Artificial Intelligence-Based Intelligent Color Matching Method, Device, and Server for 3D Models

Abstract

This invention provides an intelligent color matching method, device, and server for 3D models based on artificial intelligence, relating to the technical field of intelligent industrial design. The method includes: acquiring an original 3D model and user requirement information; rendering and extracting features from the original 3D model to obtain a high-resolution snapshot and model feature information; using a deep learning model based on semantic segmentation, performing background filtering on the high-resolution snapshot to remove non-model regions at the pixel level, obtaining an input image; and performing illumination normalization and distortion detection on the input image to determine the target input image; and sending the target input image, model feature information, and user requirement information to a preset intelligent color matching model for intelligent color matching processing to obtain the target color scheme. This invention can significantly improve the efficiency of model color matching and the accuracy of color schemes.

Description

Technical Field

[0001] This invention relates to the technical field of intelligent industrial design, and in particular to an intelligent color matching method, device and server for three-dimensional models based on artificial intelligence. Background Technology

[0002] In the field of industrial design, color scheme design is a crucial element influencing user perception and market competitiveness. Whether it's consumer electronics, home appliances, furniture, or transportation, color schemes not only determine the product's aesthetics but also directly impact brand image and user purchasing decisions. Currently, relevant technologies suggest that traditional color matching methods mainly fall into two categories: one is based on physical samples, where the final color scheme is determined through repeated color adjustments and comparisons on existing physical objects; the other is based on two-dimensional design drafts, utilizing 2D images or renderings to complete the color design. However, both of these approaches have significant limitations. Color matching based on physical samples is time-consuming and labor-intensive, requires substantial material support, and has low iteration efficiency; while color matching based on 2D design drafts is more intuitive, it still differs significantly from the final 3D product, making it difficult to accurately reflect the color effect in real space. Summary of the Invention

[0003] In view of this, the purpose of this invention is to provide an intelligent color matching method, device and server for 3D models based on artificial intelligence, which can significantly improve the efficiency of model color matching and the accuracy of color matching schemes.

[0004] In a first aspect, embodiments of the present invention provide an intelligent color matching method for 3D models based on artificial intelligence. The method includes: acquiring an original 3D model and user requirement information, and performing rendering and feature extraction processing on the original 3D model to obtain a high-resolution snapshot of the model and model feature information; using a deep learning model based on semantic segmentation, performing background filtering processing on the high-resolution snapshot of the model, removing non-model areas in the high-resolution snapshot of the model at the pixel level to obtain an input image, and performing illumination normalization processing and distortion detection processing on the input image to determine a target input image; sending the target input image, model feature information, and user requirement information to a preset intelligent color matching model for intelligent color matching processing to obtain a target color matching scheme.

[0005] In one embodiment, the steps of rendering and feature extraction of the original 3D model to obtain a high-resolution snapshot of the model and model feature information include: obtaining a high-resolution snapshot of the model by performing orthographic projection and perspective projection on the original 3D model, and obtaining model feature information by performing feature extraction on the original 3D model.

[0006] In one implementation, prior to the step of obtaining a high-resolution snapshot of the model by performing orthogonal projection and perspective projection on the original 3D model, the method includes: driving a virtual camera to rotate orbitally around a model reference point of the original 3D model based on the interaction information between the user terminal and the view cube component to switch the standard viewpoint, and determining the target viewpoint by performing multi-angle observation of the model to obtain a high-resolution snapshot of the model under the target viewpoint.

[0007] In one implementation, the steps of performing illumination normalization and distortion detection processing on the input image to determine the target input image include: performing intelligent lighting and shadow processing on the input image based on an improved lighting model; reshaping the main light source and ambient light of the input image through lighting effect corpus control and a physical lighting rendering engine to normalize the illumination of the input image; and performing distortion detection processing on the light and shadow processed input image to determine the target input image.

[0008] In one implementation, the step of performing distortion detection processing on the input image after lighting and shadow processing to determine the target input image includes: performing distortion detection processing on the input image after lighting and shadow processing using a Hausdorff distance calculation model to determine the bidirectional Hausdorff distance; if the bidirectional Hausdorff distance exceeds a preset distance threshold, then re-rendering processing is performed; if the bidirectional Hausdorff distance does not exceed the preset distance threshold, then it is determined that the geometric features of the input image are consistent with those of the original 3D model, and the input image is determined as the target input image.

