An Automated Garment Modeling Method that Reuses Sewing Information

Abstract

This invention discloses an automated garment modeling method that reuses sewing information. It extracts sewing relationship information from a completed garment model (2D pattern) and reuses this information in a new 3D pattern, achieving automated modeling. A garment model library containing sewing information and garment pattern outline information is constructed. The user inputs a 2D pattern DXF file, and the outline information of all garment pieces is extracted and constructed into closed polygons. The outline information of the garment model is represented by a multi-scale non-rigid closed polygon shape descriptor (MCC), and each garment piece is matched with garment pieces in the library to determine the seam line correspondence. Under boundary constraints, the UV coordinates of the library model are migrated to the 2D pattern closed polygons using the SLIM method. The original garment model and the garment pieces after UV mesh migration are recombined to generate a new garment model that meets the user's requirements. This invention has a high degree of automation, solving the problem that current garment CAD software requires cumbersome interactive processes for garment modeling.

Description

Technical Field

[0001] This invention relates to the field of 3D clothing CAD, specifically to an automated clothing modeling method that reuses sewing information to achieve fast and automatic clothing modeling that meets user requirements. Background Technology

[0002] Clothing modeling technology plays a vital role in film and television production, fashion design, game development, and virtual try-on. This makes clothing modeling technology one of the hottest research topics in computer graphics.

[0003] In recent years, clothing CAD software has developed rapidly, with well-known examples including Style3D from Lingdi (my country), CLO3D from South Korea, and VStitcher from the United States. Clothing modeling is a major function of clothing CAD, but existing systems all rely on interactive methods. The process involves inputting 2D garment pieces and simulating the stitching process within the CAD software to generate a 3D garment model. This process requires numerous tedious interactive operations: manually specifying the seam lines of each garment piece and their corresponding seam lines with other garment pieces, then dragging these pieces onto the human body model for stitching. If the garment designer wants to adjust the garment pieces or needs to generate 3D garment models in batches, these operations significantly reduce the designer's efficiency. To address this deficiency, this invention extracts the seam line correspondences of garment pieces from an interactively modeled garment model and transfers them to a target garment piece of the same style, improving clothing modeling efficiency and overcoming the shortcomings of current clothing CAD software. Summary of the Invention

[0004] To address the inefficiency of 2D garment modeling, this invention proposes a library-based automated garment modeling method. The aim of this invention is to provide a rapid and convenient method for automatically generating garment models. By correlating user-inputted garment piece information with garment models in a library, a new garment model is generated in an end-to-end manner, thereby improving the designer's work efficiency and reducing time-consuming operations.

[0005] The objective of this invention is achieved through the following technical solution: Firstly, this invention provides an automated garment modeling method that reuses sewing information, the method comprising the following steps:

[0006] Step 1: Construct a database of clothing models containing sewing information: Collect a template clothing models of different styles Ω={G0,G1,...,G a}, where Ω is the clothing model G aA set of clothing models (such as T-shirts, skirts, vests, pants, etc.), where clothing models of the same type can also be independent due to differences in the constituent pieces (e.g., a T-shirt can consist of a front piece, a back piece, a left sleeve, a right sleeve, a front collar, and a back collar (or no collar)), used as a reference for generating new clothing models; Clothing model G a ={V,F,U,L,C}, where V represents the 3D coordinates of the model, F represents the index of the mesh triangle, U is the UV mesh coordinates of the model, which represents the 2D garment pattern information corresponding to the garment model before sewing, L represents the seam information of the garment model, and C represents the outline information of each piece of the garment model.

[0007] Step 2: User inputs 2D clothing pattern: The user inputs a 2D clothing pattern stored in DXF file format. Several closed polygon sets S = {P0, P1, ..., P...} are extracted from the outline information of each garment piece and assembled. i ,…,P b}. Among them, the closed polygon P i ={v0,v1,…,v m The extracted closed polygons are stored as continuous point sets, with each closed polygon corresponding to a garment piece, such as (front piece, back piece, left sleeve, right sleeve, front collar, and back collar of a T-shirt). Hereafter, S represents the extracted set of closed polygons, and P represents the extracted set of closed polygons. i This represents a closed polygon that has been extracted.

