A method and system for generating a smart pattern of a fit garment based on dynamic skin deformation

Abstract

The application discloses a kind of based on dynamic skin deformation and smart generation method and system of fit clothing pattern.The method first divides trunk surface into 16 analysis regions, extracts the first principal component by principal component analysis dimension reduction, overcome the technical prejudice that skin is regarded as homogeneous isotropic material in prior art.Then, chest lumbar difference-flattening error segmented mapping function is established to adaptively compensate the area obtained by first flattening of deformation sensitive area;Then, based on the comparative analysis of fabric load tensile test and virtual fitting pressure response, the deformation rate threshold is determined, the deformation rate-province way transfer rule is constructed, and the side seam allowance is corrected;Finally, joint-skin mechanical transmission model is established, and the cascade double closed loop system verification optimization is carried out by adopting outer ring body shape compensation and inner ring deformation structure feedback.The application solves the technical problems of static design, discrete classification and open loop system in prior art, and realizes intelligent plate making with high fit and high comfort.

Description

Technical Field

[0001] This invention belongs to the field of intelligent design and manufacturing technology of clothing, specifically relating to a method and system for intelligent generation of fit garment patterns based on dynamic skin deformation. Background Technology

[0002] With the development of information technology and intelligent manufacturing, the apparel industry is gradually transforming from the traditional large-scale standardized production model to a personalized customization production model. During this transformation, consumer demand is becoming increasingly diversified and sophisticated, with rising requirements for fit, comfort, and functionality. Apparel not only needs to meet basic aesthetic and wearing needs but also needs to consider comfort and fit under various body postures. This places higher demands on garment structural design and gradually reveals the limitations of traditional pattern-making methods that rely on experience and manual operation.

[0003] Traditional methods primarily rely on static human body dimensions, using empirical formulas to determine key structural parameters for pattern drawing and modification. While these methods are applicable to standard body types and conventional styles under static conditions, they struggle to meet the comfort needs of the human body in various postures, especially in deformable areas, often resulting in insufficient fit or restricted movement. Therefore, clothing structural design is gradually shifting from a design model centered on static dimensions to one that addresses the needs of multi-posture wearing.

[0004] Meanwhile, the rapid development of technologies such as 3D human body scanning, virtual fitting, and computer-aided design has provided technical support for the digitalization and intelligentization of garment structural design. By acquiring high-precision 3D human body data and combining it with methods such as surface modeling, surface flattening, and parametric modeling, the digital transformation from human geometric information to 2D garment patterns can be achieved, breaking through the limitations of traditional pattern-making methods in terms of efficiency and accuracy. Against this backdrop, how to effectively utilize 3D human body data to establish a suitable garment structure and achieve automatic generation and optimization of garment patterns has become an important research direction in the fields of garment engineering and digital design. While some progress has been made in acquiring human skin deformation, surface flattening, and parametric pattern making, the integrated path surrounding the multi-posture characteristics of the human body—surface flattening—parametric pattern generation still needs improvement. How to introduce the laws of human skin deformation into the surface segmentation and unfolding process, and further transform it into a parametric model that can drive the automatic generation of patterns, is a key issue in realizing the digital design of well-fitting garments.

[0005] Existing 3D parametric pattern making systems generally assume a simple linear mapping relationship between skin deformation and garment structural parameters. This reflects a deep-seated technical bias: the belief that skin deformation data and garment structural parameters (darts, pleats, ease) can be directly mapped using simple linear proportions or empirical formulas. For example, traditional methods typically assume that for every 10% increase in deformation rate, ease increases by a fixed value, or directly input the deformation rate as the ease coefficient. This bias stems from existing research treating skin as a homogeneous isotropic material, neglecting the differences in Gaussian curvature across different areas of the human body surface and the nonlinear stretching characteristics of the fabric. This results in a lack of quantitative mapping between deformation data and darts, pleats, and ease. However, this invention, through extensive experimentation, reveals that when the longitudinal deformation rate β exceeds a 60% threshold, the fabric enters the nonlinear stretching zone. In this state, increasing ease linearly leads to redundant accumulation of the lower garment pieces in static conditions; conversely, reducing darts linearly leads to tearing of the shoulder line in dynamic conditions. This paradox of 'static redundancy and dynamic tension' is precisely the nonlinear coupling problem that existing technologies cannot solve through conventional linear mapping.

[0006] While existing research has applied statistical methods such as principal component analysis to human kinematics analysis, these are all purely mathematical dimensionality reduction methods to reveal deformation patterns. They have not established a direct mapping relationship between load coefficients and garment structural parameters (dart amount, pleat amount, ease allowance configuration), nor have they formed an adaptive compensation closed loop based on the dimensionality reduction results. Those skilled in the art cannot directly obtain quantitative basis for pattern-making parameters from kinematic PCA analysis, and existing parameterization systems are unidirectional data flows, lacking an automatic optimization closed loop based on pressure feedback, and cannot resolve the technical contradiction of static fit but dynamic tightness. Summary of the Invention

[0007] The purpose of this invention is to provide a method and system for intelligently generating fit garment patterns based on dynamic skin deformation.

[0008] In a first aspect, the present invention provides a method for intelligently generating a fitted garment pattern based on dynamic skin deformation, the method comprising:

[0009] Point cloud data of the upper body of the subjects were collected in different postures, and a torso model was constructed based on the collected point cloud data; the surface of the torso model was divided into multiple analysis areas.

[0010] The longitudinal and lateral deformation rates of each analysis region are obtained separately; a deformation rate matrix is ​​constructed based on the longitudinal and lateral deformation rates of each analysis region; and deformation-sensitive and deformation-mild regions are identified based on the deformation rate matrix.

[0011] Each analysis region was flattened, and the corresponding original flattened area was obtained; the compensation coefficient of the deformation-sensitive area was obtained based on the subject's body shape and chest-waist difference;

[0012] The flattened area of ​​the deformation-sensitive region is corrected using a compensation coefficient; the flattened relative error is obtained based on the corrected flattened area; if the flattened relative error is greater than the error threshold, the compensation coefficient is adjusted, and the above process is repeated until the flattened relative error is not greater than the error threshold; the flattened area of ​​the last correction is taken as the final flattened area of ​​the deformation-sensitive region.

[0013] The amount of material saved is adjusted based on the longitudinal and lateral deformation rates of the deformation-sensitive and deformation-moderate zones.

[0014] An initial garment pattern is generated based on the flattened area and sag of all analysis regions, and a virtual fitting is performed based on the initial garment pattern to detect the pressure distribution. If the pressure distribution does not meet the iteration termination condition, the coordinates of the sag cusp are adjusted and the garment pattern is updated. The above process is repeated until the iteration termination condition is met. The last obtained garment pattern is used as the final generated fitted garment pattern.

[0015] Preferably, the method for identifying the deformation-sensitive region and the deformation-mild region is as follows: performing principal component analysis on the deformation rate matrix to extract the first principal component of the deformation rate matrix; constructing a factor loading matrix based on the first principal component; obtaining the loading coefficients of each analysis region from the factor loading matrix, and identifying the deformation-sensitive region and the deformation-mild region from the analysis region based on the loading coefficients; wherein, the loading coefficient of each analysis region is taken as the larger of the absolute values ​​of the longitudinal deformation rate and the transverse deformation rate of that region on the first principal component.

[0016] Preferably, the method for obtaining the corrected flattened area is as follows: based on the flattened area before correction, add the product of the flattened area before correction and the compensation coefficient to obtain the corrected flattened area.

[0017] Preferably, the method for adjusting the coordinates of the provincial cusp is as follows: using the pressure gradient direction as a guide, adjust the coordinates of the provincial cusp by a set step size.