[0009] In one implementation, the step of sending the target input image, model feature information, and user requirement information to a preset intelligent color matching model for intelligent color matching processing to obtain the target color scheme includes: acquiring brand database corpus information and extracting color features from the brand style map in the brand database corpus information to obtain target color information; after locking the target color information, performing intelligent color matching processing on the target input image, model feature information, and user requirement information through the preset intelligent color matching model to obtain the target color scheme.

[0010] In one implementation, the step of obtaining target color information by extracting color features from the brand style map in the brand database corpus includes: performing cluster analysis on each pixel in the brand style map using the K-means clustering algorithm to obtain the pixel proportion of each cluster; determining the primary color and each secondary color, as well as the corresponding proportion information of the primary color and each secondary color, based on the pixel proportion of the clusters; and determining the primary color, each secondary color, and the proportion information as the target color information.

[0011] Secondly, embodiments of the present invention also provide an intelligent color matching method for 3D models based on artificial intelligence. The method includes: an information acquisition module, which acquires the original 3D model and user requirement information, and performs rendering and feature extraction processing on the original 3D model to obtain a high-resolution snapshot of the model and model feature information; an image processing module, which performs background filtering processing on the high-resolution snapshot of the model using a deep learning model based on semantic segmentation, removes non-model areas in the high-resolution snapshot of the model at the pixel level to obtain an input image, and performs illumination normalization processing and distortion detection processing on the input image to determine a target input image; and an intelligent color matching module, which sends the target input image, model feature information, and user requirement information to a preset intelligent color matching model for intelligent color matching processing to obtain a target color matching scheme, wherein the target color matching scheme includes the proportion and color code corresponding to each color.

[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 an intelligent color matching method, device, and server for 3D models based on artificial intelligence. The method acquires the original 3D model and user requirement information, performs rendering and feature extraction on the original 3D model to obtain a high-resolution snapshot and model feature information. Then, using a deep learning model based on semantic segmentation, background filtering is performed on the high-resolution snapshot, removing non-model areas at the pixel level to obtain the input image. The input image is then subjected to illumination normalization and distortion detection to determine the target input image. Finally, the target input image, model feature information, and user requirement information are sent to a preset intelligent color matching model for intelligent color matching processing to obtain the target color scheme. This invention significantly improves model color matching efficiency.

[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 an intelligent color matching method for 3D models based on artificial intelligence, provided in an embodiment of the present invention;

[0020] Figure 2 A schematic diagram illustrating the specific process of an intelligent color matching method for 3D models based on artificial intelligence, provided in an embodiment of the present invention;

[0021] Figure 3 A schematic diagram of the structure of an artificial intelligence-based three-dimensional model intelligent color matching device 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, in the color matching process of industrial design, common solutions mainly rely on manual operation by designers. Designers usually complete the color matching and rendering of products manually based on their own professional experience. However, this method has obvious limitations: First, the quality of the design effect is highly dependent on the individual ability of the designer, and the overall level varies. Second, clients cannot view and freely adjust the product in the design drawings from multiple angles according to their own intentions, and can only rely on the fixed-angle renderings provided by the designer. In addition, the design cycle of manual color matching is long, often requiring multiple rounds of communication and repeated revisions, making it impossible to produce a large number of design drafts and color matching reports in a short period of time, which is difficult to meet the market's demand for rapid iteration and diversified designs.

[0025] On the other hand, with the development of artificial intelligence technology, some intelligent color matching software products based on image algorithms have emerged in the market. However, existing intelligent color matching tools generally perform color replacement or style transfer based on defined two-dimensional renderings and do not support direct operation in a three-dimensional model environment. This two-dimensional image-based approach also has significant drawbacks: firstly, the expression method of two-dimensional images is inefficient and cannot fully demonstrate the color effects of products under different lighting and angles from a spatial perspective; secondly, the input quality of two-dimensional renderings is difficult to standardize. If the input image lacks clarity or the angle is unreasonable, it will seriously affect the authenticity and applicability of the generated design, thereby weakening the reference value of the color matching scheme. Furthermore, it cannot automatically generate color matching schemes based on three-dimensional models, style setting information, and design drafts.

