p_{c h e s t - b a c k R i g h t}^{h u m a n}$ , $p_{c h e s t - b a c k L e f t}^{h u m a n}$ are determined by a similar method.

Figure 4.

Key points of the trunk: (a) key points of the chest, (b) key points of the chest Key points and auxiliary points of the chest, (c) key points of the neckline Key, and (d) points of the waist.

In order to form a fit chest style contour, in addition to generating sweater feature points on the chest cutting ring, it is also necessary to build auxiliary chest key points. Two striped chest cutting rings are constructed between the underarm cutting ring $L_{t o r s o}$ and chest cutting ring $L_{c h e s t}$ . The chest feature points are extracted from the chest cutting ring by extracting feature points from the chest contour line. The chest key points and auxiliary key points obtained are shown in Figure 4(b).

After obtaining 3D human body chest feature points, in order to construct multiple sweater neckline shapes, it is necessary to establish four key points ( $p_{n e c k - f r o n t}^{h u m a n}$ , $p_{n e c k - r i g h t}^{h u m a n}$ , $p_{n e c k - l e f t}^{h u m a n}$ and $p_{n e c k - b a c k}^{h u m a n}$ ) at the neck. The four points are obtained by intersecting the four rays passing through the center point of the neck cutting ring $L_{n e c k}$ with $L_{n e c k}$ , as shown in Figure 4(c). Besides the four neckline feature points at the neck, the feature points should also be established for the low-cut neckline, which should be set to $p_{c o l l o r - 0}^{h u m a n}$ , and $p_{c o l l o r - 1}^{h u m a n}$ . . . $p_{c o l l o r - 7}^{h u m a n}$ , where $p_{c o l l o r - 0}^{h u m a n}$ and $p_{c o l l o r - 1}^{h u m a n}$ are the positions of shoulder straps, and $p_{c o l l o r - 2}^{h u m a n}$ is the feature point on the right shoulder cutting ring line $L_{s h o u l d e r - r i g h t}$ and is obtained by intersecting with the ray passing through $L_{s h o u l d e r - r i g h t}$ center and in the direction of (0,0,1) with $L_{s h o u l d e r - r i g h t}$ . $p_{c o l l o r - 6}^{h u m a n}$ is the feature point on the left cutting ring line and is obtained by intersecting the ray passing through $L_{s h o u l d e r - l e f t}$ center and in the direction of (0,0,1) with $L_{s h o u l d e r - l e f t}$ . $p_{c o l l o r - 4}^{h u m a n}$ is the midpoint between $p_{c o l l o r - 2}^{h u m a n}$ and $p_{c o l l o r - 6}^{h u m a n}$ ; $p_{c o l l o r - 3}^{h u m a n}$ a point between $p_{c o l l o r - 2}^{h u m a n}$ and $p_{c o l l o r - 4}^{h u m a n}$ ; $p_{c o l l o r - 3}^{h u m a n}$ a point between $p_{c o l l o r - 2}^{h u m a n}$ and $p_{c o l l o r - 4}^{h u m a n}$ ; and $p_{c o l l o r - 3}^{h u m a n}$ and $p_{c o l l o r - 5}^{h u m a n}$ are used to adjust the width of low-cut neckline. $p_{c o l l o r - 7}^{h u m a n}$ is the feature point on the $L_{t o r s o}$ cutting ring and is obtained by intersecting the ray passing through $L_{t o r s o}$ center and in the direction of (0,0,1) with $L_{t o r s o}$ . $p_{c o l l o r - 7}^{h u m a n}$ is used to adjust the length of the low-cut neckline.

The waist point of the human body is the fulcrum of human body activities and the key point connecting the chest and buttocks. According to the fashion trend of styles, the position of the hem of the garment body will be between the waist and hip. The hem of relatively short sweaters is generally at the waist point. The hem of regular style is generally at the high hip. The hem of medium and long styles is at the hip point. Therefore, it is necessary to determine the position of the midpoint between the waist point and the hip point to adjust the height position of the hem. According to the analysis above, the feature points of sweater styles are extracted on the waist feature cutting ring, as shown in Figure 4(d). The cutting ring $L_{w a i s t - m}$ is obtained by plane cutting between 3D human body cutting rings $L_{w a i s t - h}$ and $L_{w a i s t - l}$ . The waist 3D sweater style structure can be constructed by finding the extreme points on cutting rings $L_{w a i s t - h}$ , $L_{w a i s t - m}$ and $L_{w a i s t - l}$ .