[0018] Preferably, the pressure distribution includes shoulder pressure and waist pressure; the iteration termination condition is: shoulder pressure is less than the shoulder pressure threshold and waist pressure is less than the waist pressure threshold, and the pressure change rate is less than the change rate threshold for multiple consecutive iterations.

[0019] Preferably, the longitudinal deformation rate is the ratio of the difference between the dynamic length and the static length to the static length; the transverse deformation rate is the ratio of the difference between the dynamic girth and the static girth to the static girth.

[0020] As a preferred embodiment, the method for dividing the analysis area is as follows: dividing the area horizontally along the neck circumference line, chest circumference line, waist circumference line and armpit circumference line respectively, and then taking multiple points on the horizontal dividing lines for vertical division.

[0021] As a preferred approach, after adjusting the amount of savings, a robust design is implemented for the adjusted amount of savings based on the upper limit of the confidence interval.

[0022] Preferably, the compensation coefficient is adjusted using a proportional-integral control algorithm; the proportional coefficient and integral coefficient in the proportional-integral control algorithm are calibrated based on the dynamic response of fabric shear stiffness and surface flattening error.

[0023] Preferably, the flattening relative error is the ratio of the difference in area before and after flattening to the area before flattening.

[0024] Preferably, the longitudinal deformation rate and the lateral deformation rate of the analysis region are the maximum longitudinal deformation rate and the maximum lateral deformation rate in different postures of the analysis region.

[0025] Secondly, this invention provides an intelligent system for generating well-fitting garment patterns based on dynamic skin deformation, which is used to execute the aforementioned intelligent method for generating well-fitting garment patterns. This intelligent system includes a data acquisition module, a model construction module, a flattening module, a flattening area correction module, a side seam dart correction module, a pattern generation module, and a pattern optimization module. The data acquisition module is used to collect the body shape data of the subject; the model construction module is used to construct a torso model of the subject; the flattening module is used to divide the torso model into analysis areas and flatten each analysis area; the flattening area correction module is used to correct the flattening area of ​​the analysis areas; the side seam dart correction module is used to correct the side seam darts of the analysis areas; the pattern generation module is used to generate an initial garment pattern based on the adjusted flattening area and side seam darts; and the pattern optimization module is used to perform a virtual fitting based on the initial garment pattern to adjust the dart point coordinates and obtain the final well-fitting garment pattern.

[0026] The beneficial effects of this invention are:

[0027] 1. This invention uses data dimensionality reduction and pattern recognition technology to accurately locate key sensitive areas of human body dynamic deformation, achieving precision and differentiation in looseness configuration. It effectively solves the contradiction between insufficient looseness in key activity areas and excessive looseness in non-critical areas in traditional methods. On the validation set, R² increased from 0.68 to 0.89, significantly improving the fit and wearing comfort of the pattern.

[0028] 2. This invention establishes a piecewise mapping relationship between body shape parameters and surface flattening error, realizing full-range adaptive compensation from slender to full-figured body types, eliminating the compensation blind spot of traditional discrete classification methods on transitional body types, and improving the fitting accuracy for different body types.

[0029] 3. By constructing a dynamic transfer mechanism for deformation rate and dart amount, this invention achieves an intelligent balance between releasing looseness in high deformation areas and recovering excess material in low deformation areas. This effectively solves the problem of sacrificing one aspect for the other in traditional pattern making, namely static fit and dynamic restraint, dart tearing and fabric accumulation, ensuring that the garment maintains a good fit both at rest and during commuting.

[0030] 4. This invention establishes a mechanical transmission chain of joint movement-skin deformation-clothing ease, realizing active ease prediction and configuration based on movement posture, replacing the traditional passive measurement method that relies on experience and trial and error, and greatly improving the design efficiency and scientific nature of personalized pattern making.

[0031] 5. By constructing a cascaded dual closed-loop feedback control system, this invention achieves automatic error correction and optimization throughout the entire process from body feature recognition to template generation, ensuring stable convergence and output quality of the system under different input conditions, and significantly improving the first-pass yield and manufacturing reliability of the finished product. Attached Figure Description

[0032] Figure 1 This is the overall flowchart of the present invention.

[0033] Figure 2 This is a schematic diagram of the front and back marking points of the upper body of the human body in this invention; where (a) is the front and (b) is the back.

[0034] Figure 3 This is a schematic diagram of the mechanical transmission chain in this invention.

[0035] Figure 4 This is a diagram of the cascaded dual-closed-loop intelligent feedback system architecture in this invention.

[0036] Figure 5 This is a schematic diagram of the curved surface flattening and its prototype of the Y-shaped subject in the single-arm raised posture in Embodiment 1 of the present invention.

[0037] Figure 6 This is a schematic diagram of the surface mesh division and PCA load coefficient distribution of the 16 analysis regions in Embodiment 1 of the present invention.

[0038] Figure 7 This is a schematic diagram of the complete two-dimensional template generated after optimization by double closed-loop feedback in Embodiment 1 of the present invention.

[0039] Figure 8 The diagram shows a comparison of the traditional prototype and the virtual try-on effect of Example 1 for a Y-shaped body; where (a) is the virtual try-on effect of the traditional prototype and (b) is the virtual try-on effect of the prototype of Example 1.

[0040] Figure 9The diagram shows the pressure distribution before and after optimization of the B-type template; where (a) is the pressure distribution before optimization and (b) is the pressure distribution after optimization.

[0041] Figure 10 The diagram shows a comparison of virtual try-on for the traditional prototype and the prototype of Example 2 for body type B; where (a) shows the virtual try-on effect of the traditional prototype and (b) shows the virtual try-on effect of the prototype of Example 2.

[0042] Figure 11 This is a schematic diagram of the flattening of the curved surface and its prototype in the bent-arm posture of a C-type subject in Embodiment 3 of the present invention.

[0043] Figure 12 The diagram shows a comparison of the traditional prototype and the virtual try-on effect of the prototype in Example 3 for a C-body type; where (a) is the virtual try-on effect of the traditional prototype and (b) is the virtual try-on effect of the prototype in Example 3.

[0044] Figure 13 This is a schematic diagram of the curved surface flattening and its prototype of the A-type subject in the arms-raised posture in Embodiment 4 of the present invention.

[0045] Figure 14 A schematic diagram showing the prototype virtual fitting comparison of embodiment 4 for body type A.

[0046] Figure 15 This is a schematic diagram of the iterative convergence characteristics of the cascaded dual closed-loop feedback system of the present invention.

[0047] Figure 16 This is a schematic diagram of the pressure distribution verification of the virtual fitting for multiple body types and postures according to the present invention; wherein, (a) is the stress distribution of the Y body type under different commuting postures; (b) is the stress distribution of the A body type under different commuting postures; (c) is the stress distribution of the B body type under different commuting postures; and (d) is the stress distribution of the C body type under different commuting postures. Detailed Implementation

[0048] The invention will be further described below with reference to the accompanying drawings. This embodiment employs a two-stage validation strategy: 12 exploratory subjects (3 for each of the four body types Y / A / B / C, aged 22-45 years, BMI 18-26) are used for model building in S1-S5, and 30 validation subjects (S6) are used for robustness confirmation. All parameters are calibrated based on experimental data, and the cutoff points strictly follow the GB / T1335.2-2008 body type classification standard (Y type Δ≥19cm, A type 14cm~18cm, B type 9cm~13cm, C type 4cm~8cm).

[0049] A method for intelligently generating well-fitting garment patterns based on dynamic skin deformation is disclosed. The intelligent pattern generation system comprises a data acquisition module, a model building module, a flattening module, a flattening area correction module, a side seam dart correction module, a pattern generation module, and a pattern optimization module. The data acquisition module collects body shape data of the subject. The model building module constructs a torso model of the subject. The flattening module divides the torso model into analysis areas and flattens each area. The flattening area correction module corrects the flattened area of ​​the analysis areas. The side seam dart correction module corrects the side seam dart amount of the analysis areas. The pattern generation module generates an initial garment pattern based on the adjusted flattened area and side seam dart amount. The pattern optimization module performs a virtual fitting based on the initial garment pattern to adjust the dart point coordinates and obtain the final well-fitting garment pattern.