[0026] In recent years, advancements in GPU computing power and breakthroughs in deep learning algorithms have made image-based intelligent color matching possible. Enterprises are seeking data-driven solutions to combine brand color assets, user preference big data, and real-time rendering engines to achieve one-click generation, instant visualization, and mass production-ready intelligent color matching systems. However, existing technologies largely remain at the level of two-dimensional planes or color chart mapping, lacking end-to-end tools that directly affect 3D models, making it difficult to meet the stringent requirements of industrial design for the coupled evaluation of space, materials, and lighting. Therefore, there is an urgent need for a 3D model intelligent color matching system that can be embedded in mainstream 3D software, is compatible with multiple model formats, and supports AI semantic understanding and real-time rendering to shorten design verification cycles, reduce prototyping costs, and enhance product competitiveness. Based on this, the present invention provides an AI-based 3D model intelligent color matching method, device, and server, which can significantly improve model color matching efficiency and the accuracy of color schemes.

[0027] See Figure 1 The diagram shows a flowchart of an intelligent color matching method for 3D models based on artificial intelligence. The method mainly includes the following steps S102 to S106:

[0028] Step S102: Obtain the original 3D model and user requirement information, and perform rendering and feature extraction processing on the original 3D model to obtain a high-resolution snapshot of the model and model feature information.

[0029] In one implementation, the high-resolution snapshot of the original 3D model obtained through rendering mainly provides features such as pixel-level color and texture cues. The model feature information is obtained by directly extracting features from the original 3D model. This model feature information is mainly used to provide model geometric information such as face ID, normal vector, area, and material slots. Both types of information need to be sent to the intelligent model together during intelligent color matching to achieve accurate intelligent color matching.

[0030] Step S104: Using a semantic segmentation-based deep learning model, background filtering is performed on the high-resolution snapshot of the model. Non-model regions in the high-resolution snapshot of the model are removed at the pixel level to obtain the input image. Illumination normalization and distortion detection are then performed on the input image to determine the target input image.

[0031] In one implementation, background removal through semantic segmentation and illumination standardization are performed to ensure that the intelligent model can only see the product itself and can make color judgments under standard lighting conditions. For example, if the lighting and shadows on the kettle model are inconsistent, the intelligent model may mistakenly identify areas with inconsistent lighting as other materials. Distortion detection is mainly used to ensure that the geometric features of the target input image are consistent with those of the original 3D model. Only after passing distortion detection can a high-resolution snapshot of the model be allowed to be input into the intelligent model for training. Otherwise, projection distortion may be mistakenly sent to the AI ​​as a color boundary, resulting in color mismatch.

[0032] Step S106: Send the target input image, model feature information and user requirement information to the preset intelligent color matching model for intelligent color matching processing to obtain the target color matching scheme.

[0033] In one implementation, during intelligent color matching, the primary and secondary colors are automatically extracted and locked based on the input brand library style map and the user-selected interactive information. This step is equivalent to building a color constraint table, ensuring that the subsequent intelligent model color matching does not deviate from the boundaries between creativity and brand specifications. Then, the target input image, model feature information, and user requirement information are sent to the intelligent color matching model, and color matching is performed based on the above color constraints to complete the color matching. In addition, a detailed report containing process cards, color difference, and cost estimates can be automatically generated based on the above color matching process, thereby automating the entire process from requirements to production documents and significantly reducing time costs.

[0034] The artificial intelligence-based intelligent color matching method for 3D models provided in this embodiment of the invention can significantly improve the efficiency of model color matching and the accuracy of color matching schemes.

[0035] The entire AI algorithm technology roadmap, based on brand, material, process, and lighting configuration libraries, effectively improves the quality of intelligent color matching design drafts through cross-checking and cyclical association. It also empowers the color matching system to automatically adapt to accurate AI corpus through simple user interaction. (See [link to relevant documentation]). Figure 2 The diagram illustrates a specific process for an AI-based intelligent color matching method for 3D models, which is divided into offline and online states:

[0036] The offline component refers to the construction of the brand library and report template library, which can be expanded and arranged without interaction, representing an independent information construction mode. AI corpus for the brand library and reports is built by retrieving and extracting popular styles and colors of popular brands in various scenarios. The online component requires user interaction to obtain information. When a user starts intelligent color matching, they first need to upload a 3D model file. The system then parses the 3D model file and extracts feature data, while simultaneously collecting data based on the set perspective and view format (orthogonal or perspective view). After this step, the data is refined in two dimensions: one is collecting additional user requirement text / reference images (including basic color schemes, processes, materials, and other effect settings corresponding to the brand), and using this information to retrieve relevant brand library data for intelligent creative generation; the other is data optimization. First, the basic white model image undergoes intelligent lighting and shadow processing and distortion detection until the data meets the standard requirements. Then, intelligent color matching is performed using the aforementioned intelligent creative generation corpus, and the corresponding design drawing is output based on the image's basic information requirements (clarity, quantity, and proportion). This invention also provides an implementation method for intelligent color matching of 3D models using artificial intelligence, as detailed in (1) to (3) below:

