Model digital generative design method and device based on continuous implicit function

Abstract

The application provides a model digital generation type design method and device based on a continuous implicit function, and relates to the technical field of artificial intelligence, which comprises the following steps: performing copyright and trademark similarity detection processing on design requirement information to obtain target design requirement information, and sending the target design requirement information to a conditional diffusion model to generate a two-dimensional concept map of a to-be-rendered jewelry; performing implicit three-dimensional reconstruction processing on the two-dimensional concept map to obtain a jewelry three-dimensional geometric model, and performing automatic matching processing on the jewelry three-dimensional geometric model to obtain a jewelry target three-dimensional geometric model containing physical material parameters; performing physical rendering processing and image enhancement processing on the jewelry target three-dimensional geometric model to obtain a multi-angle high-definition rendering map of the jewelry, and sending the jewelry target three-dimensional geometric model and the multi-angle high-definition rendering map as target output to a user end. The application can significantly improve the accuracy and model construction efficiency of jewelry three-dimensional model construction.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence, and in particular to a method and apparatus for digital generative design of models based on continuous implicit functions. Background Technology

[0002] Currently, related technologies suggest that AI-assisted tools can replace traditional handmade jewelry design. Existing jewelry modeling tools mainly use structural indexing or triangular networks to create 3D models of jewelry. However, due to the complex structures of jewelry, such as openwork carvings or twisted curved surfaces, the accuracy of 3D jewelry models built using existing 3D modeling methods is relatively low. In addition, existing AI-assisted tools often lack accurate simulation of the physical properties of jewelry and lack verification of the infringement risks and engineering feasibility of jewelry designs. As a result, although the designed 3D jewelry models are beautiful, they do not have practical production value. Summary of the Invention

[0003] In view of this, the purpose of this invention is to provide a model digital generative design method and apparatus based on continuous implicit functions, which can significantly improve the accuracy and efficiency of jewelry 3D model construction.

[0004] In a first aspect, embodiments of the present invention provide a model digital generative design method based on continuous implicit functions. The method includes: performing copyright and trademark similarity detection processing based on the design requirements information of the jewelry to be rendered, obtaining target design requirements information, and sending the target design requirements information to a conditional diffusion model to generate a two-dimensional concept map of the jewelry to be rendered; performing implicit three-dimensional reconstruction processing on the two-dimensional concept map to obtain a three-dimensional geometric model of the jewelry, and obtaining a target three-dimensional geometric model of the jewelry containing physical material parameters by performing automatic matching processing of materials and colors on the three-dimensional geometric model of the jewelry; obtaining a multi-angle high-definition rendering image of the jewelry by performing physical rendering processing and image enhancement processing on the target three-dimensional geometric model of the jewelry, and sending the target three-dimensional geometric model of the jewelry and the multi-angle high-definition rendering image as target outputs to the user terminal.

[0005] In one implementation, the step of performing copyright and trademark similarity detection processing based on the design requirements information of the jewelry to be rendered to obtain target design requirements information includes: converting the design requirements information into semantic feature vectors through a semantic encoding model, and performing cosine similarity calculation processing between the semantic feature vectors and vectors in a preset copyright and trademark feature database to obtain a cosine similarity score; when the cosine similarity score is greater than a preset similarity threshold, intercepting the vector corresponding to the cosine similarity score to obtain the filtered target design requirements information.

[0006] In one implementation, the step of sending target design requirement information to a conditional diffusion model to generate a two-dimensional concept map of the jewelry to be rendered includes: mapping the target design requirement information into a semantic conditional vector through a text encoder, and fusing the semantic conditional vector with a noise vector to obtain a latent vector; sending the latent vector to the conditional diffusion model for reverse denoising processing to obtain a two-dimensional concept map.

[0007] In one implementation, the step of performing implicit 3D reconstruction processing on a 2D concept map to obtain a 3D geometric model of jewelry includes: performing feature extraction processing on the 2D concept map to obtain image feature vectors, and sending the image feature vectors and a preset set of sampling points to an implicit neural field model to obtain a signed distance function corresponding to each sampling point; determining the distance information between each sampling point and the 3D geometric model of jewelry to be generated based on the signed distance function, and using the distance information to extract zero-value isosurfaces to generate the 3D geometric model of jewelry.

[0008] In one embodiment, after the step of generating a three-dimensional geometric model of jewelry, the process includes: performing a manufacturing inspection on the three-dimensional geometric model of jewelry; if there are problem areas in the three-dimensional geometric model of jewelry that do not meet the preset manufacturing constraints, then performing local geometric correction processing on the problem areas.

[0009] In one implementation, the step of obtaining a target three-dimensional geometric model of jewelry containing physical material parameters by automatically matching the material and color of the jewelry three-dimensional geometric model includes: matching corresponding material parameters and color information from a preset material database according to the geometric characteristics of the jewelry three-dimensional geometric model; and configuring the material and color of the jewelry three-dimensional geometric model based on the material parameters and color information using a micro-surface reflection model to obtain a target three-dimensional geometric model of jewelry containing physical material parameters.

[0010] In one implementation, the step of obtaining a multi-angle high-definition rendering image of the jewelry by performing physical rendering and image enhancement processing on the three-dimensional geometric model of the jewelry target includes: using a path tracing-based global illumination rendering model to perform physical rendering processing on the three-dimensional geometric model of the jewelry target by simulating the reflection, refraction and scattering process of light in the scene, and performing image enhancement processing on the rendered image to obtain a multi-angle high-definition rendering image of the jewelry.