Key points of limbs

The sleeves of sweaters are generally based on the basic sleeve shape. The position of the sleeves is generally the length from the shoulder point to the wrist point. The position of the sleeve width requires obtaining the most protruding arm root point of the upper arm. Meanwhile, the shape of the front of the sleeve requires obtaining the most convex key points of the forearm. In order to obtain the arm root point, the following steps should be followed:

As shown in Figure 5, the temporary point $p_{t e m p}^{h u m a n}$ is constructed through the 3D human body feature point $p_{s h o u l d e r - r i g h t}^{h u m a n}$ and 3D human body cutting ring $L_{a r m - r i g h t}$ . The x value of $p_{t e m p}^{h u m a n}$ is the X direction value of $L_{a r m - r i g h t}$ cutting ring; and y value and z value of the y value and z value of $p_{s h o u l d e r - r i g h t}^{h u m a n}$ .

All cutting rings from $p_{s h o u l d e r - r i g h t}^{h u m a n}$ to $L_{a r m - r i g h t}$ are obtained, with the maximum point in the X direction of each cutting ring solved. The black points as shown in Figure 4 are set to $p_{o u t X - i}^{h u m a n}$ and $i = 0, 1, 3, \dots n$ .

The point with the maximum distance from $p_{t e m p}^{h u m a n}$ in $p_{o u t X - i}^{h u m a n}$ to be used as $p_{a r m R o o t - r i g h t}^{h u m a n}$ . In order to obtain the key point of the sweater at the elbow (taking the right arm as an example), We establish three directions perpendicular to each other $V_{0}$ , $V_{3}$ , $V_{2}$ , where $V_{0} = p_{w r i s t - r i g h t}^{h u m a n} - p_{e l b o w - r i g h t}^{h u m a n}$ , $V_{1} = C_{L} - p_{e l b o w - r i g h t}^{h u m a n}$ , $C_{L}$ is the center of $L_{e l b o w}$ , $V_{2} = V_{1} \times V_{0}$ ,Vector $V_{3}$ is obtained by contrarotating $V_{0}$ around $V_{2}$ with $V_{2}$ as the axis and by 900.the intersection with the right arm is obtained with a ray with a starting point of $p_{e l b o w - r i g h t}^{h u m a n}$ and with $V_{3}$ as the direction is obtained and set to $p_{e l b o w - r i g h t 0}^{h u m a n}$ . Similarly, the front and rear feature points of the sweater are obtained by the ray passing through the midpoint of $P_{e l b o w - r i g h t}^{h u m a n}$ and $p_{e l b o w - r i g h t 0}^{h u m a n}$ , and in a direction perpendicular to $V_{3}$ .

Figure 5.

Key points of the gusset.

The four key points at the elbow of the sweater sleeve can be established by the above-mentioned method. The four key points at the wrist can be established by a similar method. In addition to the key points at the shoulder, the elbow and the wrist, in order to make the sweater generated better fit the arm of the 3D human body, the human body feature cutting ring $L_{f o r e a r m}$ should be found between forearms $L_{e l b o w}$ and $L_{w a i s t}$ to obtain the key points of sweater style at this part.

In order to facilitate the construction of the bottom style, it is necessary to obtain the most convex point of the thigh, that is, the key point of the leg root. In the subsequent modeling of skirts or trousers, the key circumference of the lower body shape is obtained according to the contour line of the leg. In style construction, the knee point is generally used as the reference point to determine the height of the skirt. The key points of sweater style finally constructed are shown in Figure 6. Basic styles of sweaters, such as tops and skirts, can be constructed according to these feature points of sweater style.

