Intelligent derivative method and system for ip of clothing and household articles

Abstract

The application provides an IP intelligent derivation method and system for clothing and household articles, relates to the technical field of intelligent design, and comprises the following steps: IP representation is subjected to manifold transformation and decomposition by constructing a cross-modal mapping space, guided projection is carried out in combination with function and structure constraints to obtain adaptive representation, and candidate forms are generated and screened. Manufacturability evaluation and conflict solving are carried out based on a manufacturing constraint graph, and a physical implementation scheme is obtained. The application realizes intelligent form derivation and manufacturability guarantee from IP to product.

Description

Technical Field

[0001] This invention relates to the field of intelligent design technology, and in particular to an IP intelligent derivation method and system for clothing and daily necessities. Background Technology

[0002] In the field of apparel and lifestyle product design, transforming cultural and creative IPs into concrete products is an important design direction. Existing technologies typically extract the visual characteristics of the IP and apply them directly to preset templates or basic models of the target product. The mapping process often relies on pre-established product form libraries and style element libraries, generating preliminary derivative design schemes through matching and combination.

[0003] Conventional approaches focus on the surface transfer of IP visual features, with the core being the attachment of style and texture or the simple adaptation of shape and outline. However, solutions generated by conventional derivation methods may only satisfy visual similarity, while ignoring the functional constraints and physical structural limitations necessary for the product category itself. This makes it difficult for the initial design concept to be implemented in the subsequent engineering phase, requiring a lot of manual adjustments and redesign.

[0004] Meanwhile, the design evaluation phase of existing methods is usually placed after the form and style are generated, or even relies entirely on manual review in the later stages. The early derivation process fails to effectively incorporate physical manufacturing constraints such as material properties, production processes, and assembly logic, resulting in many creative solutions being identified as unmanufacturable or too costly before being put into production, causing problems such as long design iteration cycles and low efficiency. Summary of the Invention

[0005] This invention provides a method and system for intelligent IP derivatives for clothing and daily necessities, which can at least solve some of the problems existing in the prior art.

[0006] A first aspect of this invention provides a method for intelligent IP derivatives for clothing and daily necessities, comprising:

[0007] Obtain an IP prototype and extract the corresponding IP representation. Construct a cross-modal mapping space based on the IP representation and the functional and structural constraints corresponding to the target product category. Perform a manifold transformation on the IP representation in the cross-modal mapping space to obtain a product domain representation. Perform component decomposition to obtain morphological and style components. Calculate the coupling relationship between the morphological components and the functional constraints, and the compatibility relationship between the style components and the structural constraints. Based on the coupling and compatibility relationships, perform constraint-guided projection on the product domain representation to obtain an adaptive representation.

[0008] The adaptation representation is projected onto the product form space and geometric transformation is performed based on preset form topology constraints to generate a candidate form set. The feature consistency metric between each candidate form in the candidate form set and the IP prototype is calculated. Based on the feature consistency metric, the candidate form set is filtered to obtain a derived form.

[0009] An initial constraint diagram is constructed based on the geometric structure of the derived form, and physical attribute nodes and operation sequence nodes corresponding to the target product category are added to obtain a manufacturing constraint diagram. Constraint propagation calculation is performed on the manufacturing constraint diagram to obtain the manufacturability evaluation value corresponding to the derived form. Conflict regions are identified based on the manufacturability evaluation value, and the geometric parameters of the conflict regions are constrained and satisfied to obtain a physical implementation scheme.

[0010] In one alternative implementation,

[0011] Obtain the IP prototype and extract the corresponding IP representation. Based on the IP representation and the functional and structural constraints corresponding to the target product category, construct a cross-modal mapping space. Perform a manifold transformation on the IP representation in the cross-modal mapping space to obtain the product domain representation, and perform component decomposition to obtain morphological and style components, including:

[0012] Multi-level feature extraction is performed on the IP prototype to obtain a visual feature set and a semantic feature set. Multimodal attention mechanism is then applied to the visual feature set and the semantic feature set respectively, and weighted aggregation is performed to obtain the IP representation.

[0013] Obtain the product sample set corresponding to the target product category, perform graph structure parsing on the structural topology of each product sample in the product sample set and extract graph convolution features to determine structural constraints, and perform semantic parsing on the functional description of the product sample set to obtain the functional constraints;

[0014] Based on the IP representation and the product sample set, a source domain space and a target domain space are constructed. The distribution difference metric between the source domain space and the target domain space is calculated and a domain mapping matrix is ​​constructed. Matrix multiplication and manifold regularization constraint calculations are performed on the IP representation and the domain mapping matrix to obtain a manifold embedding vector. The manifold embedding vector is then locally linearly reconstructed in the target domain space to obtain the product domain representation.

[0015] Independent component analysis is performed on the product domain characterization to obtain multiple independent components. The correlation between each independent component and the structural constraints is calculated and the structurally sensitive components are labeled. The structurally sensitive components are aggregated to obtain morphological components. The matching degree between each independent component and the functional constraints is calculated and the style-sensitive components are labeled. The style-sensitive components are aggregated to obtain style components.

[0016] In one alternative implementation,

[0017] Calculating the coupling relationship between the morphological components and the functional constraints, and the compatibility relationship between the style components and the structural constraints, and performing constraint-guided projection on the product domain representation based on the coupling and compatibility relationships to obtain an adapted representation includes:

[0018] The morphological components are represented as morphological feature vectors, the functional constraints are represented as functional requirement vectors, and the cosine similarity between the morphological feature vectors and the functional requirement vectors is calculated to obtain the initial coupling degree. The functional requirement vectors are orthogonally decomposed to obtain rigid requirement components. The projection length between the morphological feature vectors and the rigid requirement components is calculated, and the coupling relationship is obtained by combining the initial coupling degree.

[0019] The style components are represented as style feature vectors and the structural constraints are represented as structural topology vectors. The interaction response value between the style feature vectors and the structural topology vectors is calculated using a bilinear mapping matrix to obtain the initial compatibility. The structural topology vectors are decomposed into a graph Laplace decomposition to obtain a set of structural eigenmodes. The projection coefficients of the style feature vectors on the set of structural eigenmodes are calculated. The compatibility relationship is calculated based on the projection coefficients and the initial compatibility.

[0020] The coupling relationship and the compatibility relationship are converted into a joint constraint vector. The gradient of the joint constraint vector is calculated to obtain the guiding direction vector. The product domain representation is projected along the guiding direction vector to obtain an intermediate representation. The intermediate representation is corrected by manifold constraints and the residual is calculated. The correction is repeated iteratively until the residual converges to obtain the adaptive representation.

[0021] In one alternative implementation,

[0022] The adaptation representation is projected onto the product form space and geometrically transformed based on preset form topology constraints to generate a candidate form set. The feature consistency metric between each candidate form in the candidate form set and the IP prototype is calculated, including:

[0023] Obtain a parameterized template of the product form space corresponding to the target product category, map the adaptation representation to the control parameters of the parameterized template to obtain an initial parameter configuration, instantiate the parameterized template based on the initial parameter configuration to obtain an initial product form, perform three-dimensional meshing on the initial product form and extract the mesh vertex coordinates;

[0024] The position and size constraints of key components are extracted from the preset morphological topology constraints. Based on the position and size constraints, the coordinates of the mesh vertices are subjected to local deformation operations to obtain multiple product forms. The geometric feasibility of the multiple product forms is tested and the product forms that do not meet the preset manufacturing requirements are filtered to obtain the candidate form set.

[0025] Contour extraction is performed on the IP prototype and each candidate form in the candidate form set. Shape moments are calculated based on the extracted contours. The Euclidean distance between the shape moments of the candidate forms and the shape moments of the IP prototype is calculated and normalized to obtain the shape difference degree. Visual element features are extracted from the IP prototype and the candidate forms and feature matching is performed to obtain the element retention rate. The feature consistency metric is obtained by inverse weighting the shape difference degree and the element retention rate.

[0026] In one alternative implementation,

[0027] Based on the feature consistency metric, the candidate morphology set is filtered to obtain the following derived morphologies:

[0028] For each candidate form in the candidate form set, determine the corresponding feature consistency metric and calculate the statistical distribution parameter. Based on the statistical distribution parameter, construct an adaptive screening threshold and screen the candidate form set to obtain a subset of highly consistent candidate forms.

[0029] The manufacturing-related parameters and cost-related parameters corresponding to each candidate form in the high-consistency candidate form subset are extracted. The manufacturing constraint satisfaction is calculated based on the manufacturing-related parameters and preset manufacturing constraints. The cost constraint satisfaction is calculated based on the cost-related parameters and preset cost constraints. The cost constraint satisfaction and the manufacturing constraint satisfaction are optimized through Pareto front analysis to obtain a feasibility score. The feature consistency metric and the feasibility score are jointly evaluated to obtain a comprehensive evaluation score.

[0030] Feature vectors are extracted from the candidate morphologies in the high consistency candidate morphology subset, and a feature matrix is ​​constructed. The Hausdorff distance between any two candidate morphologies in the feature matrix is ​​calculated to obtain a morphological difference matrix. Based on the morphological difference matrix, the difference index between candidate morphologies is calculated and combined with the comprehensive evaluation score to obtain a dual-criteria ranking result. The derived morphology is extracted based on the dual-criteria ranking result.

[0031] In one alternative implementation,

[0032] An initial constraint diagram is constructed based on the geometric structure of the derived form, and physical attribute nodes and operation sequence nodes corresponding to the target product category are added to obtain a manufacturing constraint diagram. Constraint propagation calculations are performed on the manufacturing constraint diagram to obtain the manufacturability evaluation value corresponding to the derived form, including:

[0033] Extract the surface topology information and boundary feature information corresponding to the derived morphology, identify the surface connection relationship based on the surface topology information and construct a topology connection matrix, identify feature boundaries based on the boundary feature information and extract boundary constraints, perform a structured expression of the topology connection matrix and the boundary constraints and construct an initial constraint graph, retrieve physical attribute requirements from the manufacturing knowledge base corresponding to the target product category and generate physical attribute nodes, retrieve the processing operation sequence from the process knowledge base of the target product category and generate operation sequence nodes, insert the physical attribute nodes and the operation sequence nodes into the initial constraint graph and establish dependencies to obtain the manufacturing constraint graph;

[0034] The nodes in the manufacturing constraint graph are labeled with types and a node type mapping table is constructed. All nodes in the manufacturing constraint graph are traversed and the constraint propagation path is determined. Iterative constraint propagation calculation is performed along the constraint propagation path and constraint conflicts are detected. Conflicting nodes are recorded and constraint relaxation is performed. This process is repeated until the states corresponding to all nodes converge. The degree of constraint satisfaction is statistically analyzed and the manufacturability evaluation value corresponding to the derived form is calculated.