[0011] Secondly, embodiments of the present invention also provide a model digital generative design device based on continuous implicit functions. The device includes: an image generation module, which performs copyright and trademark similarity detection processing based on the design requirements information of the jewelry to be rendered, obtains target design requirements information, and sends the target design requirements information to a conditional diffusion model to generate a two-dimensional concept map of the jewelry to be rendered; an implicit three-dimensional reconstruction module, which performs implicit three-dimensional reconstruction processing on the two-dimensional concept map to obtain a three-dimensional geometric model of the jewelry, and obtains a target three-dimensional geometric model of the jewelry containing physical material parameters by performing automatic matching processing of materials and colors on the three-dimensional geometric model of the jewelry; and a model rendering module, which obtains a multi-angle high-definition rendering image of the jewelry by performing physical rendering processing and image enhancement processing on the target three-dimensional geometric model of the jewelry, and sends the target three-dimensional geometric model of the jewelry and the multi-angle high-definition rendering image as target outputs to the user terminal.

[0012] Thirdly, embodiments of the present invention also provide a server, including a processor and a memory, the memory storing computer-executable instructions executable by the processor, the processor executing the computer-executable instructions to implement any of the methods provided in the first aspect.

[0013] Fourthly, embodiments of the present invention also provide a computer-readable storage medium storing computer-executable instructions, which, when invoked and executed by a processor, cause the processor to implement any of the methods provided in the first aspect.

[0014] The embodiments of the present invention bring the following beneficial effects:

[0015] This invention provides a digital generative design method and apparatus for jewelry based on continuous implicit functions. The method performs copyright and trademark similarity detection based on the design requirements of the jewelry to be rendered, obtaining target design requirements information. This target design requirements information is then sent to a conditional diffusion model to generate a two-dimensional concept map of the jewelry. Subsequently, implicit three-dimensional reconstruction processing is performed on the two-dimensional concept map to obtain a three-dimensional geometric model of the jewelry. Automated material and color matching processing is then performed on the three-dimensional geometric model to obtain a target three-dimensional geometric model of the jewelry containing physical material parameters. Finally, physical rendering and image enhancement processing are performed on the target three-dimensional geometric model of the jewelry to obtain a multi-angle high-definition rendering image of the jewelry. The target three-dimensional geometric model and the multi-angle high-definition rendering image are then sent as target outputs to the user terminal. This invention can significantly improve the accuracy and efficiency of jewelry three-dimensional model construction.

[0016] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention are realized and obtained in accordance with the structures particularly pointed out in the description, claims and drawings.

[0017] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0018] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0019] Figure 1 A flowchart illustrating a model digital generative design method based on continuous implicit functions provided in an embodiment of the present invention;

[0020] Figure 2 A schematic diagram illustrating the specific process of a model digital generative design method based on continuous implicit functions provided in an embodiment of the present invention;

[0021] Figure 3 A schematic diagram of the structure of a model digital generative design device based on continuous implicit functions provided in an embodiment of the present invention;

[0022] Figure 4 This is a schematic diagram of the structure of a server provided in an embodiment of the present invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, 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.

[0024] Currently, traditional handcrafted design is the cornerstone of the jewelry industry, typically relying on hand-drawn sketches, wax prototypes, or metal casting. This method is highly dependent on individual skills and experience, resulting in a lengthy design process that often takes weeks or even months to complete a complex piece. The drawbacks of the handcrafted approach are particularly prominent: first, significant subjective bias, where different designers' interpretations can lead to discrepancies between the final product and the initial concept; second, difficulty in iteration, as errors in prototype creation necessitate a complete overhaul, increasing material waste and costs; and third, difficulty in handling fine details, such as the precise curvature of metal curves or the light and shadow reflections on gemstone facets, which are easily compromised by human error. Handcrafted design struggles to quickly respond to personalized customization needs. Therefore, traditional methods rely on hand-drawn sketches, physical prototypes, or computer-aided design software, which are time-consuming and require repeated manual adjustments to capture details such as gemstone settings, metal curves, and textures. In the high-end customization sector, personalized needs are limited by time and skill, leading to extended iteration cycles, increased costs, and design deviations.

[0025] In the wave of digitalization, the industry needs innovative tools to accelerate the transformation from concept to finished product. While existing AI drawing software can quickly generate 2D images of jewelry based on text prompts, these programs primarily focus on image generation and cannot convert 2D to 3D models. This results in output limited to static planar views, preventing users from generating, editing, or adjusting the angle of 3D models—for example, viewing the three-dimensional structure of jewelry from a side or rotating perspective. This keeps designs at the visual concept stage, preventing them from entering the actual prototype development phase. Furthermore, these AI tools lack simulation of the physical properties of jewelry, such as material reflection or light interaction. While the generated images may be aesthetically pleasing, they lack engineering feasibility and are difficult to integrate with downstream manufacturing processes. Existing auxiliary software, although capable of partial modeling or rendering, is mostly limited to a single stage, lacking end-to-end integration. It cannot meet the needs of rapid prototyping and multi-angle visualization, nor can it satisfy the efficiency and precision requirements of the modern jewelry industry.

[0026] Based on this, the model digital generative design method and device based on continuous implicit functions provided by this invention can adopt multiple steps such as semantic detection, diffusion model generation, image to 3D modeling, physical optics rendering and multi-angle high-definition rendering enhancement, and achieve optimized adaptation to specific data of the jewelry industry through internal calibration and training, thereby constructing a scalable and compliant intelligent jewelry design system.

[0027] See Figure 1 The diagram shows a flowchart of a model digital generative design method based on continuous implicit functions. The method mainly includes the following steps S102 to S106:

[0028] Step S102: Based on the design requirements information of the jewelry to be rendered, perform copyright and trademark similarity detection processing to obtain target design requirements information, and send the target design requirements information to the conditional diffusion model to generate a two-dimensional concept map of the jewelry to be rendered.

[0029] In one implementation, the design requirements information for the jewelry to be rendered may include: text descriptions, image samples, and historical design drafts. First, the system's copyright and trademark detection module structurally organizes the internal jewelry industry database, enabling feature extraction from text descriptions, image samples, and historical design drafts. Text data is segmented using a controllable dictionary (composed of jewelry terminology, material names, setting methods, classic style names, etc.), and then a semantic encoding model is used to convert the text into computable feature vectors. Image data is extracted using a convolutional neural network model to extract multi-scale visual feature vectors, representing the image's shape and outline, gemstone facet details, setting structure, and overall design. Subsequently, the system establishes a tag mapping table based on trademark category, rights holder name, registration number, validity period, etc., and constructs a vector index structure based on all feature vectors. A combination of segmented inverted indexes and hierarchical clustering indexes is used to improve retrieval speed and comparison accuracy.