[0008] Step 3, Model Matching: Based on the template garment model collected in Step 1, search for the best matching model of S in Step 2. The best matching model must meet the following conditions: First, the number of garment pieces in the template model must be greater than or equal to the number of polygons in S; second, the shape descriptor difference must be minimized in the database, as shown in the following expression:

[0009] minE s (S,C,L) (1)

[0010] Among them, E s This measures the difference in shape descriptors between two sets of closed polygons, where C represents the set of closed garment piece outline information in the template garment model, C = {c′0, c′1, ..., c′}. j ,…,c′ f}, where L represents the set of closed sewing line information in the template garment model, L = {P′0, P′1, ..., P′}. j ,…,P′ f Similar to step two, c′ j With P′ j Each polygon is a closed polygon composed of a continuous set of points, representing the closed contour information and the closed sewing line of a garment piece in the template clothing model, respectively. Each closed polygon corresponds to one garment piece; because c′j With P′ j There exists a one-to-one correspondence, obtained by P i With c′ j The matching relationship is used to obtain P. i With P′ j The matching relationship. Hereafter, C represents the set of closed garment piece outline information in the template garment model, L represents the set of closed sewing line information in the template garment model, and P′ represents the matching relationship. j ={v0,v1,…,v w} represents the closed sewing line of a garment piece extracted from the template garment model;

[0011] Step 4, Close the sewing thread separation: For P obtained in Step 2... i The corresponding P′ obtained in step three j Feature point extraction is performed. Feature points are those points whose angle sum is less than α degrees within a 1-ring neighborhood centered on the current point, arranged according to adjacency relationships. Secondly, based on the obtained feature points, P is... i With P′ j Perform closed sewing thread separation to obtain a closed polygon PE consisting of continuous edges. i ={e i0 ,e i1 ,…,e ip ,…,e in} and closed sewing thread PE′ j ={e′ j0 ,e′ j1 ,…,e′ jq ,…,e′ jn}. The following text will all refer to PE. i Represents a closed polygon P consisting of continuous edges with the feature point as its endpoint. i , with PE′ j P′ represents a closed sewing line consisting of continuous edges with the feature point as its endpoint. j With e ip Represents a closed polygon P i Based on one side after feature point separation, with e′ jq P′ represents a closed sewing thread j Based on a sewing thread after feature point separation; where edge e ip With e′ jq All are composed of continuous point sets, e ip ={v ip0 ,v ip1 ,…,v ips}, e′ jq ={v jq0 ,v jq1 ,…,v jqk ,…,vjqt}

[0012] Step 5, Sewing Relationship Construction: Based on the feature points obtained in Step 4 and P obtained in Step 3 i With P j The matching relationship is constructed using feature points as endpoints for precise sewing relationship construction. Under the premise of consistent feature point count, the difference in sewing relationship construction is minimized, as expressed below:

[0013] minE e (PE i ,PE′ j (2)

[0014] Among them, E e It is a measure of P i With P′ j The sewing relationship constructs the difference value to the minimum, and assigns e. ip Corresponding sewing relationship {e ip ,e′ jq};

[0015] Step Six, Resampling of Garment Sewing Edges: Based on the sewing relationships obtained in Step Five, resampling is performed on the closed polygon using the chord length parameterization method, as shown in the following expression:

[0016]

[0017] Where, v′ k Represents a closed polygon P i Based on the k-th point of the chord length parameterized resampling, l(e ip ) represents e ip The chord length, l(e′) jq ) represents e′ jq String length, v ipr e ip The r-th point in the middle; n(v ipr ,v ipr-1 ) indicates that v ipr-1 Starting from v ipr The unit direction vector is the endpoint; a closed polygon P consisting of a continuous set of points is obtained. newi ={v′0,v′1,…,v′ w}, and obtain a set S of several closed polygons. new ={P new0 ,P new1 ,…,P newi ,…,P newb Each closed polygon corresponds to one piece of clothing;