[0050] like Figure 1 As shown, the intelligent generation method for the fitted garment pattern includes the following steps:

[0051] Step 1: Multi-pose 3D point cloud data acquisition and deformation pattern dimensionality reduction

[0052] The EX Scan portable 3D scanner was used to acquire upper body point cloud data of 12 exploratory subjects in a static state and in three commuting postures (both arms outstretched at θ=90°, one hand raised at θ=180°, and both arms bent at θ=60°).

[0053] Step 1-1. Surface Mesh Construction and Deformation Rate Calculation

[0054] like Figure 2 As shown, 28 skin markers were set along the neck circumference line (N, 4 points), chest circumference line (B, 5 points), waist circumference line (W, 5 points), axillary circumference line (Y, 7 points), and the line connecting the lateral neck point N3 and the acromion point S, dividing the torso surface into 16 analysis regions: left side a1-a4 and right side b1-b4 for the front piece, and left side c1-c4 and right side d1-d4 for the rear piece. Hole repair and triangular mesh reconstruction were performed using Geomagic Studio.

[0055] For each region, the maximum longitudinal deformation rate β and the maximum lateral deformation rate α are calculated under three orientations. The largest absolute value among the three orientations is selected to cover the most unfavorable deformation state (robust design). The maximum longitudinal deformation rate β and the maximum lateral deformation rate α are expressed as:

[0056]

[0057]

[0058] in, For dynamic length; It is the static length; For dynamic dimensions; This is a static dimension.

[0059] Construct a deformation rate matrix X∈R based on the maximum longitudinal deformation rate β and the maximum transverse deformation rate α. 12×32 The deformation rate matrix consists of rows representing 12 subjects and columns representing the maximum longitudinal deformation rate β and the maximum transverse deformation rate α of 16 regions (16×2=32 dimensions).

[0060] Step 1-2. Principal Component Analysis for Dimensionality Reduction

[0061] Z-score normalization is applied to the deformation rate matrix X to eliminate dimensional differences, and it is expressed as follows:

[0062]

[0063] in, These are the elements in the standardized deformation rate matrix; represents the elements in the deformation rate matrix X; Let be the mean of the j-th column in the deformation rate matrix X; Let be the standard deviation of the j-th column in the deformation rate matrix X.

[0064] Calculate the deformation rate matrix covariance matrix Then, the characteristic equation is solved to obtain multiple eigenvalues ​​λ, which are expressed as follows:

[0065]

[0066]

[0067] The first principal component, PC1 (eigenvalue λ1=3.74, contribution rate 78.3%), is extracted according to the Kaiser criterion (λ>1) to represent shoulder joint-driven longitudinal deformation. Since the contribution rate of PC1 (78.3%) does not reach the 85% information retention threshold, a secondary principal component, PC2 (eigenvalue λ2=0.89, contribution rate 12.6%), is extracted in descending order of eigenvalue to assist in representing trunk torsional lateral deformation. The eigenvalue λ2<1 of PC2 is extracted based on the information retention threshold of a cumulative variance contribution rate ≥85%, not the Kaiser criterion. The cumulative contribution rate of both is 90.9%, meeting the information retention requirement. The load distribution of PC2 is used to assist in identifying the dominant lateral deformation region (such as the back width region c2), providing a deformation pattern verification basis for the threshold calibration of the lateral provincial transfer rules (rules C and D) in step three. Region division is performed only based on the first principal component load; the secondary principal components do not participate in the calibration of deformation-sensitive and deformation-stable regions.

[0068] A factor loading matrix was constructed based on the extracted first and second principal components. Where e1 and e2 are the eigenvectors (unit vectors) corresponding to the first and second principal components, respectively. Loading coefficients for each analysis region are obtained from the factor loading matrix. The loading coefficient for each analysis region is the larger of the absolute values ​​of the longitudinal deformation rate β and the lateral deformation rate α loaded on the first principal component. Based on the loading coefficients, K-means clustering (k=3) is used for verification, and the loading coefficients naturally cluster into three clusters: high, medium, and low, with centroids of 0.85, 0.55, and 0.05, respectively. Considering the interpretability and pattern-making accuracy requirements of garment engineering, a threshold of 0.8 between the high and medium clusters and 0.3 between the medium and low clusters are used as boundary thresholds to identify deformation-sensitive and deformation-moderate regions from the analysis regions: regions with |loading coefficient|>0.8 are defined as deformation-sensitive regions; regions with 0.3≤|loading coefficient|≤0.8 are defined as deformation-moderate regions; and regions with |loading coefficient|<0.3 are defined as deformation-stable regions. The deformation-stable region refers to a region with low PCA loading, but this does not necessarily mean small geometric flattening error. The overall Gaussian curvature of the Y-shaped body surface is large, resulting in relatively large flattening errors in all regions. However, the deformation pattern in region d1 is stable, so it is still considered a deformation-stable region. The above region division is based solely on the loading of the first principal component, as the PC1 contribution rate of 78.3% already covers most of the deformation information. If the number of principal components k>2, then the high-loading regions are identified based on the factor loading matrix of each principal component, and independently characterized according to the anatomical movement pattern, not limited to the first two eigenvalues.

[0069] Repeated measures one-way ANOVA showed statistically significant differences in the scores of the first principal component (PPC) under different commuting postures (p < 0.001, partial η² > 0.70), validating the effectiveness of the PPC in capturing posture-driven deformation. It should be noted that the above principal component extraction employed a dual-criteria hierarchical strategy: the extraction of the PPC strictly followed the Kaiser criterion (λ > 1) to ensure its clear statistical significance; when the variance contribution rate of the PPC did not reach the 85% information retention threshold, the supplementary extracted secondary principal components were not bound by the Kaiser criterion, and their eigenvalues ​​could be less than 1. The extraction was based solely on the information retention threshold of a cumulative variance contribution rate ≥ 85%. This secondary principal component was only used to assist in characterizing lateral deformation patterns and did not participate in the identification of deformation-sensitive and deformation-stable regions; the region division was still performed solely based on the PPC loading.

[0070] Step 2: Continuous mapping and adaptive compensation of chest-waist difference-flattening error

[0071] The 3D mesh is flattened based on the deformation-sensitive region and the deformation-stable region to generate a 2D template. Since the deformation-sensitive region undergoes nonlinear stretching under dynamic posture, its flattening error is significantly related to the chest-waist difference, i.e., body shape characteristics. Therefore, a continuous mapping function is established for adaptive compensation.

[0072] Step 2-1. Surface flattening and error calculation

[0073] Import the 3D human body model into 3ds Max, perform UVW topology segmentation along the lines connecting the marked points, and flatten each of the 16 regions using Peel Mode; calculate the relative flattening error of each region. It is represented as:

[0074]

[0075] in, This refers to the area obtained from the initial flattening. The area before flattening is ε; a positive value of the flattening relative error ε indicates expansion after flattening, while a negative value indicates contraction.

[0076] To maximize the inter-group differences, the sensitive region with the largest first principal component load was selected as the typical sensitive region b2 (lateral waist, first principal component load 0.87) from the sensitive region, and the sensitive region with the smallest first principal component load was selected as the typical stable region d1 (back center, PC1 load 0.12) from the stable region. The flattening relative errors of the typical sensitive region and the typical stable region were compared, and the results are shown in Table 1.

[0077] Table 1 Comparison of relative flattening errors between typical sensitive and typical stable regions.