[0030] Step S104: Implicit 3D reconstruction processing is performed on the 2D concept map to obtain a 3D geometric model of the jewelry. Then, by automatically matching the material and color of the 3D geometric model of the jewelry, a 3D geometric model of the jewelry target containing physical material parameters is obtained.

[0031] In one implementation, the conversion from a 2D image to a 3D model can be achieved based on the Tripo engine. The advantage of implicit 3D reconstruction is that traditional modeling methods, such as those using structural indexing or triangular networks for jewelry 3D modeling, construct the model by describing the specific coordinates of the model frame and the orientation and connection relationships between the frames. However, this modeling method struggles to accurately describe the structural characteristics of jewelry and results in extremely large file sizes. Implicit modeling, on the other hand, provides a set of coordinates that can encompass the model to be generated. By determining the signed distance value (SDF) of each coordinate point in this set, i.e., ... The implicit modeling method defines the spatial occupancy of the model by placing it outside (positive value), inside (negative value), and on the surface (zero). This method has three advantages: 1. It excels at handling complex smooth surfaces: Jewelry often features complex structures such as openwork carvings or twisted surfaces, and implicit functions are adept at expressing such continuous, smooth surfaces. Its accuracy is much higher than methods like triangular mesh modeling, which approximate the surface plane by plane (mesh). 2. The model generated through implicit modeling is always a closed entity (without broken surfaces) because it extracts boundaries from internal and external rules. This also facilitates subsequent 3D printing manufacturing. 3. Multi-view Figure 1 Consistency: Implicit functions define a unified rule in space, so the model is reasonable and self-consistent from all perspectives. Even if only a front view is input during reconstruction, the system will use prior knowledge such as symmetry to complete the back view.

[0032] Step S106: By performing physical rendering and image enhancement processing on the three-dimensional geometric model of the jewelry target, a multi-angle high-definition rendering image of the jewelry is obtained, and the three-dimensional geometric model of the jewelry target and the multi-angle high-definition rendering image are sent to the user terminal as target output.

[0033] In one implementation, during the final output stage, the 3D model file (supporting standard formats such as OBJ and STL), material parameter files, and high-definition rendering atlas are output uniformly. Users can then directly browse and manipulate the design results in the front-end interactive interface, including operations such as rotation, scaling, material replacement, and color scheme modification. Furthermore, the system allows users to trigger a design iteration with a single click on the page, regenerating new concept images or material combinations, thus forming a rapid iteration and feedback loop. Through this interactive mechanism, users can not only efficiently obtain designs that meet their needs but also explore multiple design solutions in a short time, achieving highly personalized and large-scale jewelry design and production.

[0034] The above-described generative design method for digital models based on continuous implicit functions provided in this invention can combine creative input, copyright protection, concept generation, 3D modeling, material assignment, high-definition rendering, and interactive output, which can significantly improve the accuracy and efficiency of jewelry 3D model construction.

[0035] See Figure 2 The diagram shows a specific flowchart of a model digital generative design method based on continuous implicit functions. This invention also provides an implementation method for automated jewelry design, as detailed in (1) to (7) below:

[0036] (1) User input stage: The user submits text requirements, reference keywords or concept descriptions (i.e., design requirements information of the jewelry to be rendered). The system receives the input and enters the semantic parsing module. The design requirements information of the jewelry to be rendered can be natural language descriptions, keyword combinations, reference sketches and multimodal inputs. In order to avoid user input triggering potential copyright conflicts, the input content is first checked for copyright.

[0037] (2) Copyright and Trademark Detection Stage: The text parsing model generates semantic vectors, and the vector retrieval engine compares them with internal copyright data. If the similarity exceeds the threshold, the design is blocked and a modification is prompted. If there is no risk, the concept map generation stage begins.

[0038] In one implementation, design requirement information can be converted into semantic feature vectors using a semantic coding model. The semantic feature vectors are then compared with vectors in a preset copyright and trademark feature database using cosine similarity calculation to obtain a cosine similarity score. When the cosine similarity score is greater than a preset similarity threshold, the vector corresponding to the cosine similarity score is intercepted to obtain the filtered target design requirement information.

[0039] Specifically, after a user submits a design description or concept sketch prompt, the system uses DeepSeek-V2's semantic parsing service to convert the user input into a semantic vector in real time. This vector, after normalization, is fed into the vector retrieval engine (based on a fast similarity calculation algorithm), ensuring the comparison results are rigorously reliable. When a potentially high similarity to a registered copyright or trademark in the database is detected, the system immediately feeds the result back to the front-end design module, automatically triggering blocking or prompting the user to modify the design, and recording a complete retrieval log and risk marker. This ultimately forms a secure, transparent, and traceable copyright review closed loop from data initialization, semantic parsing, vector retrieval to front-end feedback.

[0040] Each protected object in the database is represented by a vector E(d), and the user input (text or image) is represented by a vector E(q). First, the copyright detection module uses cosine similarity to measure the semantic similarity between the user input and the protected objects in the database. The mathematical expression of the fast similarity calculation algorithm here is:

[0041]

[0042] in, This represents a semantic and image encoder function, the output of which is... The real-valued vectors, the encoder should ensure a common or comparable embedding space (i.e., multimodal alignment) between text and images; q represents user input (which can be a natural language phrase, a set of keywords, or a reference sketch, etc.); d represents the prototype representation of a single protected object in the database; This represents the cosine similarity score; values ​​closer to 1 indicate greater semantic similarity. For industrial deployment, it is necessary to... The output is batch normalized and truncated with possible principal components, and approximate nearest neighbor is used to accelerate retrieval during the indexing stage. The similarity threshold is denoted as... ,when Interception or manual review may be triggered at any time; The initial screening values ​​need to be systematically selected using precision-recall (PR) curves on the validation set: the PR curve is defined as the performance trajectory at different thresholds, and the standard definitions of precision and recall are... , Here, TP, FP, and FN represent the counts of true positives, false positives, and false negatives, respectively. The PR curve is used to weigh the costs of missed positivity against false negatives. If the business is more concerned about copyright risk, a strategy that increases the recall threshold (tolerating lower accuracy) should be chosen; if the business is more concerned about the impact of false blocking on user experience, a strategy that increases accuracy threshold should be chosen. This solution sets... The initial screening value was 0.85, which is relatively consistent.