[0037] (1) Import and unified rendering of 3D models. The specific solution includes: model file upload, format conversion and compression, web-based interpretation and rendering, and conversion and interaction functions for orthogonal and perspective perspectives. Users upload 3D model files through the web interface. Supported file formats include STEP, IGES, STL and other common industrial design formats. The system's cloud server calls the open-source interface based on FreeCAD to parse and extract geometric information from multiple source files (geometric and topological extraction, unit standardization and coordinate system normalization, etc.). Then, with the help of a general open-source 3D database, vertex coordinates, normal vectors, texture coordinates, index faces, material properties and texture resources are organized into a standardized binary buffer and packaged together with the scene hierarchy and resource index into a .zip binary data package, thus forming a unified data exchange standard. After receiving the data, the web client first decompresses and interprets the .zip package to restore the standardized 3D data structure, and then passes it to the WebGL-based rendering engine for real-time rendering and color effect display.

[0038] In one implementation, a high-resolution snapshot of the model can be obtained by performing orthographic and perspective projection processing on the original 3D model. Feature extraction processing can then be performed on the original 3D model to obtain its feature information. Specifically, during the rendering stage, the system provides both orthographic and perspective projection methods, allowing users to switch between them according to design requirements. (Orthographic projection matrix) Defined as:

[0039]

[0040] in, Indicates the left boundary of the view frustum. Indicates the right boundary. Indicates the bottom boundary. Indicates the top boundary. This represents the distance from the near plane to the camera. The distance from the far plane to the camera must satisfy the following conditions: Orthographic projection can maintain the geometric proportions of a model without distortion and is often used in engineering drawings and color matching verification.

[0041] Perspective projection matrix Defined as:

[0042]

[0043] The parameters have the following meanings: The vertical viewpoint of the camera (in radians). The viewport aspect ratio is the ratio of the window's width to its height. The distance from the near plane to the camera. Given the distance from the far plane to the camera, the required distance is... This projection matrix can achieve the effect of objects appearing smaller when farther away and larger when closer, as seen by the human eye in natural observation, significantly enhancing the spatial sense and realism of the color scheme.

[0044] Furthermore, based on the interaction information between the user and the view cube component, the virtual camera is driven to rotate orbitally around the model reference point of the original 3D model to switch the standard viewpoint. By performing multi-angle observation of the model, the target viewpoint is determined to obtain a high-resolution snapshot of the model from the target viewpoint. Specifically, in terms of interaction logic, the system provides the following operation mechanism based on mouse events:

[0045] When the user presses and drags the left mouse button, the system drives the virtual camera to rotate around the model's reference point, enabling multi-angle observation of the model. When the user presses and drags the right mouse button, the system drives the camera to translate on a plane parallel to the screen, thus adjusting the model's spatial position. When the user scrolls the mouse wheel, the system adjusts the distance between the camera and the model's center to perform zoom operations. These interactive logics ensure that users can intuitively control the viewpoint and observation position, providing strong operability for comparing and evaluating color scheme effects. After the 3D model is imported, the system provides a polyhedral cube (viewpoint cube) in the lower right corner for quick switching of standard viewpoints. The six faces of this cube correspond to the front, back, left, right, top, and bottom standard orthogonal viewpoints, while the eight corners and edges correspond to commonly used oblique or isometric viewpoints. When the user clicks on a face, edge, or corner of the cube, the system automatically calculates the view matrix based on the corresponding preset camera direction vector, smoothly updating the camera position and orientation to align the model with the target viewpoint.