[0078] Y 19cm~24cm +7.6% +7.5% Expansion (+) negative compensation A 14cm~18cm -1.7% +4.2% Mixed (±) Zone compensation B 9cm~13cm -10.1% -10.5% shrink(-) Positive compensation C 4cm~8cm -5.1% -2.6% slight contraction Slightly positive

[0079] Table 1 shows that the flattening error direction is significantly correlated with the chest-waist difference. The Y-shaped body (high curvature) exhibits an expansion error (+7.6%), the B / C-shaped body (near developable surface) exhibits a contraction error (-10.1% / -5.1%), and the A-shaped body shows mixed characteristics. This indicates that there is a compensation blind zone in the transition range of discrete body type classification, and it is necessary to establish a piecewise mapping relationship between the chest-waist difference and the error to achieve seamless adaptive compensation.

[0080] Step 2-2. Construct a continuous mapping function

[0081] Based on the GB / T 1335.2-2008 body type classification standard, the piecewise least squares fitting method was used to construct the chest-waist difference Δ and the relative error of flattening for different body types. The piecewise linear function (Δ) has its breakpoints consistent with national standards, and leave-one-out cross-validation is used to ensure prediction error <±5%. Breakpoints are set at Δ=9cm, Δ=14cm, and Δ=19cm, forming four linear mapping segments: C-shaped feature segment. B-type body characteristic segment A. Body type characteristic segment and Y-shaped body feature segment The transitional thoracolumbar differences (e.g., 8–9 cm, 13–14 cm, 18–19 cm) are calculated using linear interpolation of the compensation coefficients of adjacent characteristic segments. The piecewise linear function described above is expressed by the formula:

[0082]

[0083] in, The difference between the chest and waist is significant. For coefficient terms; This is a constant term.

[0084] Table 2 shows the piecewise mapping coefficients for each body shape feature segment after leave-one-out cross-validation calibration with 12 subjects.

[0085] Table 2. Segmentation mapping coefficients for each body shape feature segment

[0086] Y-shaped feature segment [19, 24] ε = 0.00312Δ - 0.0015 η = -0.00312Δ + 0.0015 3 A-type body shape characteristic segment [14, 18] ε = -1.7% η = +1.7% 3 B-type characteristic segment [9, 13] ε = -0.045Δ + 0.52 η = 0.045Δ - 0.52 3 C-shaped feature segment [4, 8] ε = -0.01Δ + 0.03 η = 0.01Δ - 0.03 3

[0087] Note: In Table 2, ε(Δ) represents the piecewise least squares fitted value based on data from 12 subjects, while Table 1 shows the measured mean of the corresponding typical sensitive area (b2) for body type. There are fitting residuals between the two (e.g., when Δ=22 cm for Y-type, the fitted value +6.71% differs from the measured mean +7.6% by approximately 0.9 percentage points). After leave-one-out cross-validation, all residuals were controlled within ±5%, meeting the prediction accuracy requirements. Further linear regression significance testing was performed on the samples within this segment. The regression slope of the chest-waist difference Δ and the relative error of flattening ε was not significant by the t-test (p>0.05), indicating that the flattening error and chest-waist difference within the A-type characteristic segment have no linear trend, and therefore are simplified to a constant term.

[0088] In the Y-shaped feature segment, that is, when Since the flattening relative error is positive (expansion error), the compensation coefficient η(Δ) is negative, and negative compensation is performed to shrink the area.

[0089] In the A-type feature segment, that is, when Since the relative error of flattening is negative (shrinkage error), the compensation coefficient η(Δ) is positive, and positive compensation is performed to expand the area.

[0090] In the B-type feature segment, that is, when Since the relative error of flattening is negative (shrinkage error), the compensation coefficient η(Δ) is positive, and positive compensation is performed to expand the area.

[0091] In the C-type feature segment, that is, when The overall flattening relative error shows a negative trend or enters the shrinkage range (shrinkage error) after passing zero. Therefore, the compensation coefficient η(Δ) is positive, and positive compensation is performed to expand the area.

[0092] The above piecewise mapping function is used to predict the flattening error trend of different body shapes and provide the initial compensation value for the outer loop PI control algorithm. In practical applications, after the initial compensation is performed on the sensitive area with this initial value, the real-time flattening error calculation in step 2-1 is used for verification and fine-tuning until the convergence criterion is met.

[0093] Steps 2-3. Adaptive Area Correction

[0094] Because the deformation-sensitive areas (b2, b3, c2) have large flattening errors under dynamic postures, and directly affect the fit between the side seam and the armhole, the deformation-stable areas have smaller errors and can use standard flattening parameters. Adaptive compensation is only established for the deformation-sensitive areas (b2, b3, c2).

[0095]

[0096] Where S is the original flattened area; This is the corrected flattened area.

[0097] After leave-one-out cross-validation on 12 subjects, the mean area error after compensation was within ±4.2%.

[0098] In some embodiments, for a chest Gaussian curvature G > 0.15 / cm -2 For areas of severe bulge, an additional -2% curvature expansion compensation is applied.

[0099] Step 3: Constructing the deformation rate-provincial highway transfer

[0100] Step 3-1. Obtain the transfer threshold

[0101] Three fabrics—cotton, polyester-cotton blend, and elastic cotton—were subjected to a constant-load tensile test (load 20 N / cm). Preliminary experiments determined the fabric's elastic recovery rate comfort threshold to be 90%. Longitudinal deformation direction: Under fitted conditions, the skin's longitudinal deformation rate β is approximately equal to the fabric's longitudinal strain (calibrated using CLO 3D virtual fitting, linear correlation coefficient R² > 0.90), therefore β is used as the equivalent index of the fabric's longitudinal strain. Combining the skin deformation data collected in step one with CLO 3D virtual try-on, it was found that when the longitudinal skin deformation rate β>60%, the fabric stretching enters the non-linear region, which is manifested as: the fabric elastic recovery rate drops to below 85% (below the comfort threshold of 90%, as determined by the stretching test), and the slope of the side waist pressure P increases from 0.32kPa / % to 0.58kPa / % (as determined by the virtual try-on), which is identified as the critical region of high longitudinal stretching; when the longitudinal skin deformation rate β<-30%, the longitudinal contraction of the skin under dynamic posture leads to longitudinal redundancy of the garment piece, with obvious wrinkles and accumulation in the front waist and side waist areas (visualized by the virtual try-on), and the longitudinal relaxation recovery rate of the fabric changes abruptly, which is identified as the critical region of longitudinal contraction.

[0102] Lateral deformation direction: When the lateral deformation rate α < -30% (mainly occurring in the back width area with arms outstretched), obvious wrinkles and accumulation appear in the fabric at the shoulder line (visualized judgment in virtual fitting), and the lateral shear stiffness of the fabric changes abruptly, which is determined to be the critical zone of lateral high shrinkage; when the lateral deformation rate α > 60% (mainly occurring in the back width area), the lateral stretching of the fabric enters the nonlinear region, manifested as a sudden increase in pressure in the back width area and an elastic recovery rate dropping below 85% (determined by tensile testing), and insufficient ease at the back sleeve cap, which is determined to be the critical zone of lateral high expansion. In summary, the dart transfer thresholds are determined as follows: longitudinal high stretching threshold β0 = 60%, longitudinal shrinkage threshold β1 = -30%; lateral high shrinkage threshold α0 = -30%, lateral high expansion threshold α1 = 60%.

[0103] Step 3-2. Transfer rule operation logic

[0104] For deformation-sensitive areas, the following priority rules are applied sequentially (once a rule is met, the process is executed and exits, and subsequent rules are not executed), constructing a deformation rate-province transfer rule: The deformation-sensitive areas are then modified based on vertical and horizontal transfer thresholds.

