A virtual article rapid generation method for three-dimensional printing of a text and creative product

Abstract

The application discloses a kind of virtual goods fast generation methods for three-dimensional printing of creative products.It converts user needs into computer instructions, including creative style, cultural elements or traditional pattern features, using generative AI models, image segmentation techniques and advanced three-dimensional reconstruction techniques, ultimately outputting three-dimensional virtual goods models that combine geometric precision and texture details and meet the three-dimensional printing manufacturability requirements of creative products.The application innovatively realizes the intelligent conversion from text description to three-dimensional model, and through multi-stage AI collaborative processing including image generation, multi-view expansion, geometric feature extraction, implicit surface reconstruction and other steps, it can automatically retain and restore cultural ornamentation, ware proportion and other creative design features, significantly improving modeling efficiency, supporting real-time feedback adjustment, and effectively solving the pain points of efficiency bottleneck, high professional dependence and high iteration cost in traditional three-dimensional modeling.

Description

Technical Field

[0001] This invention relates to a method for rapidly generating virtual objects for 3D printing of cultural and creative products. Specifically, it relates to a method that utilizes generative AI models, image segmentation technology, and advanced 3D reconstruction technology to automatically generate 3D virtual object models that can be used for the design and 3D printing of cultural and creative products, starting from the user's input of cultural and creative design requirements. Background Technology

[0002] With the booming development of the metaverse, virtual reality (VR), augmented reality (AR), digital twins, and the gaming and entertainment industry, the demand for high-quality, diverse 3D virtual items (such as game props, virtual goods, scene components, and digital human assets) is exploding. However, the traditional virtual item creation process heavily relies on the manual labor of professional 3D modelers. Designers need to use complex modeling software (such as Maya, Blender, and 3ds Max) to perform a series of tedious steps, including geometry construction, UV unwrapping, texture painting, material setting, rigging, and animation. This process is not only time-consuming, labor-intensive, and costly, but also requires highly specialized skills, making it difficult to meet the market's demand for rapid content iteration, large-scale customization, and personalized innovation. Long modification feedback cycles and high communication costs severely restrict the enrichment and development speed of the virtual content ecosystem.

[0003] In recent years, artificial intelligence, especially deep learning, has made significant progress in image generation and understanding. For example, text-to-image generation models, such as Stable Diffusion, can quickly generate creative and detailed 2D images based on natural language descriptions, greatly improving the efficiency of conceptual design. Meanwhile, image segmentation techniques (such as DeepLabv3+) can accurately identify and extract target objects from complex backgrounds. Furthermore, techniques for reconstructing 3D models from 2D images are constantly evolving, such as those utilizing multi-view geometry, neural radiation fields (NeRF), and symbolic distance fields (SDF).

[0004] While existing technologies offer individual AI tools for image generation, segmentation, and 3D reconstruction, effectively integrating these advanced technologies into an end-to-end automated and efficient process that directly translates user textual requirements into high-quality, usable 3D virtual objects remains a challenge. In particular, ensuring the accurate and efficient transformation of AI-generated initial concept maps into 3D models with precise geometry and surface details, and addressing issues such as background removal, multi-view information acquisition, and geometric and texture detail restoration, are critical challenges that current technologies urgently need to overcome. The lack of such an integrated process prevents the full potential of AI technology from being realized in the virtual object production stage, and the bottlenecks of traditional methods persist.