[0046] (2) Background removal and distortion detection. In the specific implementation process, the backend not only relies on the two-dimensional projection results of the 3D model snapshot, but also combines the geometric feature data of the 3D model itself (such as volume feature set, surface feature set, normal vector set and material mapping information) to construct the input parameters of the large artificial intelligence model. This ensures the intuitiveness of the image while improving the structured and semantic quality of the input data. The system first generates a high-resolution snapshot during the model rendering stage and extracts the model feature information simultaneously, encapsulating it in JSON format. This allows the large model to not only rely on the image pixel features during inference, but also to use geometric information for morphological constraints. Subsequently, the backend performs a background filtering step, using a deep learning model based on semantic segmentation to remove non-model areas in the snapshot image at the pixel level, ensuring that the input image retains only the significant features related to the product.

[0047] Furthermore, after background filtering, the system enters the intelligent lighting and shadow processing stage. This stage, based on lighting effect corpus control and a physically based lighting rendering engine, recolors the main light source and ambient light to ensure that the light and shadow distribution conforms to the realism of industrial design. This process requires intelligent lighting and shadow processing of the input image based on an improved lighting model. Through lighting effect corpus control and a physically based lighting rendering engine, the main light source and ambient light of the input image are recolored to standardize the lighting of the input image. Specifically, this controls the lighting of the main light source and ambient light to be consistent, avoiding color misjudgment caused by arbitrary lighting. Here, a lighting formula based on the improved Blinn–Phong model is used:

[0048]

[0049] in, Represents points on the model surface The final brightness value (which can be expanded into an RGB three-channel vector). For ambient light intensity, The environmental reflectance coefficient of the material; and The first Diffuse reflection and high light intensity of a light source , These are the diffuse reflection and specular reflection coefficients of the material, respectively; For point The unit normal vector; To point to the first Unit vector of each light source; For half-range vectors, The unit vector in the direction of observation; The highlight index controls the sharpness of the highlight areas; This represents the number of light sources. Using this formula, snapshots can maintain a consistent visual context under different lighting conditions, avoiding AI misjudgments of color and material properties.

[0050] After completing the lighting and shadow processing, the system further performs distortion detection. It performs distortion detection processing on the input image after lighting and shadow processing to determine the target input image, so as to ensure the geometric consistency between the two-dimensional snapshot and the original three-dimensional model and avoid the deviation of the design draft outline due to projection or rendering errors.

[0051] In one implementation, distortion detection processing can be performed on the input image after lighting and shadow processing using the Hausdorff distance calculation model to determine the bidirectional Hausdorff distance: if the bidirectional Hausdorff distance exceeds a preset distance threshold, the rendering process is repeated; if the bidirectional Hausdorff distance does not exceed the preset distance threshold, it is determined that the geometric features of the input image are consistent with those of the original 3D model, and the input image is identified as the target input image. The Hausdorff distance formula used in the above process has the following core calculation formula:

[0052]

[0053] in, This represents the set of two-dimensional contour points extracted from a snapshot image. This is the theoretical contour point set obtained from the original 3D model under the same projection matrix; Indicates Euclidean distance; , Let these represent the supremum and infimum, respectively. The bidirectional Hausdorff distance is calculated. This can measure the maximum deviation between the image contour and the projection of the 3D model. If Exceeding the threshold The system will readjust the rendering or lighting parameters until they meet the requirements. ≤ε. Through two steps of intelligent lighting and distortion detection, the system can ensure that the input data fed into the AI ​​model is strictly aligned with the original model in terms of lighting, contour, and geometric semantics. Combined with user needs and brand database information, the final output is a high-precision color matching result with realistic lighting and shadow effects.

[0054] (3) In the intelligent color matching process of the system, the brand library provides mainstream style data of different brands. Each brand style corresponds to a set of basic primary and secondary colors. These primary and secondary colors are not set manually, but are obtained by extracting color features from typical brand style images in the brand library. In other words, the brand library corpus information can be obtained, and the target color information can be obtained by extracting color features from the brand style images in the brand library corpus information. After locking the target color information, the target input image, model feature information and user demand information are intelligently matched using a preset intelligent color matching model to obtain the target color matching scheme. In one implementation, the K-means clustering algorithm can be used to perform cluster analysis on the pixels in the brand style image to obtain the pixel ratio of each cluster. The primary color and each secondary color, as well as the ratio information corresponding to the primary color and each secondary color, are determined based on the pixel ratio of the clusters. The primary color, each secondary color and the ratio information are then determined as the target color information.