[0105] Rule A (High longitudinal stretch priority): If the longitudinal deformation rate β of the skin is greater than 60% and the region ∈ {b2, b3, c2}, the side seam allowance D' = D × (1 - 0.3 × β) is transferred to the adjacent low deformation region (d region) pleat.

[0106] Rule B (Longitudinal Contraction): If the longitudinal deformation rate β of the skin is less than -30%, and the region ∈ {b2, b3}, D' = D × (1 + 0.1 × |β|). Longitudinal contraction leads to skin laxity. Appropriately increase the dart allowance to prevent redundant accumulation of the dynamic lower garment piece. The dart increase should not exceed 10% of the original dart allowance to avoid the static lower garment piece being too tight.

[0107] Rule C (lateral contraction): If the skin lateral deformation rate α is less than -30% and the region ∈ {c2} (back width), the shoulder dart is converted into shoulder line contraction, and the contraction is = |α|×0.02×L; where L is the back length.

[0108] Rule D (Horizontal High Expansion): If the skin's horizontal deformation rate α is greater than 60%, and the area ∈ {c2} (back width), the back shoulder dart is reduced according to D'=D×(1-0.25×α), and the reduction is converted into the back width horizontal ease |α|×0.015×back width; at the same time, the bottom point of the back armhole is moved outward by 0.3cm~0.5cm towards the side seam (increasing linearly according to α=60%-80%), and the back sleeve cap ease is reduced by 0.3cm~0.6cm.

[0109] Rule E (Default to Maintain): If none of the above conditions are met, maintain the standard provincial highway configuration.

[0110] For the stable deformation region, maintain the standard provincial highway configuration; for the mild deformation region: trigger the above rules only when β>60% (triggering rule A) or β<−30% (triggering rule B) or α>60% (triggering rule D) or α<−30% (triggering rule C), otherwise maintain the original configuration.

[0111] When both longitudinal and lateral triggering conditions are met in the same area, the longitudinal rule is executed first, and the lateral deformation is fine-tuned through subsequent pressure feedback iterations.

[0112] Step 4: Joint-Skin-Template Mechanism Transfer Model and Robust Design

[0113] Step 4-1. Construct the mechanical transmission chain

[0114] like Figure 3 As shown, a quantitative transfer model from joint movement to clothing ease was established. Based on the literature on sports anatomy (Gu Deming et al., 2013) and pre-experimental data, the calibration coefficients were determined as follows: shoulder flexion angle θ (180° with one arm raised / 90° with both arms extended / 60° with both arms bent) → scapular upward rotation angle φ = 0.6θ (coupling coefficient 0.6, scapular upward rotation accounts for 60% during shoulder flexion) → vertical displacement Δy = 0.15φ at the axillary base Y4 (proportional coefficient 0.15, unit cm / °, R² = 0.89, p < 0.001 obtained by regression analysis of 3D scan data from 12 subjects). The lateral waist reference length Ls (the longitudinal length of the skin from the axillary base to the lumbar segment under static conditions, unit cm) was taken, and the theoretical longitudinal deformation rate β was calculated. 理论 =Δy / Ls×100%. If β 理论 If the stretching rate is >60%, the skin stretching in this posture is determined to have entered the nonlinear region, thus triggering the sag transfer in step three; this mechanical transmission chain simultaneously outputs the theoretical sag requirement value S based on rigid body geometric displacement. 理论 =k×Δy (k is the fabric elasticity correction coefficient, cotton k=1.2, polyester-cotton k=1.1, elastic cotton k=0.8).

[0115] Theoretical requirement value S for side seam saving 理论 The theoretical upper limit value is not considered for the elastic recovery of the fabric and the cushioning of the human soft tissue. It is only used to estimate the structural parameter requirements and is not directly input as the final saving amount. The final saving amount is determined by the deformation rate-dart transfer rule in step three combined with the robustness confidence upper limit.

[0116] Step 4-2. Robust Design

[0117] Using the Bootstrap resampling method (n=1000), based on the deformation rate data of 12 subjects in step one, the 95% confidence intervals for longitudinal deformation rate β and lateral deformation rate α were calculated. The upper confidence limit was taken as the slack design benchmark (covering the most unfavorable deformation of 95% of the population), and the lower confidence limit was taken as the fit validation benchmark (ensuring static fit), thus achieving design robustness at the 95% confidence level.

[0118] Step 5: Generation of MATLAB Cascaded Double Closed-Loop Intelligent Feedback

[0119] like Figure 4 As shown, a MATLAB GUI parametric plate-making system is constructed, which has a built-in cascaded dual closed-loop feedback mechanism. The outer loop is a body shape-compensation closed loop, which is used to eliminate flattening errors caused by body shape differences. The inner loop is a deformation-structure closed loop, which is used to eliminate posture-specific deformation residual errors based on the net body model corrected by the outer loop.

[0120] The outer loop uses a proportional-integral (PI) control algorithm: η(t) = η(t-1) + K p ·e(t)+Kᵢ·Σe(t), where K p In this embodiment, Kᵢ is calibrated based on the dynamic response of fabric shear stiffness and surface flattening error. p =0.5, Kᵢ=0.1, using the chest-lumbar difference Δ as the input parameter, the compensation coefficient η is calculated based on the continuous mapping function constructed in step S2. Δ Based on this, adaptive correction is performed on the surface flattening area of ​​the deformation-sensitive region; the convergence criterion of this closed loop is set as the absolute value of the current flattening error e(t) satisfies |e(t)|≤5%.

[0121] The inner ring employs a heuristic iterative algorithm, guided by the pressure gradient direction (∂P / ∂x), adjusting the cusp coordinates (P) by a step size Δs (Δs≤0.5cm). n =P n-1 ±Δs·sign(ΔP)). Where P n The coordinates of the sag apex after the nth iteration are given; ΔP is the pressure change measured by CLO 3D virtual fitting between the current iteration step and the previous step. The maximum number of iterations is 10. Taking the posture type as input, the mechanical transfer model in step four is called sequentially to calculate the initial looseness, the PCA deformation mode in step one is used to locate the adjustment area, a pre-looseness configuration including sag and pleat amounts is generated, and the DXF is exported to CLO 3D virtual fitting to detect the pressure distribution. The convergence criterion is that the shoulder pressure is ≤25kPa and the waist pressure is ≤20kPa, and the pressure change rate between two consecutive iterations is <2%.

[0122] When the inner loop fails to converge after 10 iterations, the system outputs a pressure distribution report, suggesting the replacement of the fabric with a high-elasticity fabric (current fabric elastic modulus > 0.9 N / cm recommended value), and saves the current optimal parameters to ensure system robustness.

[0123] Example 1

[0124] A method for intelligently generating fitted clothing patterns based on dynamic skin deformation, used to generate patterns for a Y-shaped (chest-waist difference 22cm) commuter shirt in a single-arm raised posture, includes the following steps:

[0125] Step S1: Input Y-shaped body data. Using PCA (Principal Component Analysis), the contribution rate of the first principal component is 78.3%. The loadings in region b2 are 0.87 and c2 are 0.81, indicating a deformation-sensitive region. The loading in region d1 is 0.12, indicating a deformation-stable region. The KMO test value is 0.71, suitable for factor analysis.