[0018] Step 7, Garment Piece Mesh Transfer: Using the template garment model UV mesh as a reference and the resampled sewing edges obtained in Step 6 as constraints, a parametric method based on fixed boundary points is used to transfer the template UV mesh to a closed polygon, generating a new UV mesh. The expression is as follows:

[0019] F(S new ,U)=U new (4)

[0020] Among them, U new The resampled closed polygon S obtained in step six new New UV mesh generated using parametric methods;

[0021] Step 8, Clothing Generation: Using the 3D clothing model G = {V, F, U, C, L} as a template, generate the new UV mesh U obtained in Step 7. new Replace the U in the template to generate a new clothing model, using the S obtained in step six. new Replace C and L in the template, which will become the final clothing model G. new ={V,F,U new ,S new}

[0022] Furthermore, in step one, the template clothing model is a mesh model, including 3D coordinates, UV coordinates, vertex connection relationships, seam information, and clothing pattern outline information; the template clothing model is stored in OBJ file format.

[0023] Furthermore, in step three, the expression representing the difference in overall shape similarity between the two models is as follows:

[0024] minE S (S,C,L)=min∑ ij ||M(P i )-M(c′ j )|| 2 (5)

[0025] Among them, M(P i M(c′) represents the shape descriptor of the i-th closed polygon in S. j (j) represents the shape descriptor of the j-th closed garment piece contour in the 3D clothing model. Using dynamic programming, and based on the premise that each closed polygon and the closed garment piece contour participate in at most one calculation, the shape descriptor difference is minimized. The non-multi-scale, non-rigid closed polygon shape descriptor (MCC) is used as the metric, and its calculation method is as follows:

[0026] M(P i )=matrix(d(μ,θ,P i (6)

[0027] (7)

[0028] x θ (μ)=∫x(μ)φ θ (δ-μ)dδ

[0029]

[0030] C θ ={v θ0 ,v θ1 ,…,v θu}

[0031] k = {1 if (x θ-1 (μ),y θ-1 (μ))inside C θ (μ)else 0}

[0032] Where θ is a constant representing the scale in MCC, δ,μ represents the δ,μ-th point out of a total of u points sampled on the contour, x(μ) is the X-coordinate of the μ-th point sample, and φ θ (t) represents the Gaussian convolution kernel at scale θ, v θμ =(x θ (μ),y θ (μ)) represents the μ-th feature coordinate at scale θ, x θ (μ),y θ (μ) represents the abscissa and ordinate of the feature, C θ Let d(μ,θ,P) be a characteristic closed curve at scale θ. i ) represents the feature of the closed contour at the μ-th point at scale θ. Gaussian convolution is used to obtain d(μ,θ,P) for each scale (1-θ) and each point. i The descriptors are composed of a matrix and used as shape descriptors. The differences between the descriptors are measured using methods including but not limited to the 2-norm. The goal is to obtain the best matching information between the closed polygon and the closed sewing line.

[0033] Furthermore, in step four, α is used to identify feature points of closed polygons or closed sewing lines, which is a threshold that is adjusted according to the characteristics of the garment model.

[0034] Furthermore, in step five, the metric P is represented. i With P j The expression for the difference in sewing relationships is as follows:

[0035] E e (PE i ,PE′ j ) = min q (∑ p||n(v p )-n(v′ q+p )|| 2 ,∑ p ||n(v p )-n(v′ q-p )|| 2 (8)

[0036] Among them, v p Represents a closed polygon PE i v′ is the p-th point under the sequential corner point. q±p PE′ represents closed sewing thread j The q±pth point under the sequential corner point, n(v p ) represents v p The normal vector of a point, E e (PE i ,PE′ j The goal is to find the optimal corner matching order, thereby constructing the best sewing relationship.

[0037] Furthermore, in step six, for each e ip , e must be retained ip The endpoints are used as the basis for e ip The endpoint during resampling.

[0038] Furthermore, in step seven, a new UV mesh is generated on the closed polygon using the Scalable Local Injection Model (SLIM) method based on fixed boundary points.