[0043] (3) Concept map generation stage: The semantic condition vector is input into the concept map diffusion model. The diffusion model is used to gradually denoise and generate the concept map. In one implementation, the target design requirement information can be mapped into a semantic condition vector by a text encoder, and the semantic condition vector is fused with the noise vector to obtain a latent vector. The latent vector is sent to the condition diffusion model for reverse denoising to obtain a two-dimensional concept map. The latent vector is a combination of highly compressed mathematical coordinate information and random noise. Its function is to be input into the diffusion model so that the model can reverse denoise the latent vector, which contains both randomness and model coordinates, thereby obtaining a clear two-dimensional concept map.

[0044] Specifically, in the conditional diffusion model, the potential initial noise vector of the diffusion process is denoted as... User semantics are mapped to semantic vectors through a text encoder. To inject semantic information into the latent space of the diffusion model, a trainable mapping network can be used. (Here, we use a multilayer perceptron (MLP) to represent this), mapping T to a vector compatible with the dimensions of the latent space. The fusion of semantic and visual latent vectors employs two rigorous and numerically stable forms: additive fusion or convex combinatorial fusion.

[0045] Additive fusion:

[0046]

[0047] Convex combination fusion:

[0048]

[0049] in, These are semantic-visual fusion coefficients, used to control the influence of semantic mapping on the initial latent representation. This indicates that semantics are not injected. (This means that the original noise is completely replaced by a semantic mapping). The necessary numerical normalization operation is indicated to ensure the stability of model training and inference; z represents the initial Gaussian noise vector (dimension m) used to generate diversity; T represents the semantic vector output by the text encoder (dimension k). (i.e., MLP) represents a trainable mapping that maps semantic vectors to the latent space; z' represents the initial latent vector after semantic injection, which serves as the starting point for backsampling in the diffusion model.

[0050] The MLP (Multilayer Perceptron) should include residual connections and layer normalization to ensure gradient flow. The input dimension k should match the text encoder output, and the output dimension m should match the latent noise z. During training, a diffusion decoder with conditional cross-attention can be used. To further integrate z' with the text conditions, the backdiffusion sampling process is executed iteratively through multiple steps:

[0051]

[0052] in, Let c be the noisy image representation at time t, and c be the condition vector. The parameterization function for the trained denoising network is responsible for deriving z' into a clear conceptual image.

[0053] Here settings The initial value is 0.5 to control the influence of semantic information in the generation process. Based on this mechanism, the randomness and diversity of the results are preserved, while maintaining consistency between the generated image and user needs. Finally, the backdiffusion process transforms z′ into a concept map with a clear jewelry appearance outline, style characteristics, and basic color scheme, laying the foundation for the next stage of 3D reconstruction.

[0054] (4) Three-dimensional modeling stage: The concept map is input into the implicit modeling engine to generate a three-dimensional model, and the basic topology and normal information are automatically generated. In one implementation, the two-dimensional concept map can be processed by feature extraction to obtain image feature vectors, and the image feature vectors and the preset sampling point set are sent to the implicit neural field model to obtain the signed distance function corresponding to each sampling point. Then, the distance information between each sampling point and the jewelry three-dimensional geometric model to be generated is determined according to the signed distance function, and the zero value isosurface is extracted using the distance information to generate the jewelry three-dimensional geometric model. Furthermore, the jewelry three-dimensional geometric model is subjected to manufacturing inspection processing. If there are problem areas in the jewelry three-dimensional geometric model that do not meet the preset manufacturing constraints, the problem areas are subjected to local geometric correction processing.

[0055] Specifically, the Tripo engine can be used to convert 2D images into 3D models. This is based on implicit function representation, which samples spatial points to learn and predict whether a point lies on or inside an object, thereby reconstructing a complete 3D geometric model. Compared to traditional mesh modeling methods, implicit representation can better capture complex surfaces and fine structures, which is particularly crucial for jewelry design, as jewelry often contains delicate curves, settings, and highly precise symmetrical shapes.

[0056] Based on the Tripo engine, parameterized implicit functions are used. To perform conditional representation and reconstruction of 3D shapes, where, for any 3D point... and from two-dimensional input images via image encoder Extracted latent vector Joint modeling is performed, and the signed distance function (SDF) is expressed as follows: This causes the target surface to change from a zero isosurface. Given. For the set of sampling points. Apply the following loss function:

[0057]

[0058] Where N represents the total number of sampling points used for this sample; The reference (true) SDF can be calculated from the implicitly true model. To ensure that the implicit field is a well-defined range field and to improve the consistency of the surface normal, an Eikonal regularization term is introduced:

[0059]

[0060] In addition, normal consistency loss can be added. (Matching the predicted gradient direction with the true surface normal) and symmetry loss (On axially symmetric / centrally symmetric objects) (Consistent with its symmetric point mapping output). The total loss is in weighted sum form:

[0061]

[0062] Among them, each weight For hyperparameters, the following values ​​can be used in the examples: , , , ( , The network architecture is adjusted to balance details and prior knowledge based on task requirements. An image encoder is used. Map the input image to a global latent vector. It also preserves multi-scale two-dimensional feature maps.

[0063] Implicit decoder A multilayer perceptron is employed, and position encoding is applied to the input (x) to enhance high-frequency representation capabilities. Specifically, the coordinates (x) are mapped as follows:

[0064]

[0065] In this example, K is set to 6 to 10 to control the highest frequency; the MLP uses 8 layers with 256 to 512 neurons per layer, and combines skip connections and layer normalization to stabilize training.