Figure 6.

Key points of sweater.

Sweater surface construction

On the basis of human body feature points and key points of style obtained, the feature contour line can be obtained according to the line between the corresponding points. Since the contour line is wrapped in the human body, the two are closely linked, and there is certain ease between sweater garments and the human body, when designing sweaters, the value of the ease at each part is mainly determined according to the desired style effect. The specific styling of regular styles of sweaters, that is, tight-fitting, fitting, semi-loose-fitting and loose- loose amount, processing values in production are added to the algorithm, such as the addition of sleeve loose amount in the top and the addition of skirt loose amount in the bottom of the body.

The reasonable allocation of the ease amount can highlight the stylish effect of different styles, whereas the ease amount of measurements is not specifically evenly distributed among parts. In order to meet the requirements for styling, a sweater ease amount model will be established by the segmented distribution method in this section. Women’s regular tops at the chest, waist, and hips will be taken as an example to illustrate the algorithm. $H_{B L}$ and $G_{B L}$ represent the section chest girth of the human body and the girth of styles at relative parts. $g_{i}$ represents the numeric value of the ease amount of these styles. Since the section chest of the human body is divided into front and rear chest, according to the requirements of pattern making for these two, the section chest of the human body can be divided into four segments, with the front and rear segments used as calculation objects in ease amount modeling. The left chest line is divided into four parts, a, b, c and d, as shown in Figure 7(a). In front and rear chest parts, b and d are segments with the maximum curvature and represent the most convex part at the chest on the human body structure. Since sweaters are affected by the loop structure, in case of any change in human body stretches and bends, the dimensional stability of knitted fabric is poor. Considering human body activity and physiological comfort, the ease amount distribution here should focus on b and d to make the length of the chest $G_{B L} = H_{B L} + g_{i}$ , where $H_{B L} = (a + b + c + d) \times 2$ . The position of chest feature points is corrected by adjusting b and d, forming the algorithm flow chart for empty chest margin, as shown in Figure 8, in which $ρ = G_{B L} - (H_{B L} + g_{i})$ . $G_{B L}$ is a constant value determined by original $H_{B L}$ and $g_{i}$ . $ς$ is set to 0.001. The adjustment factor is $λ$ and is initialized to 2.

Figure 7.

Loose quantity distribution: (a) ease t distribution of the Chest, (b) ease t distribution of the waist, and (c) ease t distribution of the haunch.

Figure 8.

Flow chart of the ease algorithm.

In order to show the beauty of women’s contours, the east amount on both sides of the waist will be reduced. Meanwhile, the influence of garment deformation during motion on the human body should be considered. The ease amount at the rear waist should be greater than that at the front waist. The left waistline is divided into four parts, a, b, c and d. The waist girth reaches $G_{W L} = H_{W L} + g_{i}$ as shown in Figure 7(b) by simultaneously adjusting b and d.

The contour line of the hip is generally the position of the hem of women’s tops. Since the hem of sweaters is mostly ribbed with better lateral stretch ability and covers the human body, the hem fits the contour line of the hip. However, taking into account the changes in human body motion, there should be a greater ease amount at the most protruding part of the hip; and the ease amount of the back hip should be greater than that of the front hip. The left hip line is divided into three parts, a, b and c. The hip girth reaches $G_{H L} = H_{H L} + g_{i}$ as shown in Figure 7(c) by adjusting a and c. The key points of the style of tops are adjusted by the above-mentioned algorithm. Sweater styles of different effects can be formed by changing the ease amount expression for different girths.