[0055] Specifically, the input style image is first converted from the RGB color space to the CIELab color space to enhance the balance of color brightness and perception. Then, the K-means clustering algorithm is used to cluster the pixels in the image. By calculating the pixel proportion of each cluster, the system can automatically identify the primary color (the cluster center with the largest proportion) and secondary colors (the remaining cluster centers), obtaining several representative colors as candidate primary and secondary colors. Based on the pixel proportion of each cluster, the system automatically identifies the color with the highest proportion as the primary color (default is one), and the rest as secondary colors (default is two), while recording the proportion information of each color to ensure that the output color scheme is consistent with the brand style.

[0056] In the subsequent automated report generation process, the system will further extract and label the colors used in the design drawings, forming a quantitative description of color proportions and color codes. Specifically, it will perform pixel traversal on the rendered design drawings, calculate the pixel percentage of each extracted color in the entire image, and generate a color proportion distribution histogram. Simultaneously, the system will call a standard color library for color mapping, performing nearest neighbor matching between the extracted cluster center color vectors and predefined colors in the standard library, and outputting the corresponding color code. For example, based on RGB values, the system will perform further RGB-to-CMYK conversion to obtain standard color codes suitable for industrial design.

[0057] The CMYK conversion formula is as follows, where C, M, Y, and K represent the cyan, magenta, yellow, and black components, respectively.

[0058] First, normalize the RGB values:

[0059]

[0060] in, These are the original pixel values. This is the normalized value.

[0061] Calculation of black component K:

[0062]

[0063] Calculation of other components:

[0064]

[0065] Each color is labeled with its proportion and corresponding CMYK code. The color codes between CMYK and Pantone can be converted again using CMYK's own open-source interface for automated report generation. Ultimately, the report will automatically generate a "Primary and Secondary Color Composition Table," a "Color Proportion Pie Chart," and "Standard Color Code Matching Results," providing designers and brands with intuitive and quantifiable color matching guidelines, ensuring the consistency and traceability of design drafts with brand tone.

[0066] In this way, the system not only achieves the extraction of primary and secondary colors based on the brand style map, but also combines color ratios and standard color systems (CMYK / Pantone) to achieve high-precision color annotation and output of design drafts.

[0067] In the system's color matching workflow, in addition to intelligent extraction of primary and secondary colors and standardized output in CMYK / Pantone format, a color matching locking mechanism is introduced to ensure that users can flexibly control the stability and divergence of color matching at different design stages. When the user completes the retrieval of basic colors from the brand library and the algorithm automatically identifies the ratio of primary to secondary colors (e.g., primary color 60%, secondary color 40%), the system provides two options: "color matching locking mode" and "free divergence mode." In color matching locking mode, the system strictly constrains the color allocation in all subsequent design drafts within the set ratio and color values. For example, when multiple areas in the design need to be colored, the algorithm uses a partitioned weighting method to ensure that the primary color is always controlled within the specified percentage range, while supplementing the secondary color, thereby avoiding deviations in the brand color style caused by manual operation or AI divergence calculations. By constraining the color distribution matrix, the color value of each pixel in the matrix is ​​weighted by the locking ratio in the scheduling function to ensure that the output color distribution conforms to the settings.

[0068] Conversely, the free-flowing design mode is more suitable for exploratory design scenarios. When users do not lock the color scheme, the system will derive and expand upon the brand's main color scheme. Through this dual-mode design, the system can balance stability and creativity: the locked color scheme mode ensures the consistency and standardization of the brand's visual identity, while the free-flowing design mode provides designers or AI-generated systems with more room for exploration, making it suitable for use when multiple solutions need to be generated quickly and creative comparisons can be made.

[0069] The system can automatically generate a complete process including brand style AI corpus, 3D model renderings, 2D design drafts, overlaid scene renderings, and color extraction and color schemes. Its core value lies in achieving consistency between brand visuals and product design. In this process, the brand database corpus first provides semantic constraints for the system, ensuring that the generation of 3D models and 2D design drafts always revolves around the brand's core style and theme. Subsequently, through semantic matching and style detection, the system automatically selects background images from the scene image database that best match the product design style, color scheme, and brand theme. For example, technology brands are matched with futuristic urban night scenes, while nature and environmental protection brands are matched with green vegetation or natural light environments. During the overlay process, the system performs semantic alignment, geometric perspective matching, and light and shadow layering on the product images and scene images to ensure a seamless visual experience.