[0035] In one alternative implementation,

[0036] Based on the manufacturability assessment value, conflict regions are identified, and the geometric parameters of the conflict regions are constrained to obtain a physical implementation scheme through solution, including:

[0037] Local gradient analysis is performed on the manufacturability assessment value, and the derived morphological regions where the gradient change rate exceeds a preset change rate threshold are marked as candidate conflict regions. The constraint conflict type and conflict intensity corresponding to the candidate conflict regions are extracted. Based on the constraint conflict type, the candidate conflict regions are divided into material conflict regions and process conflict regions. Based on the conflict intensity, the conflict degree of the material conflict regions and the process conflict regions is quantitatively evaluated. Adjacent conflict regions of the same type are merged, and the region boundary is determined based on the conflict degree quantitative evaluation result to obtain the conflict region.

[0038] Extract the geometric parameters corresponding to the conflict region, perform parameter space mapping, and construct a parameter constraint network. Mark hard constraint edges and soft constraint edges in the parameter constraint network. Construct a constraint propagation tree based on the hard constraint edges and perform layer-by-layer constraint solving to obtain the hard constraint solution space. Construct an objective function based on the soft constraint edges and solve it to obtain the soft constraint solution space. Calculate the intersection of the hard constraint solution space and the soft constraint solution space and filter parameter configurations that satisfy constraint compatibility. Adjust the parameters of the geometry of the conflict region based on the parameter configuration to obtain corrected geometric data. Determine the processing sequence and assembly sequence based on the corrected geometric data and associate them to obtain a physical implementation scheme.

[0039] A second aspect of this invention provides an IP-based intelligent derivative system for clothing and daily necessities, comprising:

[0040] The IP adaptation unit is used to acquire an IP prototype and extract the corresponding IP representation. Based on the IP representation and the functional and structural constraints corresponding to the target product category, a cross-modal mapping space is constructed. In the cross-modal mapping space, the IP representation is subjected to manifold transformation to obtain a product domain representation. Component decomposition is then performed to obtain morphological and style components. The coupling relationship between the morphological components and the functional constraints, as well as the compatibility relationship between the style components and the structural constraints, are calculated. Based on the coupling and compatibility relationships, the product domain representation is subjected to constraint-guided projection to obtain an adaptation representation.

[0041] The morphology derivation unit is used to project the adaptation representation onto the product morphology space and perform geometric transformations based on preset morphology topology constraints to generate a candidate morphology set, calculate the feature consistency metric value between each candidate morphology in the candidate morphology set and the IP prototype, and filter the candidate morphology set based on the feature consistency metric value to obtain a derived morphology.

[0042] The manufacturing constraint unit is used to construct an initial constraint diagram based on the geometric structure of the derived form and add physical attribute nodes and operation sequence nodes corresponding to the target product category to obtain a manufacturing constraint diagram. It performs constraint propagation calculation on the manufacturing constraint diagram to obtain the manufacturability evaluation value corresponding to the derived form. Based on the manufacturability evaluation value, it identifies conflict areas and performs constraint satisfaction calculation on the geometric parameters of the conflict areas to obtain a physical implementation scheme.

[0043] A third aspect of the present invention provides an electronic device, comprising:

[0044] A processor and a memory for storing processor-executable instructions, wherein the processor is configured to invoke instructions stored in the memory to perform the aforementioned method.

[0045] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0046] In this invention, by constructing a cross-modal mapping space and performing manifold transformation, IP representation is transformed into product domain representation, which is then decomposed into morphological and style components. The calculation of the coupling relationship between morphological components and functional constraints ensures that the derived form meets the practical needs of the target product, while the analysis of the compatibility relationship between style components and structural constraints ensures the effective continuation of the IP's aesthetic characteristics. Constraint-guided projection generates adaptive representations based on the aforementioned relationships, achieving deep adaptation between core IP features and product functions and structural requirements. This lays a precise semantic foundation for subsequent form generation. After projecting the adaptive representations onto the product form space, geometric transformations are performed based on form topology constraints, systematically generating a diverse set of candidate forms. By calculating and filtering the feature consistency metrics between candidate forms and the IP prototype, visual and semantic features highly correlated with the IP prototype are effectively preserved. While ensuring form innovation, the recognizability and stylistic consistency of the IP are maintained, making the derived forms both novel and faithful to the original IP. Based on the selected derived forms, a manufacturing constraint graph is constructed, integrating physical attributes and operation sequence nodes to achieve knowledge mapping from the design domain to the manufacturing domain. By constraining and solving the geometric parameters of conflict areas, a physical implementation scheme that meets actual production conditions can be automatically generated, significantly shortening the iteration cycle from conceptual design to a manufacturable solution and reducing later modification costs. Attached Figure Description

[0047] Figure 1 This is a flowchart illustrating the IP smart derivation method for clothing and daily necessities according to an embodiment of the present invention;

[0048] Figure 2 This is a flowchart illustrating the manufacturability assessment of the derivative forms of the IP smart derivative method for clothing and daily necessities according to an embodiment of the present invention. Detailed Implementation

[0049] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0050] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0051] Figure 1 This is a flowchart illustrating the IP smart derivative method for clothing and daily necessities according to an embodiment of the present invention. Figure 1 As shown, the method includes:

[0052] Obtain an IP prototype and extract the corresponding IP representation. Construct a cross-modal mapping space based on the IP representation and the functional and structural constraints corresponding to the target product category. Perform a manifold transformation on the IP representation in the cross-modal mapping space to obtain a product domain representation. Perform component decomposition to obtain morphological and style components. Calculate the coupling relationship between the morphological components and the functional constraints, and the compatibility relationship between the style components and the structural constraints. Based on the coupling and compatibility relationships, perform constraint-guided projection on the product domain representation to obtain an adaptive representation.

[0053] The adaptation representation is projected onto the product form space and geometric transformation is performed based on preset form topology constraints to generate a candidate form set. The feature consistency metric between each candidate form in the candidate form set and the IP prototype is calculated. Based on the feature consistency metric, the candidate form set is filtered to obtain a derived form.

[0054] An initial constraint diagram is constructed based on the geometric structure of the derived form, and physical attribute nodes and operation sequence nodes corresponding to the target product category are added to obtain a manufacturing constraint diagram. Constraint propagation calculation is performed on the manufacturing constraint diagram to obtain the manufacturability evaluation value corresponding to the derived form. Conflict regions are identified based on the manufacturability evaluation value, and the geometric parameters of the conflict regions are constrained and satisfied to obtain a physical implementation scheme.

[0055] In one alternative implementation,

[0056] Obtain the IP prototype and extract the corresponding IP representation. Based on the IP representation and the functional and structural constraints corresponding to the target product category, construct a cross-modal mapping space. Perform a manifold transformation on the IP representation in the cross-modal mapping space to obtain the product domain representation, and perform component decomposition to obtain morphological and style components, including:

[0057] Multi-level feature extraction is performed on the IP prototype to obtain a visual feature set and a semantic feature set. Multimodal attention mechanism is then applied to the visual feature set and the semantic feature set respectively, and weighted aggregation is performed to obtain the IP representation.

[0058] Obtain the product sample set corresponding to the target product category, perform graph structure parsing on the structural topology of each product sample in the product sample set and extract graph convolution features to determine structural constraints, and perform semantic parsing on the functional description of the product sample set to obtain the functional constraints;

[0059] Based on the IP representation and the product sample set, a source domain space and a target domain space are constructed. The distribution difference metric between the source domain space and the target domain space is calculated and a domain mapping matrix is ​​constructed. Matrix multiplication and manifold regularization constraint calculations are performed on the IP representation and the domain mapping matrix to obtain a manifold embedding vector. The manifold embedding vector is then locally linearly reconstructed in the target domain space to obtain the product domain representation.

[0060] Independent component analysis is performed on the product domain characterization to obtain multiple independent components. The correlation between each independent component and the structural constraints is calculated and the structurally sensitive components are labeled. The structurally sensitive components are aggregated to obtain morphological components. The matching degree between each independent component and the functional constraints is calculated and the style-sensitive components are labeled. The style-sensitive components are aggregated to obtain style components.

[0061] Feature extraction is performed on the IP prototype. A convolutional neural network is used to extract multi-scale features from the IP prototype image, acquiring edge texture features, local shape features, and global semantic features from shallow to deep layers, forming a visual feature set. Word embedding encoding is performed on the descriptive text of the IP prototype, and semantic feature vectors are extracted using a pre-trained language model. These semantic feature vectors constitute a semantic feature set. A multimodal attention mechanism is executed to map the visual and semantic feature sets to a common feature space. An attention weight matrix is ​​calculated between the visual and semantic features, and the features are weighted and aggregated using this attention weight matrix to obtain an IP representation vector that integrates visual and semantic information.

[0062] After obtaining the product sample set corresponding to the target product category, a structural topology analysis is performed on each product sample. The geometric structure of the product is converted into a graph structure representation, where nodes represent key components and edges represent the connections between components. A graph convolutional neural network is used to extract features from the graph structure, capturing the topological connection patterns and spatial layout relationships of the product. The graph convolutional features reflect the structural constraints of the product, including the relative positional relationships between components, connection methods, and overall structural stability requirements. For functional constraint extraction, the functional description text of each product in the product sample set is collected. Semantic parsing is performed using natural language processing techniques to identify functional keywords and attributes, establishing a vectorized representation of functional requirements, forming a set of functional constraints.