[0039] Furthermore, the stitch information is extracted from the mesh model information or entered into the database simultaneously with the model; a "triangular gap" will appear at the corner point, i.e., the inflection point, of the garment piece outline. Therefore, the angle is extracted using angular features, and the feature lines between the corner points are further extracted using geometric features as stitches.

[0040] Furthermore, the garment pattern outline information is extracted from the mesh model information or by extracting the garment piece outline from the image.

[0041] Furthermore, the suture extraction method is as follows:

[0042] First, the boundary lines of the UV mesh for the garment piece are extracted using the "half-side structure" of the mesh. Then, corner features are defined:

[0043]

[0044] Where link(z) D ) represents vertex z D The 1-ring neighborhood, T DB Let A(T) be the B-th vertex adjacent to vertex D.DB ) represents the angle of that vertex angle, A(z) D ) represents the angle of the boundary vertex; when α≤A(z) D When α ≤ β, the vertex is identified as a corner point, and we take α = 210° and β = 359°.

[0045] After detecting the corner points of the garment pieces, a half-edge structure is used to extract the seam lines of adjacent corner points. A breadth-first search (BFS) algorithm is employed to extract the mesh edges that constitute the seam edge one by one. Let z be the vertex of the mesh edge currently belonging to the seam edge. D Then, among all the adjacent edges of the current edge F, search for the edge that is most similar to the current edge F. D The side with the largest included angle As the next edge, the angle is equivalently represented using the vector dot product, which is:

[0046]

[0047] Since this step only needs to be calculated once, it can be pre-calculated, and the suture information can be stored in a database.

[0048] The main advantages of this invention include:

[0049] (a) The generated model is highly automated and adopts an end-to-end approach;

[0050] (b) Compared with other methods, this method generates clothing models based on the 2D clothing pattern input by the user, which is more in line with the user's requirements. Attached Figure Description

[0051] Figure 1 This is a flowchart of an automated garment modeling method that reuses sewing information, as described in one embodiment of the present invention.

[0052] Figure 2 This is a stage diagram of an automated garment modeling method that reuses sewing information, according to one embodiment of the present invention.

[0053] Figure 3 This is a schematic diagram of the separation and sewing relationship construction of the closed sewing thread obtained by the present invention.

[0054] Figure 4 This is a schematic diagram of a garment model generated by the present invention from a 2D garment piece graphic provided by the user. Detailed Implementation

[0055] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention.

[0056] It should be noted that in the claims and specification of this patent, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one" does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0057] Example

[0058] All documents mentioned in this invention are incorporated herein by reference as if each document were individually incorporated by reference. Furthermore, it should be understood that after reading the foregoing teachings of this invention, those skilled in the art can make various alterations or modifications to this invention, and these equivalent forms also fall within the scope defined by the appended claims.

[0059] The embodiments of the present invention relate to an automated garment modeling method that reuses sewing information. Using a library of garments as templates, a multi-scale non-rigid closed polygon shape descriptor (MCC) is used to perform closed polygon matching to search for the garment model in the library that best matches the user-input 2D garment pattern. Based on the template garment model, the corresponding sewing relationship is constructed on the user's 2D garment pattern. A new mesh is generated on the closed polygon using the SLIM method, while retaining the features of the template model and the user-input 2D garment pattern. The generated mesh is then combined with the template garment model to generate a new garment model.

[0060] Figure 1 This is a flowchart of an automated garment modeling method for reusing sewing information, according to one embodiment of the present invention. The method includes the following steps: Step 101, constructing a database containing as many garment models as possible; Step 102, the user provides a 2D garment pattern; Step 103, searching the database for the garment model most similar to the obtained 2D garment pattern; Step 104, matching the garment pieces of the 2D garment pattern with the garment pieces of the searched garment model; Step 105, based on the garment piece matching results, transferring the UV mesh information of the garment pieces of the garment model to the garment pattern provided by the user; Step 106, combining the mesh generated on the 2D garment pattern with the searched garment model to finally generate a new garment model that meets the user's requirements.