[0005] This problem is even more pronounced in the design of cultural and creative products and in 3D printing applications. Cultural and creative products often need to embody specific cultural symbols, patterns, and form characteristics, while simultaneously requiring rapid iteration and multiple version generation based on personalized user needs. Traditional manual modeling methods struggle to produce a large number of detailed, stylistically consistent, and printable 3D models within a limited timeframe. Furthermore, existing AI image generation, segmentation, and 3D reconstruction tools lack a coherent automated workflow, failing to guarantee that the generated results both conform to the cultural and creative style and meet the manufacturability requirements of 3D printing. Therefore, there is an urgent need for a technological solution that can efficiently generate 3D virtual objects with geometric accuracy, textural detail, and printability, starting from the user's cultural and stylistic needs. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of existing technologies in manually creating virtual items, such as low efficiency, high cost, long cycle, reliance on professional skills, and difficulty in quickly responding to personalized needs. It provides a method for rapidly generating virtual items for 3D printing of cultural and creative products. This method is a fast, automated, and low-cost method for generating virtual items based on AI image processing and 3D reconstruction. It is especially suitable for application scenarios that require rapid creative iteration and stylized expression, such as cultural and creative products, digital cultural heritage assets, and cultural IP derivatives.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for rapidly generating virtual items for 3D printing of cultural and creative products includes the following steps: S1: Transform the customer's needs into computer-understandable instructions to drive image generation and generate an initial image containing the target object and background. The needs include cultural and creative styles, cultural elements, decorative preferences or object shape characteristics, such as cultural and creative style descriptions, regional cultural symbols, traditional pattern characteristics or cultural IP visual elements. S2: Segment the target object from the initial image and remove the background; then, apply the Cross-Domain Diffusion algorithm to generate a multi-view image of the target object; next, extract key geometric and texture features from the multi-view image, and then map the key geometric and texture features to the normal map domain to generate the normal map of the target object. The key geometric and textural features include: cultural pattern features, used to characterize the distribution and basic outline of decorative patterns on the surface of the target object; shape proportion features, used to characterize the relative size relationship between the components in the overall shape of the target object; and traditional decorative morphology features, used to characterize the geometric outline of traditional decorative elements on the surface of the target object and their corresponding surface structure features. S3: Apply Neural Implicit Signed Distance Field (SDF) technology, combining the multi-view image and the normal map, to calculate the signed distance field of the target object's surface; generate a 3D mesh representation of the target object based on the signed distance field, completing the generation of the 3D virtual item and meeting the manufacturability requirements for 3D printing. The model closure ensures that the generated 3D mesh forms a closed solid structure without openings in its topology; the minimum wall thickness constraint ensures that the generated 3D mesh has a continuous solid thickness in each local region that meets the stability and structural strength requirements for 3D printing.

[0008] In step S1, image generation is driven by Stable Diffusion technology.

[0009] Image generation is driven by Stable Diffusion technology, which includes: transforming the image from a structured state to an unstructured state by gradually adding Gaussian noise (forward process), and then gradually denoising the noisy image to generate the target image by learning (reverse process), finally obtaining an initial image containing the target object and the background.

[0010] In step S2, the target object is segmented from the initial image and the background is removed, specifically including: The DeepLabv3+ image segmentation model is used to identify and extract the target object region in the initial image. The background region outside the target object region is removed or made transparent, and the cultural pattern details on the surface of the target object are preserved during the segmentation process to avoid the loss of patterns.

[0011] In step S3, the neural implicit symbolic distance field technique is applied, combining the multi-view image and the normal map, to calculate the symbolic distance field of the target object's surface, specifically including: Using multi-view images and normal maps as input, a neural network learns a function that outputs the minimum signed distance from any point in space to the object's surface. The normal map provides surface orientation constraints, and the multi-view images provide visual consistency constraints from multiple angles, thus completing the calculation of the signed distance field of the target object's surface.

[0012] In step S3, a three-dimensional mesh representation of the target object is generated based on the symbolic distance field and using the Marching Cubes algorithm to complete the generation of the three-dimensional virtual item, while preserving the surface geometric details corresponding to the cultural patterns.

[0013] Compared with the prior art, the present invention has the following advantages: I. High efficiency: The entire process of this invention is highly automated. The time from textual requirements to the generation of a 3D model is greatly shortened compared to traditional manual modeling. It can achieve rapid output in minutes or hours, and is especially suitable for cultural and creative design scenarios with high-frequency iteration.

[0014] II. Low cost: This invention significantly reduces reliance on expensive professional modeling manpower and time, thereby lowering the production cost of virtual items.

[0015] III. Ease of use and flexibility: In this invention, users only need to provide a text description or simple input to start the process, and the design can be iterated quickly by adjusting the input. The method is not limited to a specific object type, has good generalization ability, and can flexibly support the rapid generation of variations of different cultural styles, art schools, and thematic IPs.

[0016] IV. High quality: This invention combines advanced AI generation technology (ensuring the quality of the initial concept), precise segmentation technology, multi-view information (ensuring the integrity of the 3D structure), normal mapping (ensuring surface details), and powerful SDF reconstruction technology (ensuring geometric accuracy) to generate high-quality 3D models. This invention provides a complete and coherent technical chain from concept generation to 3D physicalization, solving the problems of fragmented and difficult integration of various stages in existing technologies, and significantly simplifying and accelerating the production process of cultural and creative products from concept to 3D finished product. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the traditional 3D modeling service process using existing technologies.

[0018] Figure 2 This is a schematic diagram of the innovative service process based on AI modeling proposed in this invention (taking the "Lingxi Creation Innovative Service Process" as an example).