[0063] A cross-modal mapping space is constructed, defining the feature space containing the IP representation as the source domain space and the feature space of the product sample set as the target domain space. The distribution difference metric between the source and target domain spaces is calculated using the maximum mean difference method, reflecting the distribution distance between the two domains. A domain mapping matrix is ​​constructed based on the distribution difference metric, which maps features from the source domain to the distribution space of the target domain. Matrix multiplication is performed on the IP representation vector and the domain mapping matrix to obtain the preliminary mapping result. To preserve the manifold structure during the mapping process, manifold regularization constraints are introduced. Laplacian eigenmaps are used to maintain the neighborhood relationships of samples in the original space, ensuring that the mapped features retain their original intrinsic geometric structure. After manifold regularization, a manifold embedding vector is obtained, which lies on the manifold of the target domain space.

[0064] When performing local linear reconstruction of a manifold embedding vector, the nearest neighbor sample points of the manifold embedding vector are found in the target domain space. The manifold embedding vector is represented by a linear combination of these nearest neighbor sample points. The optimal linear combination coefficients are obtained by minimizing the reconstruction error. Based on these linear combination coefficients, the nearest neighbor sample points are weighted and combined to obtain the reconstructed representation in the target domain space, i.e., the product domain representation. The product domain representation retains the core features of the IP prototype while adapting to the distribution characteristics of the target product category.

[0065] Independent Component Analysis (ICA) is performed on the product domain representation, decomposing the product domain representation vector into multiple statistically independent components. ICA achieves separation by maximizing the non-Gaussianity among the components, ensuring that each independent component represents an independent source of information in the product domain representation. For each independent component, the correlation between the independent component and structural constraints is calculated. Structural constraints include the product's topological and geometric layout features. By calculating the correlation coefficient between the independent components and these structural features, independent components that strongly respond to structural features are identified and marked as structurally sensitive components. Structurally sensitive components primarily capture the product's morphological features, including the shape, size proportions, and spatial arrangement of components. All structurally sensitive independent components are aggregated, and morphological components are obtained through weighted summation or concatenation.

[0066] For style component extraction, the matching degree between each independent component and the functional constraints is calculated. Functional constraints describe the functional characteristics and performance requirements that the product should possess. The degree to which each independent component expresses the functional characteristics is evaluated by calculating the cosine similarity or inner product between the feature vector of the independent component and the functional constraint vector. Independent components with a matching degree higher than a pre-set matching degree threshold are marked as style-sensitive components. Style-sensitive components mainly reflect the product's decorative features, color style, and surface texture—attributes related to the IP prototype's style. All style-sensitive independent components are aggregated to obtain the style components. Style components retain the visual style characteristics of the IP prototype and can reflect the unique aesthetic features of the IP in product design.

[0067] In this embodiment, by performing multi-level feature extraction on the IP prototype and combining it with a multimodal attention mechanism to weightedly fuse visual and semantic information, the completeness and accuracy of IP representation are effectively improved. This allows for a more comprehensive depiction of the core features of the IP, reducing information loss and semantic bias. By performing graph structure analysis and semantic analysis on the sample set of the target product category, structural constraints and functional constraints are constructed respectively, enabling explicit modeling of structural relationships and functional attributes in product design. This significantly enhances the accuracy and computability of constraint expression, thereby improving the rationality and adaptability of design constraints. By constructing source and target domain spaces and calculating distribution differences, a domain mapping matrix and manifold regularization mechanism are introduced to achieve effective alignment and smooth mapping of cross-domain features. This reduces information distortion caused by inter-domain distribution shifts and improves the stability and consistency of IP features during migration to the product domain. By performing local linear reconstruction in the target domain space, the local structural characteristics of the data are further maintained, making the generated product domain representation more consistent with the inherent distribution patterns of the real product space. This effectively reduces structural distortion and semantic shifts, improving the interpretability and reliability of the representation.

[0068] In one alternative implementation,

[0069] Calculating the coupling relationship between the morphological components and the functional constraints, and the compatibility relationship between the style components and the structural constraints, and performing constraint-guided projection on the product domain representation based on the coupling and compatibility relationships to obtain an adapted representation includes:

[0070] The morphological components are represented as morphological feature vectors, the functional constraints are represented as functional requirement vectors, and the cosine similarity between the morphological feature vectors and the functional requirement vectors is calculated to obtain the initial coupling degree. The functional requirement vectors are orthogonally decomposed to obtain rigid requirement components. The projection length between the morphological feature vectors and the rigid requirement components is calculated, and the coupling relationship is obtained by combining the initial coupling degree.

[0071] The style components are represented as style feature vectors and the structural constraints are represented as structural topology vectors. The interaction response value between the style feature vectors and the structural topology vectors is calculated using a bilinear mapping matrix to obtain the initial compatibility. The structural topology vectors are decomposed into a graph Laplace decomposition to obtain a set of structural eigenmodes. The projection coefficients of the style feature vectors on the set of structural eigenmodes are calculated. The compatibility relationship is calculated based on the projection coefficients and the initial compatibility.

[0072] The coupling relationship and the compatibility relationship are converted into a joint constraint vector. The gradient of the joint constraint vector is calculated to obtain the guiding direction vector. The product domain representation is projected along the guiding direction vector to obtain an intermediate representation. The intermediate representation is corrected by manifold constraints and the residual is calculated. The correction is repeated iteratively until the residual converges to obtain the adaptive representation.

[0073] To complete manifold transformations and obtain morphological and style components in the cross-modal mapping space, a precise correlation mechanism between these components and product constraints needs to be established. Morphological components typically contain spatial structural information such as the geometric outline, proportional relationships, and local details of the IP prototype. These components are mapped to morphological feature vectors, the dimensions of which are determined by the complexity of the product category. For apparel, these vectors can be set to 128 or 256 dimensions, while for household goods, they can be set to 64 or 128 dimensions. Functional constraints reflect the performance requirements that the target product must meet in actual use. For example, clothing needs to consider wearing comfort, freedom of movement, warmth, and breathability; cups need to consider capacity, grip stability, and heat insulation. These functional requirements are transformed into a functional requirement vector, where each component corresponds to a specific functional indicator.

[0074] The initial coupling degree is calculated using the cosine similarity method, assuming the morphological feature vector is... The functional requirement vector is The initial coupling degree The initial coupling degree is calculated by dividing the dot product of the two vectors by the product of their respective magnitudes. It reflects the degree of matching between the morphological components and functional requirements in the overall direction, ranging from 0 to 1. A value closer to 1 indicates a higher initial matching degree between form and function. However, the initial coupling degree only considers global similarity and fails to reflect the mandatory characteristics of certain functional requirements. Orthogonal decomposition of the functional requirement vector can identify rigid requirement components. Rigid requirement components are the hard indicators that a product must meet, such as the sealing performance of a water cup or the basic size specifications of clothing. The orthogonal decomposition uses the Schmidt orthogonalization method, identifying the component with the highest weight in the functional requirement vector as the first principal axis, then gradually constructing an orthogonal basis, and finally extracting the rigid requirement components. .

[0075] The projection length of the morphological feature vector onto the rigidity requirement component reflects the degree to which the form satisfies the mandatory function. The coupling relationship is obtained through vector dot product calculation. The projection length is combined with the initial coupling degree, and a weighted fusion method is used to solve for the final coupling relationship. The fusion weights are adjusted based on the characteristics of the product category. For product categories with strict functional requirements, the weight of the projection length is increased, while for decorative products, the weight of the initial coupling degree is relatively increased. The higher the coupling value, the better the morphological components can meet the functional requirements of the product.

[0076] The style component carries non-geometric information such as the artistic expression, color tendency, and texture features of the IP prototype, and is represented as a style feature vector. Structural constraints define the topological properties of a target product, including its composition, component connections, and material distribution. For apparel, these include neckline design, sleeve structure, and opening / closing mechanisms; for household goods, they include handle placement, cover fit, and base support structure. This structural information is encoded into structural topology vectors. .

[0077] Bilinear mapping matrix It can establish an interaction channel between style features and structural topology. The number of rows in the bilinear mapping matrix is ​​the same as the dimension of the style feature vector, and the number of columns is the same as the dimension of the structural topology vector. The interaction response value is obtained by left-multiplying the style feature vector by the bilinear mapping matrix and then right-multiplying it by the transpose of the structural topology vector, and is used as the initial compatibility. The parameters of the bilinear mapping matrix were obtained by training with data from historical successful cases. The training samples included commercially available IP derivatives and their corresponding style and structure annotations.

[0078] The graph Laplace decomposition of structural topological vectors can reveal the intrinsic vibration modes of a structure. Treating structural constraints as a graph structure, where nodes represent functional components of a product and edges represent connections between components, the Laplace matrix of this graph structure is constructed and its eigenvalues ​​are decomposed. The resulting eigenvectors constitute the set of structural intrinsic modes. Low-frequency intrinsic modes correspond to the macroscopic morphology of the overall structure, while high-frequency intrinsic modes correspond to the microscopic changes in local details. The style feature vector is projected onto each structural intrinsic mode, and the projection coefficients... It reflects the distribution of style across different structural scales.

[0079] Compatibility The calculation comprehensively considers the initial compatibility and the distribution characteristics of the projection coefficients, employing a weighted summation method. The weight coefficients are determined based on the importance of each intrinsic mode, with low-frequency modes typically having higher weights. When the projection coefficient of a certain intrinsic mode is too large or too small, it indicates a mismatch between style and structure at the current scale, requiring a reduction in the overall compatibility score. The compatibility score also ranges from 0 to 1, with higher values ​​indicating a better match between style components and structural constraints.

[0080] After obtaining the coupling and compatibility relationships, convert them into a joint constraint vector. The transformation process requires extending the scalar relational measure into a vector form. This involves scaling the morphological component based on the coupling relationship and rotating or translating the style component based on the compatibility relationship. The processed morphological and style components are then concatenated to form a joint constraint vector. This joint constraint vector indicates the direction and magnitude of adjustment required for the product domain representation.

[0081] Calculate the gradient of the joint constraint vector The gradient calculation employs a numerical differentiation method, which involves applying small perturbations to each dimension of the joint constraint vector and observing the response changes. The resulting gradient vector is the guiding direction vector. .