[0061] Specifically, in one Figure 2In this embodiment, the automated garment modeling method for reusing sewing information provided by the present invention includes the following steps:

[0062] Step 1: Construct a database of clothing models containing sewing information: Collect a template clothing models of different styles Ω={G0,G1,...,G a}, where Ω is the clothing model G a Collections (such as T-shirts, skirts, vests, pants, etc.), including common types like T-shirts (front piece, back piece, front collar, back collar, left sleeve, right sleeve) and dresses (front piece, back piece, belt, front left hem, front right hem, back left hem, back right hem), are stored in OBJ file format and used as references for generating new clothing models; clothing model G a ={V,F,U,L,C}, where V represents the 3D coordinates of the model, F represents the index of the mesh triangle, U is the UV mesh coordinates of the model, which represents the 2D garment pattern information corresponding to the garment model before sewing, L represents the seam information of the garment model, and C represents the outline information of each piece of the garment model.

[0063] Step 2, User provides 2D garment pattern: The user stores a 2D garment pattern S of a T-shirt in DXF file format. T恤 ={P 前衣片 ,P 后衣片 ,P 前衣领 ,P 后衣领 ,P 左衣袖 ,P 右衣袖}, and extract the sequence point set of each piece of clothing;

[0064] Step 3, Model Matching: Based on the template clothing model collected in Step 1, search for the model matching the S model from Step 2. T恤 The optimal matching model must meet the following conditions: First, the number of garment pieces in the template model must be greater than or equal to S. T恤 The number of garment pieces; secondly, the minimum difference in the multi-scale non-rigid shape descriptor (MCC) must be satisfied with the database, as expressed below:

[0065] minE s (S T恤 ,C,L)

[0066] Among them, E s Measuring the MCC difference between two sets of closed polygons, where C represents the set of closed garment piece outline information in the template garment model, C = {c′0, c′1, ..., c′}. j ,…,c′ f}, where L represents the set of closed sewing line information in the template garment model, L = {P′0, P′1, ..., P′}. j ,…,P′f}, because c′ j With P′ j There exists a one-to-one correspondence, obtained by obtaining P′ i With c′ j The matching relationship is used to obtain P. i With P′ j Matching relationships;

[0067] Step 4, Close the sewing thread separation: For P obtained in Step 2... i The corresponding P obtained in step three j Feature point extraction is performed. Feature points refer to points whose angle sum is less than 160 degrees within a 1-ring neighborhood centered on the current point, arranged according to adjacency relationships. Secondly, based on the obtained feature points, P is... i With P j Perform closed sewing thread separation to obtain a closed polygon PE consisting of continuous edges. i ={e i0 ,e i1 ,…,e ip ,…,e in} and closed sewing thread PE′ j ={e′ j0 ,e′ j1 ,…,e′ jq ,…,e′ jn},like Figure 3 As shown;

[0068] Step 5, Sewing Relationship Construction: Based on the feature points obtained in Step 4 and P obtained in Step 3 i With P j The matching relationship is constructed using feature points as endpoints for precise sewing relationship construction. Under the premise of consistent feature point count, the difference in sewing relationship construction is minimized, as expressed below:

[0069] minE e (PE i ,PE′ j )

[0070] Among them, E e It is a measure of P i With P′ j The sewing relationship constructs the difference value to the minimum, and assigns e. ip Corresponding sewing relationship {e ip ,e′ jq};

[0071] Step Six, Resampling of Garment Sewing Edges: Based on the sewing relationships obtained in Step Five, resampling is performed on the closed polygon using the chord length parameterization method, as shown in the following expression:

[0072]

[0073] Where, v′ k Represents a closed polygon P′ i Based on the k-th point of the chord length parameterized resampling, l(e ip ) represents e ip The chord length, l(e′) jq ) represents e′ jq String length, v ipr e ip The r-th point in the middle; n(v ipr ,v ipr-1 ) indicates that v ipr-1 Starting from v ipr The unit direction vector is the endpoint; this yields the closed polygon P. new衣片 ={v′0,v′1,…,v′ w}

[0074] With S newT恤 ={P new前衣片 ,P new后衣片 ,P new前衣领 ,P new后衣领 ,P new左衣袖 ,P new右衣袖},