[0019] Figure 3 This is a schematic diagram of the core AI modeling technology process in the method of this invention.

[0020] Figure 4 It is a three-dimensional model of a square wooden treasure chest generated by the method of the present invention based on provided prompts, as well as multi-view images and surface texture details observed from different preset virtual camera angles.

[0021] Figure 5 This is the operation page used in the method of the present invention to generate three-dimensional cultural and creative product models, which includes multiple functions such as input, viewing, and selection.

[0022] Figure 6 This is a schematic diagram of the three-dimensional model results of various cultural and creative products generated by the method of this invention.

[0023] Figure 7 This is a schematic diagram of the three-dimensional model results generated by the method of the present invention for different object shapes and texture features. It is used to show the effect of different surface texture features in the three-dimensional model while maintaining the consistency of the object shape proportions. Detailed Implementation

[0024] A schematic diagram of the traditional 3D modeling service process using existing technologies is shown below. Figure 1 As shown, the interaction process between the client, account manager, and modeler is illustrated, highlighting its multiple steps, long communication chain, and reliance on manual modeling.

[0025] This invention discloses a method for rapidly generating virtual objects based on AI image processing and 3D reconstruction, comprising the following steps: S1: Initial Image Generation: Transforms the user's specific requirements (e.g., text description of the item, style keywords, reference imagery, etc.) into computer-understandable instructions (e.g., optimized text hints), driving a powerful text-to-image generation model (e.g., Stable Diffusion) to generate an initial 2D image containing the target object and its environmental background.

[0026] S2: Object segmentation, multi-view generation, and normal map extraction: First, an advanced image segmentation model (e.g., DeepLabv3+) is applied to the initial image generated in step S1 to accurately identify and extract the pixel regions of the target object, generate an object mask, and separate the target object from the background (e.g., make the background transparent or solid color).

[0027] The segmented target object image is then input into a cross-domain diffusion algorithm model. This model is responsible for generating images of the target object from multiple different virtual camera perspectives (multi-view images) based on a single or a small number of view inputs.

[0028] Simultaneously or subsequently, using a cross-domain diffusion algorithm or a dedicated normal estimation network, key geometric surface details are extracted from the generated multi-view images, and this information is mapped to the normal map domain to generate normal maps corresponding to the multi-view images. Normal maps record the orientation of points on the object's surface and are crucial for accurate reconstruction.

[0029] S3: SDF-based 3D Reconstruction: The Neural Implicit Signed Distance Field (SDF) technique is applied, using the multi-view image set and corresponding normal map set generated in step S2 as input or supervision signals. A function is learned by training or optimizing a neural network (typically a multilayer perceptron MLP) that outputs the shortest signed distance from any point in space to the object's surface. The multi-view images provide global shape and appearance constraints, while the normal maps provide strong constraints on local surface orientation. Finally, from the learned SDF field, an explicit 3D surface representation (such as a triangular mesh or voxels) is extracted using an isosurface extraction algorithm (e.g., Marching Cubes), completing the 3D model generation of the virtual object.

[0030] The schematic diagram of the AI-based modeling-based innovative service process proposed in this invention (taking the "Lingxi Creation Innovative Service Process" as an example) is shown below. Figure 2 As shown, the process demonstrates the direct interaction between the client and the AI ​​modeler, highlighting the combination of AI-generated renderings for rapid feedback, AI-automated modeling, and human refinement, aiming to improve efficiency and shorten the cycle.

[0031] A schematic diagram of the core AI modeling technology process in the method of this invention. Figure 3 The diagram illustrates in detail the technical steps from the "requirements list" to the final textured 3D model ("Completed"). This process includes: translating the requirements into a model language (Prompt); generating a "final image" (typically a key view such as the front view) using diffusion-based modeling techniques (including a text encoder, multiple refinement / denoising steps, and possibly a VAE decoder); performing "multi-view generation" based on this image; reconstructing 3D using the multi-view information (illustrated as cube processing, representing implicit representations such as SDF or voxelization) and performing "mesh extraction"; and finally, performing "texture mapping" (possibly combined with information such as normal maps) to obtain the final textured model. The diagram also illustrates "manual optimization" (referring to adjustments made during the Prompt process).

[0032] The following is in conjunction with the attached diagram ( Figure 2 and Figure 3 The method of the present invention will be described in more detail below, but this is not a limitation of the present invention.