[0070] The most crucial step is color consistency correction, ensuring that the product's color scheme and the background scene are harmonious and unified in color gamut. The correction steps include: converting the scene image and product color scheme from RGB space to CIELab space, then comparing their mean and standard deviation, performing statistical matching to bring the overall scene tone closer to the brand's primary and secondary colors. The formula for calculating the primary and secondary color difference in the CIELab color space is:

[0071]

[0072] in( , , This indicates the Lab value of the product's primary color. , , () indicates the average color of the scene. If (Threshold) The system will adjust it through the color mapping function. Taking into account the rendering rate, the threshold here is set in the range of 2 to 5, which means that it is perceptible to the average person in a normal environment, but is still considered acceptable.

[0073] To support the rapid implementation of industrial color scheme design across industries and scenarios, five core capabilities—brand library, material library, process library, scene library, and report library—are abstracted into independent and pluggable microservice components, forming a "modular component-based rapid development platform." The brand library component incorporates mainstream brand visual assets, supports automatic extraction of primary and secondary colors through K-means clustering, and provides interfaces to return CMYK and Pantone color values. The material library component encapsulates material parameters for metals, fabrics, and plastics. The process library component digitizes surface processes such as spraying, IMD, and hot stamping into process cards. The scene library component includes scene templates that designers can drag and drop to replace with a single click. The report library component automatically injects AI-extracted primary and secondary colors, materials, processes, and scene renderings into defined report templates, achieving one-click rendering in seconds. All components communicate via a unified event bus: after the brand library publishes a "primary color update" event, the material library automatically filters matching materials, the process library synchronously marks compatible processes, the scene library immediately refreshes its rendering results, and the report library generates a new PDF in real time. Developers can assemble customized color matching tools for specific industries within hours simply by selecting the required components and configuring parameters on the platform. This significantly shortens the iteration cycle from requirements to product, reduces secondary development costs, and truly achieves rapid expansion in a modular fashion.

[0074] In practical applications, taking the intelligent color matching scenario in the cup and kettle industry as an example, the first step in the entire implementation process is on the model upload page of the cloud platform. At this stage, users need to upload the 3D model file of the cup or kettle to the system. By uploading the file, the system can obtain complete product structure and appearance information, thus laying the foundation for subsequent material selection, intelligent color matching, and process settings. This step is not only the starting point for data input but also a crucial step in realizing personalized design and brand style presentation, ensuring that the subsequently generated design scheme can highly match the actual product.

[0075] After the model is uploaded, the system will proceed to the model rendering stage. During this stage, users can freely switch between different viewing modes, including orthographic and perspective views, to more comprehensively observe the overall appearance and details of the cup and kettle model. At the same time, users can also adjust the angle of the model and confirm the view to ensure that every detail of the product is accurately displayed. After confirming the model view, the system will guide users to the intelligent color matching stage, where they can flexibly select and apply materials, colors, and styles.

[0076] Within the interactive panel, users can select a brand style based on their needs and further personalize settings such as materials, craftsmanship, and color schemes. Upon receiving these settings, the system automatically generates corresponding model design drafts, providing an intuitive visual presentation of the design process. This generation process not only demonstrates the efficiency and intelligence of the technology but also incorporates a deep product thinking process, integrating the brand's core concepts, design logic, and value proposition. By visualizing the brand analysis and design thinking process, users can more intuitively understand the cultural connotations and differentiating advantages behind the product, thereby achieving a high degree of unity between brand image and product design, enhancing the overall design value and communication effectiveness.

[0077] After the user completes the color scheme selection, the system automatically and intelligently integrates the primary and secondary colors with the set material and process parameters, as well as the rendered images, to generate a complete color scheme report. This report not only includes accurate color code information and color difference comparison data, but also includes high-quality design renderings that intuitively showcase the visual presentation of the final product. Simultaneously, the system combines brand analysis and product style to generate corresponding scenario-based display images, making the scheme more realistic and impactful. Finally, all content is compiled into a professional PDF report, which users can quickly download with one click for internal reviews, client presentations, or marketing promotions, significantly improving design communication efficiency and brand value expression.

[0078] In summary, this invention has the following advantages: 1. It imports basic AI input based on 3D models, with adjustable viewing angles and controllable input effects, thereby effectively improving the output effect of subsequent design drafts; 2. Based on AI algorithms using brand, material, process, and lighting configuration libraries, through cross-checking and cyclic association, it effectively improves the effect of intelligent color matching design drafts and empowers the color matching system to automatically adapt to accurate AI corpus through simple user interaction; 3. It can automatically extract color ratios and provide a color matching lock function; 4. It can automatically generate color schemes containing brand style AI corpus, 3D model renderings, design drawings, overlaid scene renderings, and color extraction; 5. It can form modular components, facilitating the expansion of color matching design tools in various industrial scenarios.