[0075] And by retaining the sewing thread endpoints after the original closed polygon is separated by the sewing thread, the features of the user-input pattern are preserved;

[0076] Step 7, Garment Piece Mesh Transfer: Using the template garment model UV mesh as a reference and the resampled sewing edges obtained in Step 6 as constraints, the template UV mesh is transferred to a closed polygon using the Scalable Local Injection Model (SLIM) parameterization method based on fixed boundary points, generating a new UV mesh. The expression is as follows:

[0077] F(S newT恤 ,U)=U newT恤

[0078] Among them, U newT恤 The resampled closed polygon S obtained in step six newT恤 The new UV mesh generated by the SLIM method, such as Figure 4 As shown in the second row;

[0079] Step 8, Clothing Generation: Using the 3D clothing model G = {V, F, U, C, L} as a template, generate the new UV mesh U obtained in Step 7. newT恤 Replace the U in the template to generate a new clothing model, using the S obtained in step six. newT恤 Replace C and L in the template, which will become the final clothing model G. new={V,F,U newT恤 ,S newT恤},like Figure 4 As shown, the first row contains the 2D clothing pattern input by the user, the second row shows the result of transferring the UV mesh of the template clothing model to the 2D clothing pattern, and the third row shows the final generated clothing model. It is clear that the clothing model generated by this method meets the requirements of the user-input 2D clothing pattern. Furthermore, this invention has high automation efficiency; the user only needs to provide the 2D clothing pattern.

[0080] It should be noted that in the claims and specification of this patent, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one" does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0081] Although the invention has been illustrated and described with reference to certain preferred embodiments thereof, those skilled in the art will understand that various changes in form and detail may be made therein without departing from the spirit and scope of the invention.