[0066] For single-view input, a projection conditionalization strategy is adopted: for each 3D point (x), the camera parameters (P) are used to project (x) onto the image plane to obtain the pixel coordinates. Local features are obtained by bilinear sampling from the two-dimensional feature map. ,Will The input is obtained by concatenating (x) with the global vector (z). To enhance visible details, a multi-level sampling strategy is employed: coarse sampling is performed during training on each sample. (e.g., 5,000 to 20,000 points) to cover the entire structure, and to add fine sampling within the surface neighborhood determined by the current prediction or reference model. (For example, 5,000 to 30,000 points, distance threshold) Use a sampling density of 0.01 to 0.05 units to preserve small-scale structures such as setting grooves, thin walls, and sharp edges; for jewelry details, it is recommended to force higher sampling density around setting positions, curved corners, and axes of symmetry.

[0067] Training details include the use of the Adam optimizer, an initial learning rate of lr = 1e-4, a batch size of 8–32, and training epochs ranging from 200k to 1Msteps depending on the amount of data and convergence. Hardware usage includes GPUs with high video memory (multi-GPU distributed training). In the inference phase, after predicting the implicit field for a given input (I), isosurfaces are extracted in the target voxel domain using the MarchingCubes algorithm. ( For SDF, For the occupied field), post-mesh processing is then performed: connected component analysis to remove isolated fragments, void filling, normal smoothing, and curvature-guided boundary protection simplification (preserving high curvature regions to maintain groove detail), followed by manufacturability checks (minimum wall thickness detection threshold). In the embodiments, the value is 0.5~1.0 mm, depending on the material and process. For areas that do not meet the manufacturing constraints, local geometric modifications are automatically triggered (for example, the fillet radius is set to 0.1~0.5 mm).

[0068] To reduce computational and storage overhead, local multi-resolution feature meshes can be used in the implementation to accelerate convergence and improve detail representation; to enhance the plausibility of the unseen back face region, symmetric loss can be employed. Shape prior and (if available) multi-view fusion. Evaluation metrics include ChamferDistance, EarthMover'sDistance, occupancy IoU, and normal consistency to quantify reconstruction accuracy, while also performing visualization rendering comparison and manufacturing inspection.

[0069] In summary, by conditionalizing 2D visual information into a continuous implicit field, employing surface-aware sampling and physical regularization, and combining isosurface extraction with manufacturing-oriented post-processing, high-fidelity automated reconstruction of 3D jewelry models containing complex curved surfaces, fine settings, and high-precision symmetrical shapes can be achieved. Key parameters include the implicit network structure (number of layers and width), position encoding frequency (K), and sampling scale. Eikonal weights Manufacturing minimum wall thickness The above parameter ranges can be adjusted according to the specific jewelry type and craftsmanship requirements to obtain the best results.

[0070] To further ensure the accuracy and continuity of geometric modeling, a constraint mechanism is introduced during the training process, resulting in high stability of the 3D model in both surface details and overall structure. This method enables the system to quickly convert 2D concept maps into high-fidelity 3D mesh models, laying a solid foundation for subsequent material assignment and rendering.

[0071] (5) Material and color matching stage: The user selects the material or the system automatically recommends it to generate a three-dimensional model with material attributes. In one implementation, according to the geometric characteristics of the jewelry three-dimensional geometric model, the corresponding material parameters and color matching information are matched from the preset material database. Then, through the micro-surface reflection model, based on the material parameters and color matching information, the material configuration and color matching of the jewelry three-dimensional geometric model are performed to obtain the jewelry target three-dimensional geometric model containing physical material parameters.

[0072] Specifically, the system utilizes a built-in database of jewelry materials and optical properties, containing physical and optical parameters of various common precious metals (such as gold, white gold, and platinum) and gemstones (such as diamonds, sapphires, and emeralds), including refractive index, reflectivity, roughness, and dispersion coefficient. Using an AI recommendation algorithm, it automatically matches suitable materials and color schemes to the model based on its surface geometry and user needs. For the representation of reflection and refraction, the system employs a physically based rendering model, the Cook-Torrance microsurface reflection model, to simulate metallic highlights, gemstone refraction, and multi-angle reflection effects. The Cook-Torrance microsurface reflection serves as the core, and its standardized expression is as follows:

[0073]

[0074] Wherein, the vector is defined as the normal n, the incident light vector l, the observation vector v, and the half-vector. The three core sub-items and their parameters are explained below:

[0075] 1. The normal distribution function (D(h)) describes the orientation ( The micro-facet density; to balance the strong mirror-like luster and metallic sheen of gemstones, the present invention preferably uses the GGX form:

[0076]

[0077] Among them, parameters Characterizes surface roughness. It tends to resemble a mirror.

[0078] 2. Fresnel term Describes the wavelength / polarization-dependent reflectivity at the angle between the incident and half-vectors; for engineering implementation, the Schlick approximation is used to ensure computational efficiency and physical consistency.

[0079]

[0080] in, To view the incident light directly (maintain) Substrate reflectivity under ( ) conditions: for nonconductors Can be determined by refractive index Pre-calculations are performed using the Fresnel formula, and for metals, the spectral values ​​of complex refractive index and reflectivity are directly used to reflect the metal's dispersion and wavelength-selective absorption.

[0081] 3. Occlusion-Shading Functions This is used to correct for energy attenuation caused by mutual occlusion between micro-area elements; a practical form can be approximated by the Smith method and the Schlick-GGX method.

[0082]

[0083] This microsurface framework allows for the direct manipulation of the material's physical and optical parameters (refractive index). Absorption coefficient k or dispersion coefficient, surface roughness (metallic) mapping to The terms D, G, etc., are used to reproduce real phenomena such as diamond refraction, high dispersion, metallic mirror highlights, and energy distribution of rough metals in the global illumination solution of path tracing.