[0126] Step S2: Substitute the chest-waist difference Δ of the Y-shaped body into the continuous mapping function to obtain the compensation coefficient η(22) = -0.00312 × 22 + 0.0015 = -0.0671 = -6.71%. From Table 1, we know that the relative error of the first flattening of the sensitive area of ​​the Y-shaped body is +7.6%, and the absolute value is greater than 5%, triggering the outer loop proportional-integral control algorithm. Using η(22) = -6.71% obtained from the piecewise mapping function as the initial value of the PI algorithm, after iterative verification, this initial value has reduced the error after compensation to +0.4%, satisfying the convergence criterion of |e(t)| ≤ 5%, and the outer loop closed loop converges once. Therefore, the compensation coefficient is directly used to perform adaptive compensation on the deformation sensitive area: the two-dimensional area obtained by the first flattening of area b2 (side waist) is 165.23 cm², and after compensation, S′ = 165.23 × (1 - 0.0671). =154.15cm²; the initial flattened two-dimensional area of ​​c2 zone (back width) is 148.35cm², after compensation S′=148.35×(1-0.0671)=138.40cm²; the initial flattened two-dimensional area of ​​b3 zone (front waist side) is 158.47cm², after compensation S′=158.47×(1-0.0671)=147.85cm²; the deformation stability zones d1 and d4 maintain standard flattening parameters. The relative flattening error decreased from +7.6% in the initial flattening to +0.4%.

[0127] Step S3: The longitudinal deformation rate β in zone b2 is 65.11% > 60%. The original side seam dart of 4.32cm is adjusted according to rule A: D′ = 4.32 × (1 - 0.3 × 0.6511) ≈ 3.48cm, and the reduction ΔD = 0.84cm is transferred to the pleat in zone d. The transverse deformation rate α in zone b2 is -0.12 (|α| < 0.30), which does not trigger the transverse rule; β in zone c2 is 0.58 (< 0.60), which also does not trigger the dart transfer.

[0128] Step S4: Raise one hand high θ=180°→φ=108°→Δy=16.2cm→Theoretical side seam dart requirement S 理论 =19.44cm (upper limit of rigid body geometry), in this example, the reference length of the side waist is taken as Ls=25 cm, β 理论 =16.2 / 25×100%=64.8%>60%, which corroborates the measured β=65.11% in region b2 in step S3. From the perspective of mechanical transmission chain, this posture must have entered the nonlinear region. Taking the upper limit of 95% CI (0.7733) as the robust design benchmark, the individual saving of 3.48 cm obtained in step S3 is enlarged to the final side seam saving D. final =3.48×(0.7733 / 0.6511)≈4.13cm, to cover extreme deformation in 95% of the population.

[0129] Step S5: Outer ring convergence (current flattening error e(t) = +0.4%, satisfying the convergence criterion |e(t)| ≤ 5%). Inner ring initial template CLO 3D fitting test: shoulder pressure 28kPa (>25kPa threshold), waist pressure 18.5kPa (<20kPa threshold, qualified) → 1st iteration: the tip of the back shoulder dart moves 0.5cm towards the neck, shoulder pressure drops to 24.5kPa, waist pressure 18.2kPa → 2nd iteration: moves 0.3cm, shoulder pressure drops to 22.1kPa, waist pressure 18.0kPa → 3rd iteration: moves 0.2cm, shoulder pressure drops to 21.8kPa, waist pressure 17.8kPa, change rate 1.4% → 4th iteration: moves 0.1cm, shoulder pressure drops to 21.6kPa, waist pressure 17.6kPa, change rate 0.9% < 2%. The system converged after two consecutive iterations (the 3rd and 4th iterations) with pressure change rates of <2%, and shoulder pressure ≤25 kPa (21.6 kPa) and waist pressure ≤20 kPa (17.6 kPa). The final DXF template was generated.

[0130] Figure 5 The flattened surface and prototype of the single-arm raised posture of a Y-shaped subject (chest 91 cm, waist 72 cm, chest-waist difference Δ=22 cm) were obtained. Figure 5 It can be seen that the prototype obtained by using the dynamic posture surface flattening technique is quite different from the prototype obtained by the traditional pattern making method. The main difference is that the front and back garment lengths are much greater than those obtained by the traditional pattern making method. This is because when one hand is raised high, the human skin stretches longitudinally and increases. The traditional pattern making method is based on static pattern making, which will cause insufficient length when the human body is in motion.

[0131] Figure 6 This diagram illustrates the surface mesh generation and PCA load factor distribution for 16 analysis regions. Figure 6It can be seen that the high values ​​(load coefficient > 0.8) in regions b2, b3, and c2 represent shoulder joint driven deformation; while the low values ​​(load coefficient < 0.3) in region d represent the deformation stability region. Figure 7 The complete two-dimensional pattern, including the front, back, sleeves, and collar, was generated after optimization through a double closed-loop feedback system. The tip of the back shoulder dart was moved towards the neck side through inner loop iteration optimization. The final virtual fitting verification showed a shoulder pressure of 21.6 kPa (<25 kPa threshold) and a waist pressure of 18.5 kPa (<20 kPa threshold), with a first-time fit rate of 95%.

[0132] Figure 8 A comparison of virtual try-on of the traditional prototype and the adjusted unfolded prototype, by Figure 8 It can be seen that the traditional prototype has obvious wrinkles on both the front and back pieces, while the prototype obtained using the present invention is significantly smoother.

[0133] Example 2

[0134] A method for intelligently generating fitted clothing patterns based on dynamic skin deformation, used to generate patterns for commuter shirts for body type B (chest-waist difference 13cm) with arms outstretched in a horizontal position, includes the following steps:

[0135] Step S1 (Data Dimensionality Reduction): Input the deformation data of body type B subjects (chest circumference 88cm, waist circumference 75cm, chest-waist difference 13cm) in the arms-out position. PCA analysis showed that PC1 contributed 76.8%, region b2 had a loading of 0.84 (deformation-sensitive region), region c2 had a loading of 0.79 (deformation-moderate region), and region d1 had a loading of 0.18 (deformation-stable region).

[0136] Step S2 (Transition Zone Compensation): The chest-lumbar difference Δ = 13cm, which is at the upper limit of the B-type segment and adjacent to the A / B transition boundary. Substituting this into η(13) = 0.045×13-0.52 = 0.065 (+6.5%), since the flattening error of the B-type body is shrinkage (-10.1%, see Table 1), according to the principle that "the compensation coefficient has the opposite sign to the flattening error", η(13) = +6.5% is consistent with the theoretical expectation. Therefore, the shrinkage area after flattening is expanded to compensate.

[0137] Step S3 (Special Transfer): For body type B, the traditional rear waist dart is set at 3.25cm. Based on the deformation stability zone (PC1 load in zone d is 0.15 < 0.3), the standard rear waist dart configuration (3.25 cm) is maintained according to the standard dart configuration to avoid excessive tightening of the rear waist. Zone c2 is in a mild deformation zone and |β| < 0.60, |α| < 0.30, so no dart transfer is triggered, and the standard configuration is maintained.

[0138] Step S4 (Mechanical Transfer): Arms raised horizontally θ=90° → φ=54° → Δy=8.1cm → S理论 =9.72cm. The upper limit of 95% CI (59.9%) is taken as the design benchmark.

[0139] Step S5 (Closed-loop verification): Initially, the shoulder pressure is 31 kPa (>25 kPa threshold) and the back waist pressure is 18 kPa (qualified), but fabric wrinkles accumulate on the side waist (excessive looseness) → Fine-tune the inner ring by lowering the side seam dart point by 0.3cm → The shoulder pressure drops to 22 kPa and the side waist wrinkles are eliminated → Verification passed (shoulder 22 kPa≤25 kPa and waist 19 kPa≤20 kPa).

[0140] Figure 9 The pressure distribution before and after template optimization is shown in the left figure, where the pressure is before optimization (without flattening error compensation or dynamic deformation considerations), the shoulder pressure reaches 31 kPa, exceeding the comfort threshold by 24%, and there are obvious wrinkles on the sides of the waist, indicating insufficient ease. The right figure shows the pressure distribution after optimization: after continuous mapping compensation of the outer ring chest-waist difference (η=+6.5%) and fine adjustment of the inner ring dart tip (side seam dart tip downwards by 0.3 cm), the shoulder pressure is reduced to 22 kPa and the waist pressure is reduced to 19 kPa, with the overall pressure in the acceptable range, verifying the adaptability of this invention to transitional body types (chest-waist difference 12-15 cm).