The composition of the contour area of the sweater is realized by two steps. Firstly, it is necessary to use spline curve fitting to form a feature loop line according to the key points of the adjusted style. When the girth ease amount for each part is added, there is a certain increase of space between the human body and the garment, forming a feature ring of style with a specific size. Secondly, the style feature ring with spatial relation will be obtained. The key points on the adjacent two feature rings, such as the key points at the corresponding positions of the front, back, left and right, will be connected by spline curves according to the different styles, then, the area divided by the feature line will be obtained. Figure 9(a) shows the area division for tops. In order to ensure the accuracy of the style in the experiment, corresponding style feature rings are added to the top, half skirt and dress. In order to achieve the authenticity of the style of tops, in addition to the neck ring, chest ring, waist ring and hip ring, the chest styling area should be added; and the most convex forearm ring should be added to Figure 9(b) and (c) for the area division for skirts. We combine the sweater processing values in production. For example, the loose volume of sleeve body is generally more than1.5 cm of the upper limb torso circumference, and the lower half of skirt generally adopts a uniform loose volume distribution valve, increasing to 2 cm in the waist abdomen and 4 cm in the hip circumference.

Figure 9.

Style area division of the sweater: (a) area division of the overpull, (b) area division of the skirt, and (c) area division of the dress.

Sweater style modeling is performed according to human body feature structure. The model is divided by the feature loop line into several sub-areas, in which the surface in style areas is established by the construction method for the upper surface. The sweater style established on the basis of human body features is constructed with human body feature points, key points of style, feature ring and area surface as basic elements. The style of sweaters can be changed by adjusting these basic elements. Take tops and dresses as an example, by adjusting the ratio of the chest, waist, and hips of the tops, they can be changed to straight-shouldered and straight-sleeved tops. Skirts with low-cut collars can be realized by making corresponding adjustments to the collar, waist and length of dresses. Different styles that fit the human body can be developed by adjusting basic elements. Figure 10 shows the effect after adjusting sweater styles and smoothing the entire garment surface.

Figure 10.

Sweater area splicing and smooth.

Experiment

In order to apply style models to practice, the 3D style model should be flattened to realize the corresponding sweater style pattern.^28,29 Because we use the parametric equation when constructing the three-dimensional surface, we can flatten it into the corresponding two-dimensional surface according to the girth between the key points and the corresponding parametric equation. According to the parameter equation and the length of the corresponding contour line when constructing the three-dimensional surface, the two-dimensional contour of the front and back of the coat can be obtained. Figure 11 shows two 2D patterns of tops. Process realization and machine weaving are performed on these patterns. Figure 12 shows the effect of the woven sweater.

Figure 11.

Conversion of the 3D style model to the 2D pattern model: (a) 3D style model and (b) 2D pattern.

Figure 12.

Knitted sweaters designed with our method: (a) steps from the upper garment sweater design model to the knitted garment, (b) steps from the sleeveless dress design model to the knitted garment, and (c) steps from the half-skirt design model to the knitted garment.

To validate the fit of the sweater style constructed based on human body feature points, an additional set of experiments was conducted. In order to match human body dimensions, sweater production was carried out as per the three-dimensional human size parameters shown in Table 2. Table 3 presents a comparison between the dimensions of the digitally constructed sweater style models based on human body feature points and the actual dimensions of the produced sweaters. The sweaters were labeled as Sweater 1 for the round-neck top, Sweater 2 for the sleeveless dress, and Sweater 3 for the midi skirt, as shown in Table 2 below. From the data in Table 2, it can be observed that the actual woven sweaters are longer in the length dimension compared to the three-dimensional sweater models. Particularly, the sleeves of Sweater 1, the dress of Sweater 2, and the lower part of the skirt of Sweater 3 show greater length. In the width dimension, the actual sweaters are slightly narrower than the three-dimensional sweater models.

Table 2.

Three-dimensional human body size parameters (cm).

Human	Parameters	Human	Parameters
Height	162	Pelvis	77.9
Chest	83	Knee Height	44.9
Waist	64	Shoulder Width	38.4
Hip	92	Sleeve Length	56.1

Table 3.

Comparison of parameters between three sweaters and style models (cm).