[0082] In this embodiment, by vectorizing morphological components and functional constraints respectively and constructing coupling relationships based on cosine similarity and orthogonal decomposition, a fine characterization of the matching degree between rigid components and morphological expressions in functional requirements is achieved. This effectively distinguishes between key functional constraints and non-key factors, thereby improving the accuracy and effectiveness of function-morphology matching. By introducing bilinear mapping and graph Laplace decomposition mechanisms into style components and structural constraints, a multi-dimensional interaction relationship between style features and structural topology is constructed. Furthermore, by using structural intrinsic modes to perform projection analysis on style features, the inherent coupling law between style and structure can be more deeply characterized, significantly enhancing the compatibility assessment capability between style and structure, and improving the structural rationality and style consistency of the design results. By mapping coupling and compatibility relationships into a unified joint constraint vector and constructing a guiding direction vector based on gradient information, directional projection and iterative optimization of product domain representation are performed, realizing the feature adjustment process under the collaborative guidance of multiple constraints, and improving the convergence efficiency and directional accuracy of representation optimization.

[0083] In one alternative implementation,

[0084] The adaptation representation is projected onto the product form space and geometrically transformed based on preset form topology constraints to generate a candidate form set. The feature consistency metric between each candidate form in the candidate form set and the IP prototype is calculated, including:

[0085] Obtain a parameterized template of the product form space corresponding to the target product category, map the adaptation representation to the control parameters of the parameterized template to obtain an initial parameter configuration, instantiate the parameterized template based on the initial parameter configuration to obtain an initial product form, perform three-dimensional meshing on the initial product form and extract the mesh vertex coordinates;

[0086] The position and size constraints of key components are extracted from the preset morphological topology constraints. Based on the position and size constraints, the coordinates of the mesh vertices are subjected to local deformation operations to obtain multiple product forms. The geometric feasibility of the multiple product forms is tested and the product forms that do not meet the preset manufacturing requirements are filtered to obtain the candidate form set.

[0087] Contour extraction is performed on the IP prototype and each candidate form in the candidate form set. Shape moments are calculated based on the extracted contours. The Euclidean distance between the shape moments of the candidate forms and the shape moments of the IP prototype is calculated and normalized to obtain the shape difference degree. Visual element features are extracted from the IP prototype and the candidate forms and feature matching is performed to obtain the element retention rate. The feature consistency metric is obtained by inverse weighting the shape difference degree and the element retention rate.

[0088] Transforming the adaptation representation into a concrete product form requires establishing a mapping mechanism from the abstract representation space to the concrete geometric space. For the target product category, a parametric template containing typical geometric features of that category is pre-constructed. This template defines the product's basic form and structure through a set of control parameters. For example, in clothing, the parametric template includes key control parameters such as collar curvature, sleeve length ratio, and waist tightness; for everyday items like water cups, it includes parameters such as cup height, mouth size, and bottom curvature. When mapping the adaptation representation to these control parameters, a combination of linear projection and nonlinear correction is used to ensure that the morphological components in the adaptation representation effectively drive the values ​​of the template parameters. The mapping process is achieved by establishing a correspondence between the dimensions of the adaptation representation and the template parameters. For dimensions with larger values ​​in the adaptation representation, the corresponding template parameters will receive a more significant adjustment. After obtaining the initial parameter configuration through the mapping mechanism, these parameter values ​​are substituted into the generation function of the parametric template to instantiate the geometric representation of the initial product form.

[0089] To facilitate subsequent geometric transformations, the initial product form is converted into a discrete 3D mesh representation. The 3D meshing process employs an adaptive subdivision strategy: increasing mesh density in regions where curvature changes exceed a preset threshold to ensure accurate representation of geometric details, and using a sparser mesh in flat regions where curvature changes are less than the preset threshold to improve computational efficiency. After meshing, the 3D coordinates of all mesh vertices are extracted; these coordinates form the basis for subsequent local deformation operations. Each mesh vertex not only records its spatial position but also associates it with the topological connections of adjacent vertices, providing constraints for maintaining mesh continuity and smoothness.

[0090] Morphological topology constraints define the spatial layout rules and size limitations of key product components. Taking a T-shirt derivative design as an example, morphological topology constraints stipulate that the neckline center must be located on the top center axis of the garment, the sleeve connection must be aligned with the shoulder line, and the hem width must not exceed a specific multiple of the shoulder width, among other rules. When extracting positional constraints related to the target product category from a pre-set morphological topology constraint library, the spatial coordinate range and relative positional relationship of key components are obtained; when extracting size constraints, the allowable ranges of dimensional parameters such as length, width, and thickness of each component are obtained. When performing local deformation operations on the mesh vertex coordinates based on the aforementioned constraints, the Laplace deformation algorithm is used. The Laplace deformation algorithm maintains the relative stability of local geometric features while adjusting vertex positions. Specifically, key vertices that need to be adjusted are selected as control points, and these control points are moved to the target positions according to the positional constraints. The new positions of the remaining vertices are determined by solving an energy minimization problem, ensuring a smooth transition of the overall mesh. Size constraints are introduced as an additional optimization objective during the deformation process. When the size of a component approaches the constraint boundary, a corresponding penalty term is added to avoid generating unreasonable shapes.

[0091] By adjusting different combinations of control points and deformation parameters, multiple variant forms are derived from the initial product form. To ensure the feasibility of these forms in actual manufacturing, a geometric feasibility test is performed on each product form. The test process includes self-intersection checks, i.e., verifying whether different parts of the mesh surface penetrate each other; thickness checks, ensuring that the wall thickness of each part of the product meets the minimum requirements for material processing; and connectivity checks, verifying whether the product structure forms a complete closed or semi-closed geometry. For example, for clothing products, it is also necessary to check whether the seam positions are reasonable to avoid geometric configurations that cannot be sewn together. For container products such as cups, it is necessary to verify whether an effective containing space is formed inside. Product forms that do not meet these manufacturing requirements are filtered out, and the forms that pass all tests constitute a candidate form set.

[0092] To quantitatively evaluate the similarity between candidate forms and the IP prototype, feature consistency metrics need to be calculated from multiple dimensions. Contour extraction is performed on the IP prototype and each candidate form. During contour extraction, the 3D model is projected from multiple standard perspectives to obtain 2D contour lines for the front, side, and top views. Shape moments are calculated based on the extracted contours. Shape moments are a set of values ​​describing the geometric features of a shape, including the zeroth-order moment representing the contour area, the first-order moment representing the centroid location, and the second-order and higher-order moments representing the shape's distribution characteristics and orientation. When calculating the Euclidean distance between the shape moments of the candidate form and the IP prototype, the square root of the sum of the squares of the differences between each order of moments is used to obtain the distance value. To eliminate the influence of scale differences, the distance value is normalized, mapping it to the interval between zero and one. The normalized value is the shape difference degree; the smaller the value, the more similar the shapes.

[0093] Besides overall shape features, IP prototypes typically contain specific visual elements, such as the shape of a cartoon character's eyes, specific pattern textures, and iconic color schemes. When extracting these visual element features from IP prototypes and candidate forms, a deep learning-based feature extraction network is used to identify and encode local visual patterns. The feature matching process is achieved by calculating the cosine similarity between feature vectors. The number of successfully matched IP prototype visual elements in the candidate form is counted, and divided by the total number of visual elements contained in the IP prototype to obtain the element retention rate. The element retention rate reflects the extent to which the candidate form retains the iconic visual features of the IP prototype.

[0094] The feature consistency metric is calculated by inversely weighting shape difference and element retention rate. Inverse weighting involves negating the shape difference, as a smaller difference indicates higher consistency; therefore, 1 minus the shape difference is used as the shape consistency component. The shape consistency component and element retention rate are then summed using preset weighting coefficients. These coefficients are determined based on the specific application scenario: higher weight is given to shape consistency in scenarios emphasizing visual similarity, and higher weight is given to element retention rate in scenarios emphasizing IP element inheritance. The weighted sum is the feature consistency metric, which comprehensively reflects the degree of matching between the candidate form and the IP prototype at both the shape and visual element levels, providing a quantitative evaluation basis for subsequent selection of the optimal derivative form.

[0095] In this embodiment, by constructing a parameterized template corresponding to the target product category and mapping the adaptation representation to adjustable parameter configuration, a direct transformation from abstract feature space to specific product form is achieved, significantly improving the automation and efficiency of product form generation. At the same time, it enhances the consistency between the design result and the upper-level semantic representation. By performing three-dimensional meshing on the initial product form and combining positional and dimensional constraints to locally deform the mesh vertices, multiple candidate form schemes are generated. This expands the design space while ensuring that design constraints are met, effectively improving the diversity of schemes and design exploration capabilities. By extracting the contours of the IP prototype and candidate forms and calculating the shape difference based on shape moments, a quantitative evaluation of the overall form similarity is achieved, which can more objectively and stably measure form consistency, thereby improving the accuracy and reliability of form matching evaluation.

[0096] In one alternative implementation,

[0097] Based on the feature consistency metric, the candidate morphology set is filtered to obtain the following derived morphologies:

[0098] For each candidate form in the candidate form set, determine the corresponding feature consistency metric and calculate the statistical distribution parameter. Based on the statistical distribution parameter, construct an adaptive screening threshold and screen the candidate form set to obtain a subset of highly consistent candidate forms.

[0099] The manufacturing-related parameters and cost-related parameters corresponding to each candidate form in the high-consistency candidate form subset are extracted. The manufacturing constraint satisfaction is calculated based on the manufacturing-related parameters and preset manufacturing constraints. The cost constraint satisfaction is calculated based on the cost-related parameters and preset cost constraints. The cost constraint satisfaction and the manufacturing constraint satisfaction are optimized through Pareto front analysis to obtain a feasibility score. The feature consistency metric and the feasibility score are jointly evaluated to obtain a comprehensive evaluation score.

[0100] Feature vectors are extracted from the candidate morphologies in the high consistency candidate morphology subset, and a feature matrix is ​​constructed. The Hausdorff distance between any two candidate morphologies in the feature matrix is ​​calculated to obtain a morphological difference matrix. Based on the morphological difference matrix, the difference index between candidate morphologies is calculated and combined with the comprehensive evaluation score to obtain a dual-criteria ranking result. The derived morphology is extracted based on the dual-criteria ranking result.