[0033] This invention proposes a method for rapid generation of virtual objects based on AI image processing and 3D reconstruction, referring to... Figure 2 The service process shown is as follows: Figure 3 The core AI technology process shown below has the following specific implementation steps: Step 1: Requirements Confirmation and Instruction Conversion (corresponding to) Figure 2The triggering of "Define Requirements" -> "AI Generate Renderings" and Figure 3 (Start) Customers submit specific requirements through a user interface (which may be part of a customer service platform), forming a "requirement list" (e.g., ...). Figure 3 (As shown). Requirements can include item description, style, color, key features, reference images, etc. For example, the generator prompt could be: "Generate a 3D rendering of a classic square wooden treasure chest, constructed of light brown polished planks, reinforced with dark gray metal corner protectors and black metal clasps, featuring realistic wood grain texture, soft shadows, a clean pure white background, cartoon game asset style, high detail, 8K resolution, cinematic lighting, no unnecessary elements, and centered composition."

[0034] AI modelers (or automated systems) translate clients' natural language requirements and reference information into a computer-understandable, optimized "model language" (e.g., ...). Figure 3 As shown), this is the text prompt used to drive the subsequent image generation model. This process may involve "human optimization" (such as...). Figure 3 (As shown), for example, adjusting keywords, weights, and style descriptors to ensure that the generated results better meet expectations.

[0035] Step 2: AI generates preliminary renderings (corresponding to...) Figure 2 "AI-generated renderings" and Figure 3 (First half) Use text-to-image generation techniques based on diffusion models, such as Stable Diffusion.

[0036] like Figure 3 As shown, the process typically begins with a random noisy image, inputting the "model language" generated in the previous step.

[0037] The model understands semantics through a "text encoder" and incorporates this understanding into the core denoising network (such as...). Figure 3 As shown in the multiple blue "refined" blocks (typically based on the U-Net architecture), noise is gradually removed based on text guidance to generate a latent image.

[0038] Furthermore, the latent space representation image is decoded into a pixel space image by a "VAE decoder", resulting in a preliminary "unrefined image" that may contain noise or lack of detail.

[0039] After further "refinement" or high-definition processing (Upscaling / Refinement), a high-quality "final image" (such as...) is generated. Figure 3(As shown). This "final image" is typically a key view (e.g., a front view) of the target object, presented to the client as a preliminary rendering. The square wooden treasure chest effect generated based on the above prompts is as follows: Figure 4 As shown.

[0040] Step 3: Review and revision of the renderings (corresponding to...) Figure 2 (The process of "Acceptance" -> "Does it meet the requirements?" -> "Change the rendering" repeats) The preliminary renderings generated in step two are then presented to the client for "acceptance" via the customer service platform (e.g., ...). Figure 2 (As shown).

[0041] If the client believes the rendering "meets their requirements" ( Figure 2 If the "Yes" branch is selected, proceed to the next step (Step 4).

[0042] If the customer requests modifications ( Figure 2 (For the "No" branch) AI modelers adjust the "model language" or use image editing techniques (such as image-to-image) based on feedback, returning to step two to regenerate or "change the rendering" (e.g. Figure 2 (As shown), resubmit to the customer for acceptance until the customer is satisfied. This rapid iterative cycle significantly improves communication efficiency.

[0043] Step 4: AI Modeling - From 2D to 3D (corresponding) Figure 2 "AI Modeling" and Figure 3 (Second half) After the final image is confirmed, the core AI 3D modeling process is executed, such as... Figure 3 The second half is shown below: Object segmentation (implicit step, preparing for subsequent processing): Although Figure 3 Not explicitly shown, but typically it's necessary to first segment the target object from the "final image" (if it includes a background). An image segmentation model (such as DeepLabv3+ mentioned earlier) is used to identify and extract the target object region, removing the background. Subsequent steps process the segmented target object.

[0044] Multi-view generation: Based on the confirmed, segmented "final image" of the target object, multi-view images of the object observed from different preset or inferred virtual camera angles are generated using cross-domain diffusion algorithms or other multi-view generation techniques (e.g., ...). Figure 3 (As shown in "Multi-view Generation"). A multi-view image of a square wooden treasure chest is shown below. Figure 4 As shown.

[0045] Normal map generation (implicit or integrated): To capture surface details, cross-domain diffusion algorithms (or dedicated normal estimation networks) can simultaneously or subsequently extract geometric features from multi-view images to generate corresponding normal maps (e.g., ...). Figure 3 The colored sphere entered in the "Texture Mapping" step may represent normal information.