Claims

1. An automated garment modeling method that reuses sewing information, characterized in that, The method includes the following steps: Step 1: Construct a database of clothing models containing sewing information: Collect a template clothing models of different styles Ω = {G0, G1, ..., G...} a }, where Ω is the set of clothing models, and G a This represents the a-th clothing model. Clothing models of the same type can also be independent due to differences in the constituent garment pieces, and are used as references for generating new clothing models; Clothing model G a ={V, F, U, L, C}, where V represents the 3D coordinates of the garment model, F represents the index of the mesh triangle, U is the UV mesh coordinates of the model, which represents the 2D garment pattern information corresponding to the garment model before sewing, L represents the seam information of the garment model, and C represents the outline information of each piece of the garment model. Step 2, Obtain the 2D clothing pattern: Extract and form several closed polygon sets S = {P} from the outline information of the clothing pattern. ０ P1, ..., P i , ..., P b }; where the closed polygon P i ={v0, v1, ..., v m It consists of a continuous set of points, with each closed polygon corresponding to a piece of clothing; Step 3, Model Matching: Based on the template garment model collected in Step 1, search for the best matching model of S in Step 2. The best matching model must meet the following conditions: First, the number of garment pieces in the template garment model must be greater than or equal to the number of polygons in S; second, the shape descriptor difference must be minimized in the database, as shown in the following expression: minE S (S,C,L) Among them, E s This measures the difference in shape descriptors between two sets of closed polygons, where C represents the set of closed garment piece outline information in the template garment model, C = {c′0, c′1, ..., c′}. j , ..., c′ f }, where L represents the set of closed sewing line information in the template garment model, L = {P′0, P′1, ..., P}. j ′,...,P′ f }, c′ j With P j Each of the following is a closed polygon composed of a continuous set of points, representing the closed contour information of the j-th garment piece and the closed sewing line of the j-th garment piece in the template garment model, respectively. Each closed polygon corresponds to one garment piece; because c′ j With P j There exists a one-to-one correspondence, obtained by P. i With c′ j The matching relationship is used to obtain P. i With P i The matching relationship of '; where P j ′={v0,v1,...,v w }; Step 4, Close the sewing thread separation: Compare Pi obtained in Step 2 with the corresponding P obtained in Step 3. j Feature point extraction is performed. Feature points are those points whose angle, centered on the current point and within its 1-neighborhood vertices, is less than a threshold γ, arranged according to adjacency. Secondly, based on the obtained feature points, P is... i With P j Perform closed sewing thread separation to obtain a closed polygon PE consisting of continuous edges. i ={e i0 e i1 , ..., e ip , ..., e in } and closed sewing thread PE′ j ={e′ j0 ,e′ j1 , ..., e′ jq , ..., e′ jn }; with PE i Represents a closed polygon P consisting of continuous edges with the feature point as its endpoint. i , with PE′ j P represents a closed sewing line consisting of continuous edges with the feature point as its endpoint. j ′;with e ip Represents a closed polygon P i Based on the p-th edge after feature point separation, with e′ jq P represents closed sewing thread j The q-th sewing thread is based on feature point separation; where edge e ip With e′ jq All are composed of continuous point sets, e ip ={v ip0 v ip1 , ..., v ips }, e′ jq ={v jq0 v jq1 , ..., v jqk , ..., v jqt }; Step 5, Sewing Relationship Construction: Based on the feature points obtained in Step 4 and P obtained in Step 3 i With P j The matching relationship is constructed using feature points as endpoints for precise sewing relationship construction. Under the premise of consistent feature point count, the difference in sewing relationship construction is minimized, as expressed below: me e (ON i ,ON' j ) Among them, E e It is a measure of P i With P j The sewing relationship constructs the difference value to the minimum, and assigns e. ip Corresponding sewing relationship {e ip ,e′ jq }; Step 6, Resampling of garment seam edges: Based on the seam relationship obtained in Step 5, resampling is performed on the closed polygon Pi using the chord length parameterization method, as shown in the following expression: Where, v′ k Represents a closed polygon P i Based on the k-th point of the chord length parameterized resampling, l(e ip ) represents e ip The chord length, l(e′) jq ) represents e′ jq String length, v ipr e ip The r-th point in the middle; n(v ipr v ipr-1 ) indicates that v ipr-1 Starting from v ipr The unit direction vector is the endpoint; a closed polygon P consisting of a continuous set of points is obtained. newi ={v′0, v′1,...,v′ w }, and obtain a set S of several closed polygons. new ={P new0 P new1 , ..., P newi , ..., P newb Each closed polygon corresponds to one piece of clothing; Step 7, Garment Mesh Transfer: Using the UV mesh of the template garment model as a reference and the resampled sewing edges obtained in Step 6 as constraints, a parametric method based on fixed boundary points is used to transfer the UV mesh to a closed polygon, generating a new UV mesh. The expression is as follows: F(S new ,U)=U new Among them, U new The resampled closed polygon S obtained in step six new New UV mesh generated using parametric methods; Step 8, Clothing Model Generation: Using the 3D clothing model G... n ={V, F, U, C, L} is used as a template to apply the new UV mesh model U obtained in step seven. new Replace the U in the template to generate a new clothing model, using the S obtained in step six. new Replace C and L in the template, which will become the final clothing model G. new ={V, F, U new S new } 2. The automated garment modeling method for reusing sewing information according to claim 1, characterized in that, In step one, the template clothing model is a mesh model, including 3D coordinates, UV coordinates, vertex connection relationships, seam information, and clothing pattern outline information; the template clothing model is stored in OBJ file format.