[0101] Selecting the final derived form from the candidate form set requires comprehensive consideration of multiple dimensions, including IP feature retention, manufacturing feasibility, and form diversity. For each candidate form in the set, a corresponding feature consistency metric was obtained in the aforementioned example. This metric reflects the degree of similarity between the candidate form and the IP prototype in terms of visual, stylistic, and semantic features. To avoid selection bias caused by using a fixed threshold, an adaptive selection threshold needs to be determined based on the overall distribution characteristics of the candidate form set.

[0102] The feature consistency metrics of all candidate forms in the candidate form set are compiled into a numerical sequence. Statistical analysis is performed on this sequence, calculating its mean and standard deviation as statistical distribution parameters. An adaptive screening threshold is constructed based on these statistical distribution parameters. This threshold can typically be set as the sum of the mean and standard deviation, or the mean plus the standard deviation multiplied by an adjustment coefficient, which can be adjusted according to the required stringency of the screening. When the feature consistency metric of a candidate form is greater than or equal to the aforementioned adaptive screening threshold, that candidate form is retained in the high-consistency candidate form subset; otherwise, it is discarded.

[0103] Candidate forms that have entered the high-consistency candidate form subset have met the basic requirements in terms of IP feature preservation. The next step is to evaluate their feasibility in actual production and manufacturing. For each candidate form, parameters related to the manufacturing process are extracted. These manufacturing-related parameters include, but are not limited to, geometric complexity, minimum feature size, maximum span, wall thickness distribution, draft angle, and number of undercut structures. Simultaneously, cost-related parameters are extracted, including material usage, number of processing steps, mold complexity, and estimated processing time.

[0104] Pre-defined manufacturing constraints are set based on the target product category. For example, for injection-molded products, manufacturing constraints might include minimum wall thickness not less than a certain threshold, draft angle within a certain range, and avoidance of undercut structures that are difficult to demold. The manufacturing-related parameters of each candidate form are compared item by item with these manufacturing constraints to calculate the degree to which they are satisfied. For each constraint, a normalized satisfaction function can be used for quantification. A satisfaction score of 1 indicates that the parameter fully satisfies the constraint; a satisfaction score close to 0 indicates that the parameter severely violates the constraint; and a median value is calculated based on the degree of deviation if the parameter falls between these two values. The satisfaction scores of each constraint are then weighted and averaged or multiplied to obtain the manufacturing constraint satisfaction score of the candidate form.

[0105] Similarly, based on preset cost constraints, such as material costs not exceeding a certain upper limit and processing time not exceeding a certain threshold, the cost-related parameters of candidate forms are compared with these cost constraints to calculate the cost constraint satisfaction. The cost constraint satisfaction also adopts a normalized quantification method; the satisfaction is higher when the cost parameters are better than the constraint requirements, and lower when they exceed the constraint requirements.

[0106] Manufacturing constraint satisfaction and cost constraint satisfaction constitute two mutually restrictive optimization objectives. Generally, improving manufacturing constraint satisfaction may lead to a more conservative design, increasing material usage or processing complexity, thereby reducing cost constraint satisfaction; conversely, pursuing cost optimization may cause the design to approach manufacturing limits, thus reducing manufacturing constraint satisfaction. Therefore, a multi-objective optimization method is needed to balance these trade-offs. Using Pareto front analysis, in a two-dimensional space with manufacturing and cost constraint satisfaction as coordinate axes, a set of non-dominated solutions not strictly dominated by any other solution is identified; these non-dominated solutions constitute the Pareto front. Candidate forms located on the Pareto front achieve an optimal balance between manufacturing and cost constraint satisfaction, making it impossible to improve one objective without sacrificing the other. Feasibility scores are calculated based on the candidate form's position on or distance from the Pareto front. Typically, candidate forms located on the Pareto front receive higher feasibility scores, while those far from the Pareto front receive lower feasibility scores.

[0107] After obtaining the feature consistency metric and feasibility score for each candidate form, these two evaluation dimensions need to be jointly assessed. A weighted summation method can be used, setting weight coefficients according to the importance attached to IP feature retention and manufacturing feasibility, and calculating the comprehensive evaluation score. The weight coefficients can be adjusted according to the specific application scenario. For example, for scenarios emphasizing IP image fidelity, the weight of the feature consistency metric can be increased, and for scenarios focusing on cost control, the weight of the feasibility score can be increased.

[0108] Based on the comprehensive evaluation score of each candidate form in the highly consistent candidate form subset, directly sorting by the comprehensive evaluation score and selecting the top few candidate forms as the final derived forms may result in highly similar selected forms and a lack of design diversity. In order to provide more diverse design choices while ensuring quality, it is necessary to introduce a form difference index as an additional screening criterion.

[0109] For each candidate morphology in the high-consistency candidate morphology subset, the corresponding feature vector is extracted. This feature vector can be a geometric feature descriptor of the morphology, such as a shape description based on grid vertex coordinates, geometric features based on curvature distribution, or a shape context descriptor based on contour. The feature vectors of all candidate morphologies are arranged in rows to form a feature matrix.

[0110] The Hausdorff distance between any two candidate morphologies in the feature matrix is ​​calculated. The Hausdorff distance measures the maximum mismatch between two point sets. For two distinct candidate morphologies, the corresponding Hausdorff distance reflects the degree of difference in their geometric shapes; a larger distance indicates a more significant morphological difference. The pairwise Hausdorff distances between all candidate morphologies are organized into a morphological dissimilarity matrix, which is a symmetric matrix with zero diagonal elements. Based on the morphological dissimilarity matrix, the average dissimilarity between each candidate morphology and other candidate morphologies is calculated as an indicator of the candidate morphology's dissimilarity.

[0111] The difference index between candidate morphologies is calculated based on the morphological difference matrix. For each candidate morphology, the average Hausdorff distance between it and all other candidate morphologies is calculated as the difference index for that candidate morphology. The higher the difference index value, the more unique the candidate morphology is within the entire subset of highly consistent candidate morphologies, and the more significant its difference from other morphologies. To reduce the impact of extreme values ​​on the calculation of the average difference index, a pruning averaging method can be used, i.e., removing several candidate morphologies that are farthest and closest to the candidate morphology before calculating the average difference index.

[0112] The comprehensive evaluation score and the difference index of each candidate form are used as two evaluation dimensions to jointly rank them, resulting in a dual-criteria ranking. The joint ranking can employ either a weighted ranking method or a non-dominated ranking method. The weighted ranking method combines the two indicators into a single ranking index according to preset weights, with the weight settings reflecting the degree of preference for quality and diversity. The non-dominated ranking method, based on Pareto dominance, divides candidate forms into multiple levels of frontiers. The first level frontier contains solutions not dominated by any other candidate forms, the second level frontier contains solutions dominated only by solutions in the first level frontier, and so on. Within the same level frontier, further ranking can be achieved by calculating the crowding distance; candidate forms with larger crowding distances have higher priority.

[0113] Derivative patterns are extracted based on the ranking results using a dual-criteria approach. Two strategies can be employed: one is to directly select a fixed number of top-ranked candidate patterns as derived patterns; the other is to use a greedy selection strategy, first selecting the highest-ranking candidate pattern, and then considering not only its ranking but also its difference from the already selected pattern set when selecting the next one. A comprehensive index is defined to balance ranking and difference. The comprehensive index can be a weighted combination of the ranking score and the minimum distance from the candidate pattern to the already selected pattern set. The resulting derived patterns are then obtained.

[0114] In this embodiment, by statistically modeling the feature consistency metric of candidate forms and constructing an adaptive screening threshold to achieve dynamic screening, the screening criteria can be adaptively adjusted according to the distribution characteristics of the current candidate set, thereby improving the accuracy and robustness of high-quality form recognition and reducing the problem of missing excellent solutions or misselecting inferior solutions. By introducing manufacturing-related parameters and cost-related parameters, and combining manufacturing constraints and cost constraints for multi-objective modeling, Pareto front analysis is used to achieve comprehensive optimization, which can achieve a balance between performance and cost under multiple constraints, significantly improving the overall performance of the design results in terms of engineering feasibility and economic feasibility. By jointly evaluating the feature consistency metric and the feasibility score obtained from multi-objective optimization, a unified comprehensive evaluation score is formed, realizing the collaborative evaluation of design expression and actual implementation capability, thereby improving the comprehensiveness and rationality of the final solution selection. By constructing a form difference matrix between candidate forms and characterizing the differences between forms based on Hausdorff distance, a dual-criteria ranking is adopted using difference index and comprehensive evaluation score, which can ensure high quality while taking into account the diversity between solutions, effectively avoiding the problem of homogenization of results, and improving the richness and innovation of the final derived form set.

[0115] In one alternative implementation,

[0116] An initial constraint diagram is constructed based on the geometric structure of the derived form, and physical attribute nodes and operation sequence nodes corresponding to the target product category are added to obtain a manufacturing constraint diagram. Constraint propagation calculations are performed on the manufacturing constraint diagram to obtain the manufacturability evaluation value corresponding to the derived form, including:

[0117] Extract the surface topology information and boundary feature information corresponding to the derived morphology, identify the surface connection relationship based on the surface topology information and construct a topology connection matrix, identify feature boundaries based on the boundary feature information and extract boundary constraints, perform a structured expression of the topology connection matrix and the boundary constraints and construct an initial constraint graph, retrieve physical attribute requirements from the manufacturing knowledge base corresponding to the target product category and generate physical attribute nodes, retrieve the processing operation sequence from the process knowledge base of the target product category and generate operation sequence nodes, insert the physical attribute nodes and the operation sequence nodes into the initial constraint graph and establish dependencies to obtain the manufacturing constraint graph;

[0118] The nodes in the manufacturing constraint graph are labeled with types and a node type mapping table is constructed. All nodes in the manufacturing constraint graph are traversed and the constraint propagation path is determined. Iterative constraint propagation calculation is performed along the constraint propagation path and constraint conflicts are detected. Conflicting nodes are recorded and constraint relaxation is performed. This process is repeated until the states corresponding to all nodes converge. The degree of constraint satisfaction is statistically analyzed and the manufacturability evaluation value corresponding to the derived form is calculated.