[0046] 3D Reconstruction and Mesh Extraction: The Neural Implicit Signed Distance Field (SDF) technique is applied. The multi-view image set generated in step 4.2 and the normal map set generated in step 4.3 are used as input or supervision signals.

[0047] Training a neural network (e.g.) Figure 3 (The process is illustrated in the diagram of a cube.) The system learns the SDF representation of the object, which is the signed distance from each point in space to the object's surface. Multiple views provide shape constraints, and normal maps provide surface orientation constraints.

[0048] From the learned SDF field, algorithms such as Marching Cubes are used for "mesh extraction" (e.g., Figure 3 As shown, this generates a three-dimensional geometric mesh of the object.

[0049] Texture mapping: The color information from the original rendering or multi-view image, along with the generated normal map and possible roughness / metallization maps, are mapped onto the extracted 3D mesh (e.g., Figure 3 (As shown in "Texture Mapping"). This may involve UV unwrapping (automatic or semi-automatic) and texture baking processes.

[0050] The final result is a 3D model with complete geometry and surface texture details. The 3D texture image of the square wooden treasure chest is shown below. Figure 4 As shown.

[0051] Step 5: Artificial retouching (corresponding to) Figure 2 (Artificial retouching) While AI modeling can generate high-quality base models, there may still be some minor flaws or adjustments needed to meet specific artistic requirements.

[0052] After the AI ​​modeling is completed (i.e., after step four), a human artist or modeler will perform necessary "human touch-ups" (such as...). Figure 2 (As shown). This may include: fixing minor geometric errors, optimizing topology, adjusting texture details, enhancing material performance, adding specific decorations or effects, etc., to ensure that the final model meets the highest quality and realism requirements.

[0053] Step Six: Final Acceptance and Completion (corresponding to) Figure 2 (Acceptance -> Completion) The final 3D model, after manual refinement, is submitted to the client for final "acceptance" (e.g., ...). Figure 2 (As shown).

[0054] Once the client confirms that the model meets all requirements, the project is "completed" (e.g., ...). Figure 2 (As shown). The generated model can be delivered to the customer for use.

[0055] The results of various 3D models of cultural and creative products generated by the method of this invention are as follows: Figure 6 As shown, various cultural and creative products can be generated using the method of this invention.

[0056] The schematic diagrams of the three-dimensional model results generated by the method of this invention for different object shapes and texture features are shown below. Figure 7 As shown, this is used to demonstrate the effect of different surface texture features in a 3D model while maintaining the same proportions of the object.

[0057] Specific examples of generating 3D cultural and creative products The operation page for generating 3D cultural and creative products is as follows: Figure 5 As shown, the model is divided into three parts: left, middle, and right. Fill and select content sequentially from left to right to obtain the desired 3D model. The layout and filling requirements for the three parts are as follows: The left-hand area is primarily used to set the parameters of the generated model. Specifically, it includes the following sections: 1. Hint Words: In the "Hint Words" input box, users can enter descriptive text (up to 500 characters) related to the 3D model. Through detailed descriptions, users can specify the appearance, style, and purpose of the 3D model. To increase the weight, simply enclose the word in parentheses "()"; "()" defaults to a weight of 1. To increase the weight, simply multiply the parentheses by the weight coefficient, for example, "(flower) * 2". Hint words with higher weights have a stronger effect on model generation.

[0058] 2. Negative Hints: In the "Negative Hints" input box, users can enter features or unwanted attributes they wish to exclude, such as "ugly," "bad quality," etc. By setting negative hints, users can help the AI ​​avoid generating elements that do not meet expectations.

[0059] 3. Model Selection: Users can choose the AI ​​model used to generate the 3D model. In this example, options include "TIFA1.0 Turbo" and other model types. Different AI models will affect the generated detail and style.

[0060] 4. Style Settings: Users can choose the art style of the generated model, with options including "Realistic", "Cartoon", "Low Poly", etc., which will affect the visual effect and level of detail of the final generated model.

[0061] 5. Generation Options: Use User-Defined Seed: By enabling this option, users can provide a specific seed value to control the randomness of model generation, helping to reproduce a particular result.

[0062] After completing the input in the left-hand area, the middle area will display the real-time progress of model generation and a preview of the generated model. Specific features include: 1. Preview Generation: In this area, users can see a preview of the generated 3D model. This area displays a thumbnail of the model being generated, helping users to judge the quality of the generated model in real time. After confirming to proceed to the next step, the model will be generated. The completed image preview looks like this. Figure 4 As shown.