[0141] Figure 10 A virtual fitting comparison was conducted between the traditional prototype of body type B and the prototype of this application, by... Figure 10 It is known that when the traditional prototype is held out with arms outstretched, the skin on the back contracts both horizontally and vertically, resulting in fabric accumulation and very obvious wrinkles on the back. However, the back of the prototype developed using this application is relatively flat, and the wrinkle accumulation phenomenon is greatly improved.

[0142] Example 3

[0143] A method for intelligently generating fitted garment patterns based on dynamic skin deformation is proposed for generating patterns for commuter shirts with a C-shaped body (6cm chest-to-waist difference) and bent arms.

[0144] Step S1 (Data Dimensionality Reduction): Input deformation data of C-type subjects (chest circumference 82cm, waist circumference 76cm, chest-waist difference 6cm) in a bent-arm posture. PCA extracted PC1 contribution rate of 77.2%. The PC1 loading in region b2 was 0.83, which was identified as a deformation-sensitive region; the loading in region c2 was 0.79, which was identified as a deformation-moderate region; and the loading in region d1 was 0.16, which was identified as a deformation-stable region. The KMO test value was 0.73, and appropriate factor analysis was performed.

[0145] Step S2 (C-body type compensation): Chest-waist difference Δ=6cm, substituting into η(6)= 0.01×6-0.03=0.03(+3%). The contracted area after flattening (average error -3.7%) is expanded and compensated, and the error is reduced to -0.7% after correction.

[0146] Step S3 (Stable Zone Configuration): For C-type body type with a flat chest (side seam dart only 1.18cm), the deformation stability zone maintains the standard lumbar dart of 2.5cm. Zone C2 is a mild deformation zone and has not reached the transfer threshold, so the standard configuration is maintained.

[0147] Step S4 (Mechanical Transfer): Both arms bend θ=60° → φ=36° → Δy=5.4cm → S 理论 =6.48cm (k=1.2 for cotton fabric). Take the upper limit of 95% CI (e.g., 42.5%) as the design benchmark.

[0148] Step S5 (Closed-loop verification): With arms bent, shoulder pressure is 19 kPa (qualified), waist pressure is 15 kPa (qualified), there is no high-pressure area, verification passed.

[0149] Figure 11 The flattening of the curved surface and its prototype in the bent-arm posture of a C-type subject. Figure 12 A virtual fitting comparison was conducted between the traditional C-body prototype and the prototype of this application, by... Figure 12 It is known that when the arms are bent, the front piece and the sides of the traditional prototype have many wrinkles, while the front piece and the sides of the prototype developed using this application are relatively flat, and the wrinkling phenomenon is greatly improved.

[0150] Example 4

[0151] A method for intelligently generating fitted clothing patterns based on dynamic skin deformation is proposed for generating patterns for A-line body types (15cm chest-to-waist difference) commuter shirts with arms outstretched in a horizontal position.

[0152] Step S1 (Data Dimensionality Reduction): Input the deformation data of body type A subjects (chest circumference 91cm, waist circumference 76cm, chest-waist difference 15cm) in the arms-out posture. The PCA load distribution is consistent with body types Y and B. The PC1 load in area b2 is 0.85, and the load in area c2 is 0.81. b2 and c2 are determined to be deformation-sensitive areas.

[0153] Step S2 (Transition Zone Compensation): Within the A-type feature segment [14,18], the compensation coefficient η(15) = +1.7% is directly applied using the A-type feature segment calibration. Compensation is applied to the two-dimensional area of ​​region b2 obtained from the initial flattening, 177.18 cm² (this value corresponds to the three-dimensional area Sb = 177.18 / (1 - 0.017) = 180.24 cm²).

[0154] S′=177.18×(1+0.017)=180.20cm².

[0155] The relative error after compensation and flattening is (180.20-180.24) / 180.24=-0.02%, and the error after compensation is reduced from -1.7% to -0.02%.

[0156] Step S3 (Gartery Transfer): Body type A Y4B3, that is, the longitudinal deformation rate β of zone b2 (side waist) is 54.2% (between body type C and Y body type), which does not reach the transfer threshold of β>60%, so the original dart configuration is maintained (side seam dart 4.32cm, back waist dart 4.13cm).

[0157] Step S4 (Mechanical Transfer): Arms raised horizontally θ=90° → φ=54° → Δy=8.1cm → S 理论 =9.72cm. The upper limit of 95% CI (56.8%) is taken as the design benchmark.

[0158] Step S5 (Closed-loop verification): Virtual fitting test with arms outstretched: shoulder pressure 22kPa (pass), waist pressure 18kPa (pass), no wrinkles on the sides, verification passed.

[0159] Figure 13 The flattening of the curved surface and its prototype in the arms-out posture of a body type A subject. Figure 14 This application conducts a prototype virtual fitting comparison for body type A, by Figure 14 It can be seen that the front, side, and rear parts of the prototype developed using this application are all relatively flat.

[0160] This embodiment demonstrates that the segmented mapping function disclosed in this invention is also applicable in the transition region (Δ=15cm), and that the provincial transfer threshold (β>60%) has universality across body types.

[0161] Example 5: Handling Convergence Failure in Feedback Optimization (Robustness Verification)

[0162] Figure 15 shows the iterative convergence characteristics of the cascaded dual closed-loop feedback system of the present invention. Taking a certain body type initial sample (shoulder pressure 28 kPa, exceeding the standard) as an example, the horizontal axis represents the number of inner loop iterations (1-10 times), and the vertical axis represents the pressure value (kPa).

[0163] Figure 15The display shows that after the first iteration, the shoulder dart tip moved 0.5 cm, and the pressure dropped to 24.5 kPa; after the second iteration, it moved 0.3 cm, and the pressure dropped to 22.1 kPa; after the third iteration, it moved 0.2 cm, and the pressure dropped to 21.8 kPa. The pressure change rate for two consecutive iterations was 1.4% < 2%, indicating system convergence. If the pressure is still higher than the threshold after 10 iterations (as shown by the dotted line in the figure, the pressure in the 10th iteration is 27 kPa), the system determines that the fabric elasticity is insufficient, outputs a pressure distribution report, and prompts to replace it with a high-elasticity fabric. This ensures that the system will not loop indefinitely or crash under any input, achieving robustness assurance.

[0164] To further verify the robustness of the cascaded dual-loop feedback system under different body types and multiple postures, four body type templates (Y / A / B / C) generated based on the method of this invention were imported into the CLO 3D virtual environment. Stress distribution tests were conducted under three commuting postures: arms outstretched at θ=90°, one arm raised at θ=180°, and arms bent at θ=60°. The results show that after system optimization, except for a few cases where insufficient fabric elasticity required replacement, all body types met the requirements (shoulder pressure ≤25 kPa and waist pressure ≤20 kPa) under commuting postures. This demonstrates that the synergistic effect of the outer ring body type compensation closed loop and the inner ring variable structure closed loop has cross-body type stability. Figure 16 As shown.

[0165] Experiment 1: Single Variable Decomposition Experiment

[0166] To verify the synergistic effect of the deformation rate-provincial highway transfer rule (step three) and piecewise mapping compensation (step two), three control experiments were set up:

[0167] Control group A: Only step two (area compensation) was performed, without step three (dart transfer). Result: The Y-shaped subject had a satisfactory shoulder pressure during static fitting (22 kPa), but when one arm was raised high, the side seam dart was not released, and the shoulder pressure suddenly increased to 38 kPa, resulting in dart tear.