[0084] This method not only makes the 3D model realistic and believable in form, but also closely resembles the physical properties of actual jewelry in terms of material and optical performance, thereby effectively improving the realism and aesthetic effect of the rendering.

[0085] (6) High-definition rendering stage: The three-dimensional model is input into the renderer to generate a multi-view high-definition rendering scene. In one implementation, a path-tracing-based global illumination rendering model is adopted. By simulating the reflection, refraction and scattering process of light in the scene, the three-dimensional geometric model of the jewelry target is physically rendered. The rendered image is then enhanced to obtain a multi-view high-definition rendering image of the jewelry.

[0086] Specifically, after the materials are applied, the system enters the high-definition rendering stage. In this stage, the invention employs a path-tracing-based global illumination rendering method. By simulating the process of light reflecting, refracting, and scattering multiple times in the scene, a highly realistic visual effect is achieved. The path-tracing method can accurately reproduce the complex optical properties of jewelry materials, such as the strong refraction of diamonds, the specular reflection of metals, and the color variations between different gemstones. Furthermore, to meet the needs of multi-angle display, the system automatically generates rendered images from multiple perspectives by setting different camera parameters, ensuring that users can observe the details of the jewelry design from all angles.

[0087] In the image enhancement and style refinement stages following high-resolution rendering, this invention introduces a NanoBanana-based class model to generate high-definition, multi-angle stereoscopic rendering images. This invention formalizes this process as a learning generator. Given a low-resolution rendered image x (obtained through physical path tracing or a multi-view atlas), the network outputs a super-resolution image. The training objective employs a multi-factor composite loss: pixel reconstruction term, perceptual loss (based on a pre-trained feature extractor). The Euclidean distance and adversarial loss (improving high-frequency details and texture) are typical weighted objective functions:

[0088]

[0089] Where y is a high-quality reference (high-sampling rendering). It can be used as a perceptual feature extractor such as VGG. The discriminator D generates the weights to enhance visual realism. In engineering, cross-validation is used to determine a balance between sharpness, artifact suppression, and texture fidelity. In terms of implementation details, Nano-Banana-like architectures typically mix convolutional and self-attention Transformer modules to balance local detail with global consistency, and introduce color preservation constraints and text or logo preservation terms (when rendering inscriptions or logos) to prevent stylization processes from destroying critical information. For jewelry images, the network input can also be expanded to carry PBR parameter maps, enabling the super-resolution process to selectively enhance refractive highlights and dispersion based on material parameters.

[0090] When automatically matching materials, the system uses the geometric features of the model surface (such as curvature distribution and micro-facet orientation statistics), user preferences (such as warm gold, cool silver, and high fire color), and business constraints (cost and machinability) as input features. The neural recommender outputs a set of candidate material parameters. Furthermore, by using rapid multi-view path tracing and Nano-Banana super-resolution loops to verify visual effects, a closed loop of automated material selection, physical rendering, learning enhancement, and user feedback is formed, ensuring that the final 3D model is highly close to real jewelry in terms of form, material, and optical performance and meets commercial output standards.

[0091] (7) Front-end interaction and output stage: Users can view the 3D model and rendering on the front end, rotate, scale, replace materials, and modify color schemes. They can trigger the design iteration again with one click and export 3D model files, material parameter files and rendering atlases. Thus, starting from user input, compliance is ensured through anti-copyright detection, then visual creativity is provided through concept map generation, geometric solidification is achieved through 3D modeling, physical properties are then given by material and color generation, and finally the results are displayed and secondary iteration is achieved through high-definition rendering and interactive output. It integrates a variety of advanced technologies such as deep semantic analysis, diffusion generation modeling, implicit 3D reconstruction, physically based optical rendering and artificial intelligence enhanced rendering, forming an efficient, automated, compliant and highly scalable technical route.

[0092] In practical applications, the system generates high-definition jewelry design drawings that meet user requirements based on user interaction information. The interaction information includes a sketch file of the jewelry model. According to the requirements, a high-definition drawing is generated through semantic model transformation and construction. Then, a graph-based model engine based on implicit modeling is used to generate a 3D model. At this point, the 3D model of the corresponding jewelry can be exported. Alternatively, the material and angle information of the 3D model can be adjusted to generate a high-definition jewelry drawing that meets expectations. Furthermore, a jewelry drawing with a background can be generated according to the requirements of the jewelry environment.

[0093] In summary, this invention offers the following advantages: 1. By introducing a semantic detection mechanism based on an open-source semantic model, copyright protection is automated in the jewelry design process, fundamentally resolving potential infringement risks; 2. Based on raw image models and internal calibration training data, the capabilities of AI-generated models are specifically adapted to the jewelry industry, ensuring style and professionalism from the concept drawing stage; 3. Based on an implicit modeling engine, the limitations of traditional modeling in expressing complex surfaces are overcome, resulting in fast and accurate 3D models; 4. The introduction of a physically-driven rendering model in the material and color matching stages ensures that the final rendering effect closely matches actual physical laws, providing realism and persuasiveness in the presentation; 5. The high-definition rendering stage not only uses path tracing to simulate real lighting but also combines AI super-resolution enhancement, achieving a dual unity of physical realism and visual aesthetics; 6. The interactive output mechanism ensures that users can quickly provide feedback and iterate again, forming an efficient closed loop and meeting large-scale personalized needs.

[0094] Regarding the model digital generative design method based on continuous implicit functions provided in the foregoing embodiments, this invention provides a model digital generative design apparatus based on continuous implicit functions. (See attached image.) Figure 3 The diagram shows a structural schematic of a model digital generative design device based on continuous implicit functions. The device includes the following parts:

[0095] The image generation module 302 performs copyright and trademark similarity detection processing based on the design requirements information of the jewelry to be rendered, obtains the target design requirements information, and sends the target design requirements information to the conditional diffusion model to generate a two-dimensional concept map of the jewelry to be rendered.