[0119] For the generated derivative forms, it is necessary to verify their manufacturability to ensure that they can be physically produced through actual processing techniques. The core of manufacturability assessment lies in transforming the geometric information of the derivative form into a structured expression with manufacturing constraints, and detecting potential process conflicts through constraint propagation mechanisms.

[0120] The process involves acquiring 3D geometric data of the derived form and extracting surface topology information through mesh topology analysis. Surface topology information includes topological invariants such as the connectivity of mesh vertices, the adjacency of facets, and topological genus. Specifically, for a derived form represented by a triangular mesh, all triangular facets are traversed, and the shared edge information between each facet and its adjacent facets is recorded. By analyzing the number and distribution pattern of shared edges, the connectivity relationships of the surfaces are identified. For clothing products, surface connectivity reflects the stitching relationships between different fabric segments; for household goods, surface connectivity reflects the assembly relationships between different components. The identified surface connectivity relationships are represented as a topological connectivity matrix. The rows and columns of the topological connectivity matrix correspond to different surface regions, and the values ​​of the matrix elements represent the connection type between two surface regions, which can be stitched, glued, or mechanically fastened.

[0121] Boundary feature information extraction is achieved through a geometric feature recognition algorithm. All edges of the derived form are traversed, and the curvature and torsion parameters of each edge are calculated. When the rate of change of edge curvature exceeds a preset threshold, the edge is marked as a feature boundary. Feature boundaries include sharp edges, crease edges, and material boundary edges. For each feature boundary, its corresponding boundary constraints are extracted. Boundary constraints describe the geometric requirements that the feature boundary must meet during processing, including edge straightness requirements, corner radius requirements, and edge reinforcement requirements. For apparel products, boundary constraints also include requirements for seam width and stitch density; for injection-molded household goods, boundary constraints include requirements for demolding angle and wall thickness uniformity.

[0122] When structurally representing the topological connectivity matrix and boundary constraints, a graph structure is used for unified modeling. An initial constraint graph is created, with nodes categorized into geometric nodes and constraint nodes. Geometric nodes correspond to the surface regions and feature boundaries in the derived form, while constraint nodes correspond to the topological connectivity relationships and boundary constraints. Directed edges are established between geometric nodes and constraint nodes, with the direction of the directed edges indicating the direction of constraint application. Each non-zero element in the topological connectivity matrix corresponds to a connection constraint node in the initial constraint graph, which is connected to two related surface geometric nodes via directed edges. Each constraint term in the boundary constraints corresponds to a boundary constraint node in the initial constraint graph, which is connected to the corresponding feature boundary geometric node.

[0123] The physical property requirements are retrieved from the manufacturing knowledge base corresponding to the target product category. The manufacturing knowledge base is stored in a structured database format, containing physical performance indicators for different product categories under different usage scenarios. For example, physical property requirements include material strength requirements, abrasion resistance requirements, flexibility requirements, and thermal stability requirements. Based on the retrieved physical property requirements, physical property nodes are generated. Each physical property node contains an attribute type identifier, an attribute value range, and a testing method identifier. Physical property nodes are stored in key-value pair format, with the attribute type identifier as the key and the attribute value range and testing method identifier forming the value.

[0124] The processing operation sequence is retrieved from the process knowledge base of the target product category. The process knowledge base records the standard processing flow for a specific product category, including cutting, sewing, forming, and assembly steps and their order. The processing operation sequence is represented as a directed acyclic graph, where nodes correspond to individual processing operations, and directed edges represent the dependencies between operations. Based on the retrieved processing operation sequence, operation sequence nodes are generated. Each operation sequence node contains an operation type identifier, a set of operation parameters, and a list of prerequisite operations. The set of operation parameters describes the specific execution parameters of the operation; for sewing operations, parameters include stitch length, thread tension, and stitch type; for injection molding operations, parameters include injection temperature, pressure, and holding time.

[0125] When inserting physical property nodes and operation sequence nodes into the initial constraint graph, it is necessary to establish the dependencies between these new nodes and the existing geometric nodes. The dependencies between physical property nodes and surface geometry nodes are established through material-geometry mapping. All surface geometry nodes are traversed, and the physical property requirements that the surface region needs to satisfy are determined based on the surface's geometric characteristics (such as thickness and curvature). Directed edges are then established between the corresponding physical property nodes and surface geometry nodes. The dependencies between operation sequence nodes and geometric nodes are established through process-geometry mapping. For each operation sequence node, directed edges are established between the operation sequence node and the relevant geometric nodes based on the geometric region affected by the operation. After completing node insertion and dependency establishment, the manufacturing constraint graph is obtained.

[0126] Nodes in the manufacturing constraint graph are labeled with their types, and a node type mapping table is established. Node types include geometric nodes, topological constraint nodes, boundary constraint nodes, physical attribute nodes, and operation sequence nodes. The node type mapping table records the unique identifier and corresponding type label of each node, providing a type query interface for subsequent constraint propagation calculations.

[0127] Traverse all nodes in the manufacturing constraint graph to determine the constraint propagation path. The determination of the constraint propagation path is based on a depth-first search algorithm for the graph. Start the search from the node with an in-degree of zero, and visit subsequent nodes sequentially along the directed edges, recording the visiting order to form the constraint propagation path. The constraint propagation path describes the flow direction of constraint information in the manufacturing constraint graph, ensuring that constraint information propagates gradually from the source node to all relevant nodes.

[0128] Iterative constraint propagation computation is performed along the constraint propagation path. The goal of constraint propagation computation is to pass the constraints of each constraint node to the relevant geometric nodes and check whether the current state of the geometric nodes satisfies all propagated constraints. For each geometric node, a constraint set is maintained, which records all constraints acting on that node. The constraint propagation path is traversed, and for each constraint node on the path, its constraints are added to the constraint set of the downstream geometric nodes.

[0129] Constraint conflicts are detected during constraint propagation. A constraint conflict refers to the existence of contradictory constraints within the constraint set of a geometric node. For example, a boundary node receiving two constraints simultaneously—one with a minimum fillet radius greater than 2 mm and the other with a maximum fillet radius less than 1.5 mm—constitutes a constraint conflict. Constraint conflict detection employs a constraint compatibility check algorithm, which compares constraints of the same type and calculates their intersection. If the intersection is empty, a constraint conflict is determined to exist.

[0130] When a constraint conflict is detected, the conflicting node and related constraint conditions are recorded, and constraint relaxation is performed. Constraint relaxation is based on constraint priority. The determination of constraint priority considers the correlation between the constraint type and product functional requirements; constraints that directly affect product functionality have higher priority, while constraints that only affect aesthetics have lower priority. Constraint relaxation includes expanding the range of numerical constraints, reducing geometric accuracy requirements, or allowing additional machining processes to meet strict constraints.

[0131] Constraint propagation computation is an iterative process. After each iteration, the state changes of all nodes are checked. When the change in a node's state over two consecutive iterations is less than a preset threshold, the node's state is considered to have converged. The constraint propagation computation is complete when the states of all nodes have converged. After the computation, the degree of constraint satisfaction is statistically analyzed, including the proportion of nodes that satisfy the constraints, the severity of constraint violations, and the proportion of nodes requiring constraint relaxation.

[0132] The manufacturability assessment value for derived forms is calculated based on the statistical results of constraint satisfaction. The manufacturability assessment value uses a weighted composite index, considering three aspects: the constraint satisfaction rate, the severity of constraint violation, and the degree of constraint slack. A higher constraint satisfaction rate, lower constraint violation severity, and smaller constraint slack result in a higher manufacturability assessment value. The manufacturability assessment value is typically normalized to the range of 0 to 1; a value closer to 1 indicates better manufacturability.

[0133] In this embodiment, by systematically extracting the surface topology and boundary feature information of the derived morphology and constructing a unified expression of the topology connection matrix and boundary constraints, explicit modeling of the product's geometric structure and key boundary characteristics is achieved. This significantly improves the completeness and computability of the structural information expression, providing a reliable foundation for subsequent manufacturing constraint analysis. By integrating the physical attribute requirements in the manufacturing knowledge base and the processing operation sequence in the process knowledge base, multi-source manufacturing knowledge is embedded into the constraint graph and dependencies are established. This achieves cross-domain correlation modeling from geometric design to the manufacturing process, effectively enhancing the consistency between the design results and actual manufacturing conditions and reducing the risk of design non-manufacturability. By marking the node types of the manufacturing constraint graph and constructing constraint propagation paths, combined with iterative constraint propagation and conflict detection mechanisms, global consistency verification under multiple constraint conditions is achieved. This enables more comprehensive identification of potential conflicts and dynamic correction, thereby improving the accuracy and stability of constraint processing. By introducing constraint relaxation and iterative convergence mechanisms, conflicts are reasonably adjusted while ensuring that key constraints are met. Compared with rigid constraint processing methods, this improves the solution capability and robustness in complex constraint environments and avoids the problem of overall infeasibility due to local conflicts.

[0134] Figure 2This is a flowchart illustrating the manufacturability assessment of the derivative forms of the IP smart derivative method for clothing and daily necessities according to an embodiment of the present invention.

[0135] In one alternative implementation,

[0136] Based on the manufacturability assessment value, conflict regions are identified, and the geometric parameters of the conflict regions are constrained to obtain a physical implementation scheme through solution, including:

[0137] Local gradient analysis is performed on the manufacturability assessment value, and the derived morphological regions where the gradient change rate exceeds a preset change rate threshold are marked as candidate conflict regions. The constraint conflict type and conflict intensity corresponding to the candidate conflict regions are extracted. Based on the constraint conflict type, the candidate conflict regions are divided into material conflict regions and process conflict regions. Based on the conflict intensity, the conflict degree of the material conflict regions and the process conflict regions is quantitatively evaluated. Adjacent conflict regions of the same type are merged, and the region boundary is determined based on the conflict degree quantitative evaluation result to obtain the conflict region.