[0063] 2. Generation Progress: Users can view the generation progress, displaying "Estimated Time" and "Remaining Time" to help them understand the approximate time required for model generation. The generation process will update in real time until completion. A preview of the completed model is shown below. Figure 4 As shown.

[0064] After selecting the preview image in the middle area, the right-hand area will display the detailed rendering of the model, helping users to fully view the generated 3D model. Specific functions include: Model Viewing: The right-hand area displays the generated 3D model, allowing users to view its detailed rendering effects. The model can be rotated and scaled, enabling users to examine its details from various angles. This allows for evaluation of the model, determining whether regeneration or localized optimization of model details is necessary.

Claims

1. A virtual goods rapid generation method for three-dimensional printing of creative product-oriented products, characterized in that, Includes the following steps: S1: Translates customer requirements into computer-understandable instructions, drives image generation, and generates an initial image containing the target object and background; S2: Segment the target object from the initial image and remove the background; Then, a cross-domain diffusion algorithm is applied to generate a multi-view image of the target object; next, key geometric and texture features are extracted from the multi-view image. The key geometric and texture features are then mapped onto the normal map domain to generate the normal map of the target object. S3: Apply neural implicit symbolic distance field technology, and combine the multi-view image and the normal map to calculate the symbolic distance field of the target object surface; A three-dimensional mesh representation of the target object is generated based on the symbolic distance field, thus completing the generation of a three-dimensional virtual item for 3D printing. 2.The virtual article rapid generation method for three-dimensional printing of a text and creative product according to claim 1, characterized in that, In step S1, the requirements include cultural and creative styles, cultural elements, decorative preferences, or vessel shape characteristics.

3. The method for rapid generation of virtual items for 3D printing of cultural and creative products according to claim 1, characterized in that, In step S1, the customer's requirements are translated into computer-understandable instructions to drive image generation, generating an initial image containing the target object and background, specifically including: Image generation is driven by Stable Diffusion technology. The Stable Diffusion model is guided by style cue words or cultural semantic vectors to generate an initial image that conforms to the cultural and creative style, that is, an initial image containing the target object and the background.

4. The method for rapid generation of virtual items for 3D printing of cultural and creative products according to claim 3, characterized in that, Image generation is driven by Stable Diffusion technology, which includes: gradually adding Gaussian noise to transform the image from a structured state to an unstructured state, and then learning to gradually denoise the noisy image to generate the target image, finally obtaining an initial image containing the target object and the background.

5. The method for rapid generation of virtual items for 3D printing of cultural and creative products according to claim 1, characterized in that, In step S2, the target object is segmented from the initial image and the background is removed, specifically including: The DeepLabv3+ image segmentation model is used to identify and extract the target object region in the initial image. The background region outside the target object region is removed or made transparent, and the cultural pattern details on the surface of the target object are preserved during the segmentation process to avoid the loss of patterns.

6. The method for rapid generation of virtual items for 3D printing of cultural and creative products according to claim 1, characterized in that, In step S2, the key geometric and texture features include: Cultural pattern features are used to characterize the distribution and basic outline of decorative patterns on the surface of cultural and creative products on the surface of the target object. The proportional characteristics of the object shape are used to characterize the relative dimensional relationships between the various components of the overall shape of the target object. Traditional decorative morphological characteristics are used to characterize the geometric contours and corresponding surface structural features of traditional decorative elements on the surface of a target object.

7. The method for rapid generation of virtual items for 3D printing of cultural and creative products according to claim 1, characterized in that, In step S3, the neural implicit symbolic distance field technique is applied, combining the multi-view image and the normal map, to calculate the symbolic distance field of the target object's surface, specifically including: Using multi-view images and normal maps as input, a neural network learns a function that outputs the minimum signed distance from any point in space to the object's surface. The normal map provides surface orientation constraints, and the multi-view images provide visual consistency constraints from multiple angles, thus completing the calculation of the signed distance field of the target object's surface.

8. The method for rapid generation of virtual items for 3D printing of cultural and creative products according to claim 1, characterized in that, In step S3, a three-dimensional mesh representation of the target object is generated based on the symbolic distance field, specifically including: Based on the symbolic distance field and using the Marching Cubes algorithm, a three-dimensional mesh representation of the target object is generated, thus completing the generation of the three-dimensional virtual item while preserving the surface geometric details corresponding to the cultural patterns during the mesh generation process.

Images