[0096] The implicit 3D reconstruction module 304 performs implicit 3D reconstruction processing on the 2D concept map to obtain a 3D geometric model of the jewelry. Then, through automatic matching processing of the material and color of the 3D geometric model of the jewelry, a 3D geometric model of the jewelry target containing physical material parameters is obtained.

[0097] The model rendering module 306 performs physical rendering and image enhancement processing on the three-dimensional geometric model of the jewelry target to obtain a multi-angle high-definition rendering image of the jewelry, and sends the three-dimensional geometric model of the jewelry target and the multi-angle high-definition rendering image as the target output to the user terminal.

[0098] The above-described model digital generative design device based on continuous implicit functions provided in this application embodiment can significantly improve the accuracy and efficiency of jewelry 3D model construction.

[0099] In one embodiment, when performing the step of performing copyright and trademark similarity detection processing based on the design requirement information of the jewelry to be rendered to obtain the target design requirement information, the image generation module 302 is further configured to: convert the design requirement information into a semantic feature vector through a semantic encoding model, and perform cosine similarity calculation processing between the semantic feature vector and the vector in the preset copyright and trademark feature database to obtain a cosine similarity score; when the cosine similarity score is greater than a preset similarity threshold, the vector corresponding to the cosine similarity score is intercepted to obtain the filtered target design requirement information.

[0100] In one embodiment, when performing the step of sending the target design requirement information to the conditional diffusion model to generate a two-dimensional concept map of the jewelry to be rendered, the image generation module 302 is further configured to: map the target design requirement information into a semantic conditional vector through a text encoder, and fuse the semantic conditional vector with a noise vector to obtain a latent vector; send the latent vector to the conditional diffusion model for reverse denoising processing to obtain a two-dimensional concept map.

[0101] In one embodiment, when performing implicit 3D reconstruction processing on a 2D concept map to obtain a 3D geometric model of jewelry, the implicit 3D reconstruction module 304 is further configured to: perform feature extraction processing on the 2D concept map to obtain image feature vectors, and send the image feature vectors and a preset set of sampling points to an implicit neural field model to obtain signed distance functions corresponding to each sampling point; determine the distance information between each sampling point and the 3D geometric model of jewelry to be generated based on the signed distance function, and use the distance information to extract zero-value isosurfaces to generate the 3D geometric model of jewelry.

[0102] In one embodiment, after generating the three-dimensional geometric model of the jewelry, the implicit three-dimensional reconstruction module 304 is further used to: perform manufacturing inspection processing on the three-dimensional geometric model of the jewelry; if there are problem areas in the three-dimensional geometric model of the jewelry that do not meet the preset manufacturing constraints, then perform local geometric correction processing on the problem areas.

[0103] In one embodiment, when performing the step of automatically matching the material and color of the jewelry three-dimensional geometric model to obtain a jewelry target three-dimensional geometric model containing physical material parameters, the model rendering module 306 is further configured to: match the corresponding material parameters and color information from a preset material database according to the geometric characteristics of the jewelry three-dimensional geometric model; and configure the material and color of the jewelry three-dimensional geometric model based on the material parameters and color information using a micro-surface reflection model to obtain a jewelry target three-dimensional geometric model containing physical material parameters.

[0104] In one embodiment, when performing the step of obtaining a multi-angle high-definition rendering image of the jewelry by performing physical rendering and image enhancement processing on the three-dimensional geometric model of the jewelry target, the model rendering module 306 is further configured to: use a path tracing-based global illumination rendering model to perform physical rendering processing on the three-dimensional geometric model of the jewelry target by simulating the reflection, refraction and scattering process of light in the scene, and obtain a multi-angle high-definition rendering image of the jewelry by performing image enhancement processing on the rendered image.

[0105] The device provided in this embodiment of the invention has the same implementation principle and technical effect as the aforementioned method embodiment. For the sake of brevity, any parts not mentioned in the device embodiment can be referred to the corresponding content in the aforementioned method embodiment.

[0106] This invention provides a server, specifically, the server includes a processor and a storage device; the storage device stores a computer program, which, when run by the processor, executes the method described in any of the above embodiments.

[0107] Figure 4This is a schematic diagram of the structure of a server provided in an embodiment of the present invention. The server 100 includes: a processor 40, a memory 41, a bus 42 and a communication interface 43. The processor 40, the communication interface 43 and the memory 41 are connected through the bus 42. The processor 40 is used to execute executable modules, such as computer programs, stored in the memory 41.

[0108] The memory 41 may include high-speed random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 43 (which can be wired or wireless), such as the Internet, wide area network, local area network, metropolitan area network, etc.

[0109] Bus 42 can be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 4 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0110] The memory 41 is used to store programs. After receiving an execution instruction, the processor 40 executes the program. The method executed by the device for defining the flow process disclosed in any of the foregoing embodiments of the present invention can be applied to the processor 40 or implemented by the processor 40.

[0111] Processor 40 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of processor 40 or by instructions in software form. Processor 40 can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this invention can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in memory 41. The processor 40 reads the information in memory 41 and, in conjunction with its hardware, completes the steps of the above method.

[0112] The computer program product of the readable storage medium provided in the embodiments of the present invention includes a computer-readable storage medium storing program code. The instructions included in the program code can be used to execute the methods described in the foregoing method embodiments. For specific implementation, please refer to the foregoing method embodiments, which will not be repeated here.