[0138] Extract the geometric parameters corresponding to the conflict region, perform parameter space mapping, and construct a parameter constraint network. Mark hard constraint edges and soft constraint edges in the parameter constraint network. Construct a constraint propagation tree based on the hard constraint edges and perform layer-by-layer constraint solving to obtain the hard constraint solution space. Construct an objective function based on the soft constraint edges and solve it to obtain the soft constraint solution space. Calculate the intersection of the hard constraint solution space and the soft constraint solution space and filter parameter configurations that satisfy constraint compatibility. Adjust the parameters of the geometry of the conflict region based on the parameter configuration to obtain corrected geometric data. Determine the processing sequence and assembly sequence based on the corrected geometric data and associate them to obtain a physical implementation scheme.

[0139] After obtaining the manufacturability assessment values ​​on the manufacturing constraint map, fine-grained spatial analysis is performed on these values ​​to identify potential manufacturing conflicts. Local gradient calculations are performed on the distribution of manufacturability assessment values ​​on the derived morphological surface. A multi-scale gradient operator is used to differentiate the assessment value field, obtaining the gradient vector and its rate of change for each spatial location. When the gradient rate of change in a certain region exceeds a preset rate of change threshold, it indicates a drastic fluctuation in manufacturability in that region, and this region is marked as a candidate conflict region. The preset rate of change threshold is determined based on the manufacturing precision requirements of the target product category; it can be set to 0.15 for precision clothing accessories and 0.25 for general consumer goods.

[0140] For the identified candidate conflict areas, the specific constraint types leading to the decrease in evaluation values ​​are extracted by tracing the constraint propagation path in the manufacturing constraint diagram. Constraint conflict types include various situations such as material performance mismatch, processing limitations, and assembly interference. Conflict intensity is calculated as a quantitative indicator, obtained by weighted accumulation of the degree of constraint violation within the region, with the weight determined by the priority of each constraint in the manufacturing process. Based on the dominant factors of constraint conflict types, candidate conflict areas are divided into two main categories: material conflict areas and process conflict areas. Material conflict areas mainly involve issues such as excessive material thickness, insufficient bending radius, and stress concentration, while process conflict areas are associated with obstacles at the manufacturing execution level, such as unreachable cutting paths, abnormal stitching angles, and insufficient assembly clearances.

[0141] The degree of conflict in material conflict areas and process conflict areas is quantitatively assessed separately. The assessment of material conflict areas is based on the deviation between the material's mechanical properties and geometric parameters, calculating a matching index between the geometric features of the conflict area and the material's bearing capacity; a lower index indicates a more severe conflict. The assessment of process conflict areas is based on the processing equipment's capabilities and operational complexity, comprehensively considering factors such as tool accessibility, clamping stability, and the smoothness of process transitions to form a process feasibility score. After the quantitative assessment, spatially adjacent areas belonging to the same conflict type are merged. A region growing algorithm is used to merge adjacent areas with similar conflict characteristics into a unified conflict area. During the merging process, the region boundaries are dynamically adjusted based on the quantitative assessment results to ensure that the boundary division accurately reflects the actual impact range of the manufacturing problem, resulting in clearly defined conflict areas.

[0142] After identifying the conflict zone, all geometric parameters affecting manufacturing within the conflict zone are extracted, including surface curvature, wall thickness distribution, feature dimensions, and relative positions. These geometric parameters are then mapped from three-dimensional space to a multi-dimensional parameter space, establishing dependencies and constraints between the parameters to construct a parameter constraint network. The parameter constraint network uses parameters as nodes and constraints as edges, with each edge carrying constraint strength and type attributes. In the parameter constraint network, inviolable constraints derived from physical laws, material limits, and equipment capabilities are marked as hard constraint edges, while negotiable constraints derived from design preferences, aesthetic considerations, and cost optimization are marked as soft constraint edges. Hard constraint edges typically involve rigid requirements such as material yield strength, minimum wall thickness, and demolding angle, while soft constraint edges include flexible objectives such as desired surface finish, optimized processing time, and improved material utilization.

[0143] Based on labeled hard constraint edges, a constraint propagation algorithm is used to construct a constraint propagation tree. Starting from the core parameters of the conflict region as the root node, the algorithm expands outward according to constraint dependencies, forming a hierarchical constraint propagation structure. Layer-by-layer constraint solving is performed on the constraint propagation tree, with parameter value ranges propagated backward from the leaf nodes. The solution results of each layer serve as the input constraints for the next layer. The algorithm converges at the root node to obtain the parameter value range that satisfies all hard constraints, and this parameter value range constitutes the hard constraint solution space.

[0144] For soft-constrained edges, an optimization modeling method is used to construct the objective function. The objective function comprehensively considers the satisfaction of multiple soft constraints, and a multi-objective optimization problem is formed through weighted summation or Pareto optimization. The Lagrange multiplier method or sequential quadratic programming algorithm is introduced to solve the objective function, and the optimal equilibrium point is found under the mutual constraints of soft constraints, thus obtaining the soft-constrained solution space.

[0145] The intersection of the hard-constraint solution space and the soft-constraint solution space is calculated. This intersection represents the range of parameters that satisfy both rigid manufacturing requirements and optimization objectives. Within this intersection region, parameter configurations that meet constraint compatibility are further screened to verify the coordination of each parameter combination in the actual manufacturing scenario, eliminating configuration schemes that may cause secondary conflicts. The selected parameter configurations are then applied to adjust the geometry of the conflict region. Parametric modeling techniques are used to modify the surface shape, feature dimensions, and connection methods of this region, generating corrected geometric data that conforms to manufacturing constraints.

[0146] Based on the corrected geometric data and the manufacturing process characteristics of the target product category, a reasonable processing sequence is determined. The processing sequence needs to consider factors such as inter-process dependencies, the number of workpiece clamping changes, and tool change frequency, and an optimal processing path is generated using a process optimization algorithm. Simultaneously, for products containing multiple components, an assembly sequence is planned to ensure smooth assembly of each component. Determining the assembly sequence requires analyzing the geometric interference relationships between components, the operational space of connection methods, and the accessibility of assembly tools. The feasibility of the sequence is verified through assembly simulation. The determined processing sequence and assembly sequence are then linked and organized with the corrected geometric data to form a complete physical implementation scheme that includes a geometric model, process parameters, operating steps, and quality inspection standards. This scheme can directly guide actual production and manufacturing.

[0147] In this embodiment, by performing local gradient analysis on the manufacturability assessment value and locating gradient abrupt change regions, the potential manufacturing conflicts are accurately identified. This enables earlier and more accurate detection of local unreasonable designs, thereby improving the accuracy of problem location and the timeliness of response. By classifying conflict regions by type and quantifying them based on conflict intensity, a differentiated analysis mechanism is established between material conflicts and process conflicts. This allows for more targeted handling strategies for different conflict sources, thereby improving the refinement and effectiveness of conflict analysis. By mapping conflict regions to the parameter space and constructing a parameter constraint network, while distinguishing between hard and soft constraints and solving the solution space separately, hierarchical modeling and collaborative solving of multiple types of constraints are achieved. This significantly enhances the solution capability and solution space representation capability under complex constraints, and improves the rationality and controllability of the parameter adjustment process. By intersecting and filtering the hard constraint solution space and the soft constraint solution space, it is ensured that the parameter configuration satisfies the key constraints while taking into account the optimization objective, achieving an effective balance between constraint satisfaction and optimization performance.

[0148] A second aspect of this invention provides an IP-based intelligent derivative system for clothing and daily necessities, comprising:

[0149] The IP adaptation unit is used to acquire an IP prototype and extract the corresponding IP representation. Based on the IP representation and the functional and structural constraints corresponding to the target product category, a cross-modal mapping space is constructed. In the cross-modal mapping space, the IP representation is subjected to manifold transformation to obtain a product domain representation. Component decomposition is then performed to obtain morphological and style components. The coupling relationship between the morphological components and the functional constraints, as well as the compatibility relationship between the style components and the structural constraints, are calculated. Based on the coupling and compatibility relationships, the product domain representation is subjected to constraint-guided projection to obtain an adaptation representation.

[0150] The morphology derivation unit is used to project the adaptation representation onto the product morphology space and perform geometric transformations based on preset morphology topology constraints to generate a candidate morphology set, calculate the feature consistency metric between each candidate morphology in the candidate morphology set and the IP prototype, and filter the candidate morphology set based on the feature consistency metric to obtain a derived morphology.

[0151] The manufacturing constraint unit is used to construct an initial constraint diagram based on the geometric structure of the derived form and add physical attribute nodes and operation sequence nodes corresponding to the target product category to obtain a manufacturing constraint diagram. It performs constraint propagation calculation on the manufacturing constraint diagram to obtain the manufacturability evaluation value corresponding to the derived form. Based on the manufacturability evaluation value, it identifies conflict areas and performs constraint satisfaction calculation on the geometric parameters of the conflict areas to obtain a physical implementation scheme.

[0152] A third aspect of the present invention provides an electronic device, comprising:

[0153] A processor and a memory for storing processor-executable instructions, wherein the processor is configured to invoke instructions stored in the memory to perform the aforementioned method.

[0154] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0155] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0156] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for intelligent IP derivatives in clothing and daily necessities, characterized in that: include: Obtain an IP prototype and extract the corresponding IP representation. Construct a cross-modal mapping space based on the IP representation and the functional and structural constraints corresponding to the target product category. Perform a manifold transformation on the IP representation in the cross-modal mapping space to obtain a product domain representation. Perform component decomposition to obtain morphological and style components. Calculate the coupling relationship between the morphological components and the functional constraints, and the compatibility relationship between the style components and the structural constraints. Based on the coupling and compatibility relationships, perform constraint-guided projection on the product domain representation to obtain an adaptive representation. The adaptation representation is projected onto the product form space and geometric transformation is performed based on preset form topology constraints to generate a candidate form set. The feature consistency metric between each candidate form in the candidate form set and the IP prototype is calculated. Based on the feature consistency metric, the candidate form set is filtered to obtain a derived form. An initial constraint diagram is constructed based on the geometric structure of the derived form, and physical attribute nodes and operation sequence nodes corresponding to the target product category are added to obtain a manufacturing constraint diagram. Constraint propagation calculation is performed on the manufacturing constraint diagram to obtain the manufacturability evaluation value corresponding to the derived form. Conflict regions are identified based on the manufacturability evaluation value, and the geometric parameters of the conflict regions are constrained and satisfied to obtain a physical implementation scheme.