Category Parts	Real Sweater 1	3D Style Model 1	Real Sweater 2	3D Style Model 2	Real Sweater 3	3D Style Model 3
Length of the garment	56.3	56	/	/	/	/
Chest width	41.5	42.2	32	32.6	/	/
Shoulder width	33.1	33.6	34	34.2	/	/
Waist width	39.2	40	25.3	26	32	33
Back collar width	16	16.5	14	14.5	/	/
Front collar depth	10	8	18.3	20	/	/
Sleeve length	58.2	57.8	/	/	/	/
Sleeve width	14.5	15	/	/	/	/
Skirt length	/	/	83.8	82	62.4	61.6
Skirt length	/	/	/	/	82.5	80.3

There is a slimming effect around the shoulders, chest, and waist. According to the industry standard FZ/T 73018-2012 for knitted woolen products, the deviations are within the permissible range of sweater board sizes. The reason behind this discrepancy is that the garment pattern is influenced by the structure of the knitted loops. Weaving along the vertical direction leads to elongation, while weaving along the horizontal direction results in contraction. Elongation has a significant impact on the shape of the pattern pieces. Therefore, in actual pattern-making, adjustments need to be made based on the sample size, particularly in the length dimension. Table 4 adds a group of real fitting experience scores, ranging from 1 to 10 points, and the evaluation criteria are 1 point extremely uncomfortable and 10 points extremely comfortable. The experimental data indicates that the wearer has a good perception of the comfort of wearing the sweater.

Table 4.

Wear test evaluation for the three sweaters.

Evaluation Items	Sweater 1	Sweater 2	Sweater 3
Comfort Level	10	10	10
Fit Accuracy	9	9	10
Size Appropriateness	9	9	10
Wearing Flexibility	10	9	9

Conclusions

In this paper, three-dimensional sweater designs have been modeled. Based on the personalized characteristics of the human body and the features of the sweater design, we have proposed a method based on three-dimensional human body feature parameters. The modeled sweater design is more suitable for the 3D human body. (1) On the established three-dimensional human body model, we used the body cutting ring algorithm to obtain human body feature points. Then, in conjunction with the structural features of the sweater design, we obtained the key points of the design, effectively improving the adaptability of the constructed design. (2) Using the chest, waist, and hips as examples, we employed a segment allocation method to further establish a looseness model for loose sweater designs. By adjusting the tightness and key points, we can change the type of sweater design to meet the need for sweater design variations. (3) Based on this, we used cubic spline curves to fit the key points and construct the style contour area. Using the surface interpolation method, we obtained a three-dimensional model of a regular sweater design. In the experimental process, we mapped the style model to a two-dimensional pattern through a mapping function and implemented knitting of the wool pattern. The realized sweater design conforms to the structure of the garment and the characteristics of the sweater, proving the feasibility of this method.

In summary, we verify and test the model construction of the classic sweater design on a single body. More body types will be studied in the future to create a sweater design database combining different sweater styles to provide technical support for personalized sweater design.

Footnotes

Data availability statement

The data used to support the findings of this study are available from the corresponding author upon request.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the financial support from National Science Foundation of China (No.62202068).

ORCID iD

Duan Li

References

Goncu-Berk

Tuna

BG.

The effect of sleeve pattern and fit on E-Textile electromyography (EMG) electrode performance in smart clothing design. Sensors 2021; 21: 5621.

Markiewicz

Witkowski

Sitnik

, et al. 3D anthropometric algorithms for the estimation of measurements required for specialized garment design. Expert Syst Appl 2017; 85: 366–385.

Peng

Jiang

Cong

Rapid parametric human modeling in 3D garment simulation. Autex Res J 2019; 19: 60–67.

Liu

, et al. A data-driven editing framework for automatic 3D garment modeling. Multimed Tools Appl 2017; 76: 12597–12626.

Wang

Zeng

, et al. Prediction of garment fit level in 3D virtual environment based on artificial neural networks. Text Res J 2021; 91: 1713–1731.

Yan

Kuzmichev

VE.

A virtual e-bespoke men’s shirt based on new body measurements and method of pattern drafting. Text Res J 2020; 90: 2223–2244.

Zhang

Wang

Yang

Design 3D garments for scanned human bodies. J Mech Sci Technol 2014; 28: 2479–2487.

3D measurement of human chest and abdomen surface based on 3D Fourier transform and time phase unwrapping. Sensors 2020; 20: 1091.