[0113] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0114] Finally, it should be noted that the above-described embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit it. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the technical scope disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A generative design method for digital models based on continuous implicit functions, characterized in that, The method includes: Based on the design requirements of the jewelry to be rendered, similarity detection of copyright and trademark is performed to obtain target design requirements information, and the target design requirements information is sent to the conditional diffusion model to generate a two-dimensional concept map of the jewelry to be rendered. Implicit 3D reconstruction processing is performed on the two-dimensional concept map to obtain a three-dimensional geometric model of the jewelry. Then, by automatically matching the material and color of the three-dimensional geometric model of the jewelry, a target three-dimensional geometric model of the jewelry containing physical material parameters is obtained. By performing physical rendering and image enhancement processing on the three-dimensional geometric model of the jewelry target, a multi-angle high-definition rendering image of the jewelry is obtained, and the three-dimensional geometric model of the jewelry target and the multi-angle high-definition rendering image are sent to the user terminal as target output. The step of sending the target design requirement information to the conditional diffusion model to generate a two-dimensional concept map of the jewelry to be rendered includes: mapping the target design requirement information into a semantic conditional vector through a text encoder, and fusing the semantic conditional vector with a noise vector to obtain a latent vector; sending the latent vector to the conditional diffusion model for reverse denoising processing to obtain the two-dimensional concept map. The step of performing implicit 3D reconstruction processing on the 2D concept map to obtain a 3D geometric model of jewelry includes: performing feature extraction processing on the 2D concept map to obtain image feature vectors, and sending the image feature vectors and a preset set of sampling points to an implicit neural field model to obtain a signed distance function corresponding to each sampling point; determining the distance information between each sampling point and the 3D geometric model of jewelry to be generated based on the signed distance function, and using the distance information to extract zero-value isosurfaces to generate the 3D geometric model of jewelry. The implicit neural field model is used to provide a coordinate set that completely encloses the model to be generated, and to generate signed distance values ​​for each coordinate point. Based on the signed distance values, it determines whether each coordinate point is outside, inside, or on the surface of the model to be generated, so as to reconstruct a fully enclosed symmetrical 3D geometric model of jewelry. The step after generating the three-dimensional geometric model of the jewelry includes: performing a manufacturing inspection process on the three-dimensional geometric model of the jewelry; if there are problem areas in the three-dimensional geometric model of the jewelry that do not meet the preset manufacturing constraints, then performing a local geometric correction process on the problem areas.

2. The model digital generative design method based on continuous implicit functions according to claim 1, characterized in that, The step of performing copyright and trademark similarity detection based on the design requirements information of the jewelry to be rendered, to obtain the target design requirements information, includes: The design requirement information is converted into a semantic feature vector through a semantic coding model, and the semantic feature vector is then compared with the vectors in a preset copyright and trademark feature database to calculate the cosine similarity and obtain a cosine similarity score. When the cosine similarity score is greater than a preset similarity threshold, the vector corresponding to the cosine similarity score is intercepted to obtain the filtered target design requirement information.

3. The model digital generative design method based on continuous implicit functions according to claim 1, characterized in that, The step of obtaining a target 3D geometric model of the jewelry containing physical material parameters by automatically matching the material and color of the 3D geometric model of the jewelry includes: Based on the geometric characteristics of the jewelry's three-dimensional geometric model, the corresponding material parameters and color information are matched from a preset material database; Using a micro-surface reflection model, based on the material parameters and color information, the material and color configurations of the jewelry three-dimensional geometric model are performed to obtain the jewelry target three-dimensional geometric model containing physical material parameters.

4. The model digital generative design method based on continuous implicit functions according to claim 1, characterized in that, The step of obtaining a multi-angle high-definition rendering image of the jewelry by performing physical rendering and image enhancement processing on the three-dimensional geometric model of the jewelry target includes: A path-tracing-based global illumination rendering model is used to physically render the three-dimensional geometric model of the jewelry target by simulating the reflection, refraction, and scattering of light in the scene. The rendered image is then enhanced to obtain a multi-angle high-definition rendering image of the jewelry.

5. A model digital generative design device based on continuous implicit functions, characterized in that, The device includes: The image generation module performs copyright and trademark similarity detection processing based on the design requirements information of the jewelry to be rendered, obtains the target design requirements information, and sends the target design requirements information to the conditional diffusion model to generate a two-dimensional concept map of the jewelry to be rendered. The implicit 3D reconstruction module performs implicit 3D reconstruction processing on the 2D concept map to obtain a 3D geometric model of the jewelry. Then, by performing automatic matching processing on the material and color of the 3D geometric model of the jewelry, a 3D geometric model of the jewelry target containing physical material parameters is obtained. The model rendering module performs physical rendering and image enhancement processing on the three-dimensional geometric model of the jewelry target to obtain a multi-angle high-definition rendering image of the jewelry, and sends the three-dimensional geometric model of the jewelry target and the multi-angle high-definition rendering image as the target output to the user terminal. The step of sending the target design requirement information to the conditional diffusion model to generate a two-dimensional concept map of the jewelry to be rendered includes: mapping the target design requirement information into a semantic conditional vector through a text encoder, and fusing the semantic conditional vector with a noise vector to obtain a latent vector; sending the latent vector to the conditional diffusion model for reverse denoising processing to obtain the two-dimensional concept map. The step of performing implicit 3D reconstruction processing on the 2D concept map to obtain a 3D geometric model of jewelry includes: performing feature extraction processing on the 2D concept map to obtain image feature vectors, and sending the image feature vectors and a preset set of sampling points to an implicit neural field model to obtain a signed distance function corresponding to each sampling point; determining the distance information between each sampling point and the 3D geometric model of jewelry to be generated based on the signed distance function, and using the distance information to extract zero-value isosurfaces to generate the 3D geometric model of jewelry. The implicit neural field model is used to provide a coordinate set that completely encloses the model to be generated, and to generate signed distance values ​​for each coordinate point. Based on the signed distance values, it determines whether each coordinate point is outside, inside, or on the surface of the model to be generated, so as to reconstruct a fully enclosed symmetrical 3D geometric model of jewelry. The step after generating the three-dimensional geometric model of the jewelry includes: performing a manufacturing inspection process on the three-dimensional geometric model of the jewelry; if there are problem areas in the three-dimensional geometric model of the jewelry that do not meet the preset manufacturing constraints, then performing a local geometric correction process on the problem areas.

6. A server, characterized in that, The method includes a processor and a memory, the memory storing computer-executable instructions executable by the processor, the processor executing the computer-executable instructions to implement the method of any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions that, when invoked and executed by a processor, cause the processor to perform the method described in any one of claims 1 to 4.

Images