2. The method according to claim 1, characterized in that, Obtain the IP prototype and extract the corresponding IP representation. Based on the IP representation and the functional and structural constraints corresponding to the target product category, construct a cross-modal mapping space. Perform a manifold transformation on the IP representation in the cross-modal mapping space to obtain the product domain representation, and perform component decomposition to obtain morphological and style components, including: Multi-level feature extraction is performed on the IP prototype to obtain a visual feature set and a semantic feature set. Multimodal attention mechanism is then applied to the visual feature set and the semantic feature set respectively, and weighted aggregation is performed to obtain the IP representation. Obtain the product sample set corresponding to the target product category, perform graph structure parsing on the structural topology of each product sample in the product sample set and extract graph convolution features to determine structural constraints, and perform semantic parsing on the functional description of the product sample set to obtain the functional constraints; Based on the IP representation and the product sample set, a source domain space and a target domain space are constructed. The distribution difference metric between the source domain space and the target domain space is calculated and a domain mapping matrix is ​​constructed. Matrix multiplication and manifold regularization constraint calculations are performed on the IP representation and the domain mapping matrix to obtain a manifold embedding vector. The manifold embedding vector is then locally linearly reconstructed in the target domain space to obtain the product domain representation. Independent component analysis is performed on the product domain characterization to obtain multiple independent components. The correlation between each independent component and the structural constraints is calculated and the structurally sensitive components are labeled. The structurally sensitive components are aggregated to obtain morphological components. The matching degree between each independent component and the functional constraints is calculated and the style-sensitive components are labeled. The style-sensitive components are aggregated to obtain style components.

3. The method according to claim 1, characterized in that, Calculating the coupling relationship between the morphological components and the functional constraints, and the compatibility relationship between the style components and the structural constraints, and performing constraint-guided projection on the product domain representation based on the coupling and compatibility relationships to obtain an adapted representation includes: The morphological components are represented as morphological feature vectors, the functional constraints are represented as functional requirement vectors, and the cosine similarity between the morphological feature vectors and the functional requirement vectors is calculated to obtain the initial coupling degree. The functional requirement vectors are orthogonally decomposed to obtain rigid requirement components. The projection length between the morphological feature vectors and the rigid requirement components is calculated, and the coupling relationship is obtained by combining the initial coupling degree. The style components are represented as style feature vectors and the structural constraints are represented as structural topology vectors. The interaction response value between the style feature vectors and the structural topology vectors is calculated using a bilinear mapping matrix to obtain the initial compatibility. The structural topology vectors are decomposed into a graph Laplace decomposition to obtain a set of structural eigenmodes. The projection coefficients of the style feature vectors on the set of structural eigenmodes are calculated. The compatibility relationship is calculated based on the projection coefficients and the initial compatibility. The coupling relationship and the compatibility relationship are converted into a joint constraint vector. The gradient of the joint constraint vector is calculated to obtain the guiding direction vector. The product domain representation is projected along the guiding direction vector to obtain an intermediate representation. The intermediate representation is corrected by manifold constraints and the residual is calculated. The correction is repeated iteratively until the residual converges to obtain the adaptive representation.

4. The method according to claim 1, characterized in that, The adaptation representation is projected onto the product form space and geometrically transformed based on preset form topology constraints to generate a candidate form set. The feature consistency metric between each candidate form in the candidate form set and the IP prototype is calculated, including: Obtain a parameterized template of the product form space corresponding to the target product category, map the adaptation representation to the control parameters of the parameterized template to obtain an initial parameter configuration, instantiate the parameterized template based on the initial parameter configuration to obtain an initial product form, perform three-dimensional meshing on the initial product form and extract the mesh vertex coordinates; The position and size constraints of key components are extracted from the preset morphological topology constraints. Based on the position and size constraints, the coordinates of the mesh vertices are subjected to local deformation operations to obtain multiple product forms. The geometric feasibility of the multiple product forms is tested and the product forms that do not meet the preset manufacturing requirements are filtered to obtain the candidate form set. Contour extraction is performed on the IP prototype and each candidate form in the candidate form set. Shape moments are calculated based on the extracted contours. The Euclidean distance between the shape moments of the candidate forms and the shape moments of the IP prototype is calculated and normalized to obtain the shape difference degree. Visual element features are extracted from the IP prototype and the candidate forms and feature matching is performed to obtain the element retention rate. The feature consistency metric is obtained by inverse weighting the shape difference degree and the element retention rate.

5. The method according to claim 1, characterized in that, Based on the feature consistency metric, the candidate morphology set is filtered to obtain the following derived morphologies: For each candidate form in the candidate form set, determine the corresponding feature consistency metric and calculate the statistical distribution parameter. Based on the statistical distribution parameter, construct an adaptive screening threshold and screen the candidate form set to obtain a subset of highly consistent candidate forms. The manufacturing-related parameters and cost-related parameters corresponding to each candidate form in the high-consistency candidate form subset are extracted. The manufacturing constraint satisfaction is calculated based on the manufacturing-related parameters and preset manufacturing constraints. The cost constraint satisfaction is calculated based on the cost-related parameters and preset cost constraints. The cost constraint satisfaction and the manufacturing constraint satisfaction are optimized through Pareto front analysis to obtain a feasibility score. The feature consistency metric and the feasibility score are jointly evaluated to obtain a comprehensive evaluation score. Feature vectors are extracted from the candidate morphologies in the high consistency candidate morphology subset, and a feature matrix is ​​constructed. The Hausdorff distance between any two candidate morphologies in the feature matrix is ​​calculated to obtain a morphological difference matrix. Based on the morphological difference matrix, the difference index between candidate morphologies is calculated and combined with the comprehensive evaluation score to obtain a dual-criteria ranking result. The derived morphology is extracted based on the dual-criteria ranking result.

6. The method according to claim 1, characterized in that, An initial constraint diagram is constructed based on the geometric structure of the derived form, and physical attribute nodes and operation sequence nodes corresponding to the target product category are added to obtain a manufacturing constraint diagram. Constraint propagation calculations are performed on the manufacturing constraint diagram to obtain the manufacturability evaluation value corresponding to the derived form, including: Extract the surface topology information and boundary feature information corresponding to the derived morphology, identify the surface connection relationship based on the surface topology information and construct a topology connection matrix, identify feature boundaries based on the boundary feature information and extract boundary constraints, perform a structured expression of the topology connection matrix and the boundary constraints and construct an initial constraint graph, retrieve physical attribute requirements from the manufacturing knowledge base corresponding to the target product category and generate physical attribute nodes, retrieve the processing operation sequence from the process knowledge base of the target product category and generate operation sequence nodes, insert the physical attribute nodes and the operation sequence nodes into the initial constraint graph and establish dependencies to obtain the manufacturing constraint graph; The nodes in the manufacturing constraint graph are labeled with types and a node type mapping table is constructed. All nodes in the manufacturing constraint graph are traversed and the constraint propagation path is determined. Iterative constraint propagation calculation is performed along the constraint propagation path and constraint conflicts are detected. Conflicting nodes are recorded and constraint relaxation is performed. This process is repeated until the states corresponding to all nodes converge. The degree of constraint satisfaction is statistically analyzed and the manufacturability evaluation value corresponding to the derived form is calculated.

7. The method according to claim 1, characterized in that, Based on the manufacturability assessment value, conflict regions are identified, and the geometric parameters of the conflict regions are constrained to obtain a physical implementation scheme through solution, including: Local gradient analysis is performed on the manufacturability assessment value, and the derived morphological regions where the gradient change rate exceeds a preset change rate threshold are marked as candidate conflict regions. The constraint conflict type and conflict intensity corresponding to the candidate conflict regions are extracted. Based on the constraint conflict type, the candidate conflict regions are divided into material conflict regions and process conflict regions. Based on the conflict intensity, the conflict degree of the material conflict regions and the process conflict regions is quantitatively evaluated. Adjacent conflict regions of the same type are merged, and the region boundary is determined based on the conflict degree quantitative evaluation result to obtain the conflict region. Extract the geometric parameters corresponding to the conflict region, perform parameter space mapping, and construct a parameter constraint network. Mark hard constraint edges and soft constraint edges in the parameter constraint network. Construct a constraint propagation tree based on the hard constraint edges and perform layer-by-layer constraint solving to obtain the hard constraint solution space. Construct an objective function based on the soft constraint edges and solve it to obtain the soft constraint solution space. Calculate the intersection of the hard constraint solution space and the soft constraint solution space and filter parameter configurations that satisfy constraint compatibility. Adjust the parameters of the geometry of the conflict region based on the parameter configuration to obtain corrected geometric data. Determine the processing sequence and assembly sequence based on the corrected geometric data and associate them to obtain a physical implementation scheme.

8. An IP-based intelligent derivative system for clothing and daily necessities, used to implement the method as described in any one of claims 1-7, characterized in that, include: The IP adaptation unit is used to acquire an IP prototype and extract the corresponding IP representation. Based on the IP representation and the functional and structural constraints corresponding to the target product category, a cross-modal mapping space is constructed. In the cross-modal mapping space, the IP representation is subjected to manifold transformation to obtain a product domain representation. Component decomposition is then performed to obtain morphological and style components. The coupling relationship between the morphological components and the functional constraints, as well as the compatibility relationship between the style components and the structural constraints, are calculated. Based on the coupling and compatibility relationships, the product domain representation is subjected to constraint-guided projection to obtain an adaptation representation. The morphology derivation unit is used to project the adaptation representation onto the product morphology space and perform geometric transformations based on preset morphology topology constraints to generate a candidate morphology set, calculate the feature consistency metric value between each candidate morphology in the candidate morphology set and the IP prototype, and filter the candidate morphology set based on the feature consistency metric value to obtain a derived morphology. The manufacturing constraint unit is used to construct an initial constraint diagram based on the geometric structure of the derived form and add physical attribute nodes and operation sequence nodes corresponding to the target product category to obtain a manufacturing constraint diagram. It performs constraint propagation calculation on the manufacturing constraint diagram to obtain the manufacturability evaluation value corresponding to the derived form. Based on the manufacturability evaluation value, it identifies conflict areas and performs constraint satisfaction calculation on the geometric parameters of the conflict areas to obtain a physical implementation scheme.

9. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.

Images