3. The automated garment modeling method for reusing sewing information according to claim 1, characterized in that, In step three, the expression representing the difference in shape descriptors between two closed polygon sets is as follows: minE s (S,C,L)=min∑ ij ||M(P i )-M(c′ j )|| 2 Among them, M(P ｉ M(c′) represents the shape descriptor of the i-th closed polygon in S. j (j) represents the shape descriptor of the j-th closed garment piece contour in the 3D clothing model. Using dynamic programming, and based on the premise that each closed polygon and the closed garment piece contour participate in at most one calculation, while satisfying the condition of minimum difference between shape descriptors, the non-multi-scale, non-rigid closed polygon shape descriptor MCC is used as the metric. Its calculation method is as follows: M(P ｉ )=matrix(d(μ,θ,P i )) x θ (μ)=∫x(μ)φ θ (d-m)dδ C θ ={v θ0 ,v θ1 ,...,v θu } k={1 if(x θ-1 (m),y θ-1 (m))inside C θ (μ)else 0} Where θ is a constant representing the scale in MCC, δ and μ represent the δ-th and μ-th points out of a total of u points sampled on the contour, x(μ) is the X-coordinate of the μ-th point sample, and φ θ (t) represents the Gaussian convolution kernel at scale θ, v θμ =(x θ (μ), y θ (μ)) represents the μ-th feature coordinate at scale θ, x θ (μ), y θ (μ) represents the abscissa and ordinate of the feature, C θ Let d(μ, θ, P) be a characteristic closed curve at scale θ. i ) represents the feature of the closed contour at the μ-th point at scale θ. Gaussian convolution is used to obtain d(μ, θ, P) for each scale (1-θ) and each point. i The descriptors are composed of a matrix and used as shape descriptors. The difference between the descriptors is measured using the L2 norm method. The goal is to obtain the best matching information between the closed polygon and the closed sewing line.

4. The automated garment modeling method for reusing sewing information according to claim 1, characterized in that, In step four, α is used to identify feature points of closed polygons or closed sewing lines. It is a threshold that is adjusted according to the characteristics of the garment model.

5. The automated garment modeling method for reusing sewing information according to claim 1, characterized in that, In step five, the metric P is represented. i With P j The expression for constructing the difference value of the sewing relationship is as follows: It is e (ON i ,ON' j )=min q (∑ p ||n(v p )-n(v′ q+p )|| 2 ,∑ p ||n(v p )-n(v′ q-p )|| 2 ) Among them, v p Represents a closed polygon PE i v′ is the p-th point under the sequential corner point. q±p PE′ represents closed sewing thread j The q±pth point under the sequential corner point, n(v p ) represents v p The normal vector of a point, E e (PE i , PE′ j The goal is to find the optimal corner matching order, thereby constructing the best sewing relationship.

6. The automated garment modeling method for reusing sewing information according to claim 1, characterized in that, In step six, for each e ip , e must be retained ip The endpoints are used as the basis for e ip The endpoint during resampling.

7. The automated garment modeling method for reusing sewing information according to claim 1, characterized in that, In step seven, a new UV mesh is generated on the closed polygon using the Scalable Local Injection Model (SLIM) method based on fixed boundary points.

8. The automated garment modeling method for reusing sewing information according to claim 2, characterized in that, The stitch information is extracted from the mesh model information or entered into the database at the same time as the model; a "triangular gap" will appear at the corner point, i.e. the inflection point, of the garment piece outline. Therefore, the angle is extracted by using angle features, and the feature lines between the corner points are further extracted as stitches using geometric features.

9. The automated garment modeling method for reusing sewing information according to claim 2, characterized in that, The garment pattern outline information is extracted from the mesh model information or by extracting the garment piece outline from the image.

10. The automated garment modeling method for reusing sewing information according to claim 8, characterized in that, The specific method for suture extraction is as follows: First, the boundary lines of the garment piece UV mesh are extracted using the "half-side structure" of the mesh, and then corner features are defined: Where link(z) D ) represents vertex z D The 1-ring1 neighborhood, T DB Let A(T) be the B-th vertex adjacent to vertex D. DB ) represents the angle of that vertex angle, A(z) D ) represents the angle of the boundary vertex; when α≤A(z) D When α ≤ β, the vertex is identified as a corner point, and we take α = 210° and β = 359°. After detecting the corner points of the garment pieces, a half-edge structure is used to extract the seam lines of adjacent corner points. A breadth-first search (BFS) algorithm is employed to extract the mesh edges that constitute the seam edge one by one. Let z be the vertex of the mesh edge currently belonging to the seam edge. D Then, among all the adjacent edges of the current edge F, search for the edge that is most similar to the current edge F. D The side with the largest included angle As the next edge, the angle is equivalently represented using the vector dot product, which is: Since this step only needs to be calculated once, it can be pre-calculated, and the suture information can be stored in a database.

Images