[0168] Control Group B: Only step three (transfer to provincial highway) was performed, without step two (area compensation). Result: Dynamic pressure was satisfactory, but in static conditions, wrinkles and buildup appeared on the lateral waist due to lack of area compensation, resulting in a 35% decrease in subjective fit score.

[0169] The invention group simultaneously performed steps two and three. Results: Both static and dynamic pressure tests were satisfactory, with a 95% first-time fitting rate.

[0170] The above experiments demonstrate that steps two and three are not simply superimposed, but rather exhibit nonlinear coupling—area compensation alters the baseline length of the provincial highway, thus affecting the absolute amount of highway transfer; the highway transfer, in turn, corrects the local stress concentration after area compensation. Both are indispensable, and the synergistic effect cannot be predicted by linear extrapolation using a single method. This unexpected technical effect indicates that those skilled in the art cannot directly derive technical inspiration from existing technologies for combining PCA dimensionality reduction, piecewise mapping compensation, and deformation rate-province transfer rules, nor can they expect through conventional experimental methods that this combined scheme can simultaneously eliminate the paradox of sacrificing one for the other in terms of static redundancy and dynamic tension.

[0171] Experiment 2: Parameter Sensitivity Experiment

[0172] To determine the necessity of coefficient 0.3 in rule A, a grid search was performed on 12 subjects with coefficients k∈{0.1, 0.2, 0.3, 0.4, 0.5}. The results showed that: when k=0.1, the dynamic tension rate was 60%; when k=0.5, the static wrinkle rate was 55%; only when k=0.3 did both the static and dynamic pass rates reach their optimal levels (static wrinkle rate <5%, dynamic tension rate <3%). This coefficient is physically coupled with the fabric elastic recovery rate (85% threshold) and the nonlinear stretching inflection point (60% deformation rate), and is not an arbitrary fitted value.

[0173] Experiment 3: Extrapolation Experiment Across Body Sizes

[0174] Applying the compensation coefficient for body type B (Δ=13 cm) directly to body type A (Δ=15 cm), and the compensation coefficient for body type A to body type B, if discrete classification is used (A / B boundary 14 cm), the error for body type A is -0.015, and the error for body type B is +0.085. However, using the piecewise mapping function of this invention, the error for body type A is reduced to -0.002, and the error for body type B is reduced to +0.003. This proves that piecewise mapping eliminates the compensation blind zone in the transition interval, an effect that cannot be achieved through simple four-class discrete compensation.

[0175] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for intelligently generating fitted garment patterns based on dynamic skin deformation, characterized in that: The method includes: Point cloud data of the upper body of the subjects were collected in different postures, and a torso model was constructed based on the collected point cloud data; the surface of the torso model was divided into multiple analysis areas. The longitudinal and lateral deformation rates of each analysis region are obtained separately; a deformation rate matrix is ​​constructed based on the longitudinal and lateral deformation rates of each analysis region; and deformation-sensitive and deformation-mild regions are identified based on the deformation rate matrix. Each analysis region was flattened, and the corresponding original flattened area was obtained; the compensation coefficient of the deformation-sensitive area was obtained based on the subject's body shape and chest-waist difference; The flattened area of ​​the deformation-sensitive region is corrected using a compensation coefficient; the flattened relative error is obtained based on the corrected flattened area; if the flattened relative error is greater than the error threshold, the compensation coefficient is adjusted, and the above process is repeated until the flattened relative error is not greater than the error threshold; the flattened area of ​​the last correction is taken as the final flattened area of ​​the deformation-sensitive region. The amount of material saved is adjusted based on the longitudinal and lateral deformation rates of the deformation-sensitive and deformation-moderate zones. An initial garment pattern is generated based on the flattened area and sag of all analysis regions, and a virtual fitting is performed based on the initial garment pattern to detect the pressure distribution. If the pressure distribution does not meet the iteration termination condition, the coordinates of the sag cusp are adjusted and the garment pattern is updated. The above process is repeated until the iteration termination condition is met. The last obtained garment pattern is used as the final generated fitted garment pattern.

2. The intelligent generation method for fit-fitting garment patterns based on dynamic skin deformation according to claim 1, characterized in that: The method for identifying the deformation-sensitive region and the deformation-mild region is as follows: perform principal component analysis on the deformation rate matrix and extract the first principal component of the deformation rate matrix; Construct the factor loading matrix based on the first principal component; The loading coefficients of each analysis region are obtained from the factor loading matrix, and deformation-sensitive regions and deformation-mild regions are identified from the analysis regions based on the loading coefficients. The loading coefficient of each analysis region is the larger of the absolute values ​​of the longitudinal deformation rate and the transverse deformation rate of that region on the first principal component.

3. The intelligent generation method for fit-fitting garment patterns based on dynamic skin deformation according to claim 1, characterized in that: The method for obtaining the corrected flattened area is as follows: based on the flattened area before correction, add the product of the flattened area before correction and the compensation coefficient to obtain the corrected flattened area.

4. The intelligent generation method for fit-fitting garment patterns based on dynamic skin deformation according to claim 1, characterized in that: The method for adjusting the coordinates of the provincial cusp is as follows: using the pressure gradient direction as a guide, adjust the coordinates of the provincial cusp according to a set step size.

5. The intelligent generation method for fit-fitting garment patterns based on dynamic skin deformation according to claim 1, characterized in that: The pressure distribution includes shoulder pressure and waist pressure; the iteration termination condition is: shoulder pressure is less than the shoulder pressure threshold and waist pressure is less than the waist pressure threshold, and the pressure change rate is less than the change rate threshold for multiple consecutive iterations.

6. The intelligent generation method for fit-fitting garment patterns based on dynamic skin deformation according to claim 1, characterized in that: The longitudinal deformation rate is the ratio of the difference between the dynamic length and the static length to the static length; the transverse deformation rate is the ratio of the difference between the dynamic girth and the static girth to the static girth.

7. The intelligent generation method for fit-fitting garment patterns based on dynamic skin deformation according to claim 1, characterized in that: The method for dividing the analysis area is as follows: divide the area horizontally along the neck circumference line, chest circumference line, waist circumference line and armpit circumference line respectively, and then take multiple points on the horizontal dividing lines for vertical division.

8. The intelligent generation method for fit-fitting garment patterns based on dynamic skin deformation according to claim 1, characterized in that: After adjusting the amount of savings, a robust design is implemented for the adjusted amount of savings based on the upper limit of the confidence interval.

9. The intelligent generation method for fit-fitting garment patterns based on dynamic skin deformation according to claim 1, characterized in that: The compensation coefficient is adjusted using a proportional-integral control algorithm; the proportional coefficient and integral coefficient in the proportional-integral control algorithm are calibrated based on the dynamic response of fabric shear stiffness and surface flattening error.

10. A smart system for generating fitted garment patterns based on dynamic skin deformation, characterized in that: This system is used to implement the intelligent generation method for a fitted garment pattern based on dynamic skin deformation as described in claim 1. The intelligent generation system for a fitted garment pattern includes a data acquisition module, a model construction module, a flattening module, a flattening area correction module, a side seam dart correction module, a pattern generation module, and a pattern optimization module. The data acquisition module is used to collect the body shape data of the subject. The model construction module is used to construct a torso model of the subject. The flattening module is used to divide the torso model into analysis areas and flatten each analysis area. The flattening area correction module is used to correct the flattening area of ​​the analysis areas. The side seam dart correction module is used to correct the side seam darts of the analysis areas. The pattern generation module is used to generate an initial garment pattern based on the adjusted flattening area and side seam darts. The pattern optimization module is used to perform a virtual fitting based on the initial garment pattern to adjust the dart point coordinates and obtain the final fitted garment pattern.

Images