Wang

Newton

, et al. Modeling of cross-sectional shape for women’s jacket design. Sen i Gakkai 2007; 63: 87–86.

10.

Liu

Zhu

Tao

, et al. Parametric design of garment pattern based on body dimensions. Int J Ind Ergon 2019; 72: 212–221.

11.

Liu

Jiang

Dong

Online 3D model reconstruction and virtual display for garment. J Eng Fiber Fabr 2021; 16: 1–10.

12.

Wang

Cong

Dong

, et al. Digital design model for weft-knitted seamless yoga pants based on skin deformation. J Eng Fiber Fabr 2021; 16: 1–9.

13.

Sharma

Koehl

Bruniaux

, et al. Development of an intelligent data-driven system to recommend personalized fashion design solutions. Sensors 2021; 21: 4239.

14.

Liberadzki

Adamczyk

Witkowski

, et al. Structured-light-based system for shape measurement of the human body in Motion. Sensors 2018; 18: 2827.

15.

Wang

CCL

Chang

TKK

Yuen

MMF.

From laser-scanned data to feature human model: a system based on fuzzy logic concept. Comput Aided Des 2003; 35(3): 241–253.

16.

Shotton

Girshick

Fitzgibbon

, et al. Efficient human pose estimation from single depth images. IEEE Trans Pattern Anal Mach Intell 2013; 35(12): 2821–2840.

17.

Girshick

Shotton

Kohli

. Efficient Regression of General-Activity Human Poses from Depth Images. In: ICCV ’11 Proceedings of the 2011 International Conference on Computer Vision, pp.415–422.

18.

Zhong

, et al. A study on improving the calibration of body scanner built on multiple RGB-Depth cameras. Int J Clothing Sci Technol 2017; 29: 314–329.

19.

Cignoni

Callieri

Corsini

, et al. MeshLab: an Open-Source Mesh Processing Tool. DBLP, 2008. DOI: 10.2312/LocalChapterEvents/ItalChap/ItalianChapConf2008/129-136.

20.

3D point cloud registration denoising method for human motion image using deep learning algorithm. Multimed Syst 2020; 26: 75–82.

21.

Nazir

Afzal

Pagani

, et al. Contrastive Learning for 3D point clouds classification and shape completion. Sensors 2021; 21: 7392.

22.

Long

, et al. SAP-Net: A simple and robust 3D point cloud registration network based on local shape features. Sensors 2021; 21: 7177.

23.

Giachetti

Lovato

Piscitelli

, et al. Robust automatic measurement of 3D scanned models for the Human Body Fat Estimation. IEEE J Biomed Health Inform 2015; 19(2): 660–667.

24.

Phillips

D’Adamo

McAree

Maximum sum of Evidence—An evidence-based solution to object pose estimation in Point Cloud Data. Sensors 2021; 21: 6473.

25.

Guan

Reiss

Hirshberg

, et al. DRAPE: dressing any person. ACM Trans Graph 2012; 31: 1–10.

26.

B-Ru

Qian

Min

, et al. Automatic measurement of scanned human body in fixed posture. In: International Conference on Computer-Aided Industrial Design & Conceptual Design, 2010, pp.575–578.

27.

Zhong

, et al. Automatic body measurement based on slicing loops. Int J Clothing Sci Technol 2018; 30(3): 380–397.

28.

Wang

GD.

Customer participating 3D garment design for mass personalization. Text Res J 2011; 81: 187–204.

29.

Luo

Jiang

Cong

Conversion from 3D to 2D pattern algorithm for the 3D-shaped knitwear. Int J Clothing Sci Technol 2020; 33: 65–73.

3D sweater garment style generation based on 3D anthropometric characteristic parameters

Abstract

Keywords

Introduction

Related work

Obtaining 3D human body data

Obtaining 3D human body feature points

Methodology

Obtaining key points of style

Key points of the torso

Key points of limbs

Sweater surface construction

Experiment

Conclusions

Footnotes

Data availability statement

Declaration of conflicting interests

Funding

ORCID iD

References