[0079] Regarding the AI-based intelligent color matching method for 3D models provided in the foregoing embodiments, this invention provides an AI-based intelligent color matching device for 3D models. (See attached document.) Figure 3 The diagram shows a structural schematic of an artificial intelligence-based 3D model intelligent color matching device, which includes the following parts:

[0080] The information acquisition module 302 acquires the original 3D model and user requirement information, and performs rendering and feature extraction processing on the original 3D model to obtain a high-resolution snapshot of the model and model feature information.

[0081] The image processing module 304 uses a semantic segmentation-based deep learning model to perform background filtering on the high-resolution snapshot of the model, removes non-model regions from the high-resolution snapshot of the model at the pixel level to obtain the input image, and performs illumination normalization and distortion detection on the input image to determine the target input image.

[0082] The intelligent color matching module 306 sends the target input image, model feature information and user requirement information to the preset intelligent color matching model for intelligent color matching processing to obtain the target color matching scheme, wherein the target color matching scheme includes the proportion and color code of each color.

[0083] The artificial intelligence-based intelligent color matching device for 3D models provided in this application embodiment can significantly improve the efficiency of model color matching and the accuracy of color matching schemes.

[0084] In one embodiment, when performing rendering and feature extraction processing on the original 3D model to obtain a high-resolution snapshot of the model and model feature information, the information acquisition module 302 is further configured to: obtain a high-resolution snapshot of the model by performing orthographic projection and perspective projection processing on the original 3D model, and obtain model feature information by performing feature extraction processing on the original 3D model.

[0085] In one embodiment, before performing the step of obtaining a high-resolution snapshot of the model by performing orthogonal projection and perspective projection on the original 3D model, the information acquisition module 302 is further configured to: drive the virtual camera to rotate around the model reference point of the original 3D model in an orbital manner based on the interaction information between the user terminal and the view cube component, so as to switch the standard viewpoint, and determine the target viewpoint by performing multi-angle observation of the model, so as to obtain a high-resolution snapshot of the model under the target viewpoint.

[0086] In one embodiment, when performing illumination normalization and distortion detection processing on the input image to determine the target input image, the image processing module 304 is further configured to: perform intelligent lighting and shadow processing on the input image based on the improved lighting model; recolor the main light source and ambient light of the input image through lighting effect corpus control and physical lighting rendering engine to normalize the illumination of the input image; and perform distortion detection processing on the input image after lighting and shadow processing to determine the target input image.

[0087] In one embodiment, when performing distortion detection processing on the input image after lighting and shadow processing to determine the target input image, the image processing module 304 is further configured to: perform distortion detection processing on the input image after lighting and shadow processing using a Hausdorff distance calculation model to determine the bidirectional Hausdorff distance; if the bidirectional Hausdorff distance exceeds a preset distance threshold, then re-render; if the bidirectional Hausdorff distance does not exceed the preset distance threshold, then determine that the geometric features of the input image are consistent with those of the original 3D model, and determine the input image as the target input image.

[0088] In one embodiment, when sending the target input image, model feature information, and user requirement information to a preset intelligent color matching model for intelligent color matching processing to obtain the target color scheme, the aforementioned intelligent color matching module 306 is further configured to: acquire brand database corpus information, and extract color features from the brand style map in the brand database corpus information to obtain target color information; after locking the target color information, perform intelligent color matching processing on the target input image, model feature information, and user requirement information through the preset intelligent color matching model to obtain the target color scheme.

[0089] In one embodiment, when performing the step of extracting color features from the brand style map in the brand database corpus to obtain the target color information, the intelligent color matching module 306 is further configured to: perform cluster analysis on each pixel in the brand style map using the K-means clustering algorithm to obtain the pixel proportion of each cluster; determine the primary color and each secondary color, as well as the proportion information corresponding to the primary color and each secondary color, based on the pixel proportion of the clusters, and determine the primary color, each secondary color, and the proportion information as the target color information.

[0090] 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.

[0091] 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.

[0092] Figure 4 This 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.

[0093] 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.

[0094] 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.

[0095] 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.

[0096] 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.

[0097] 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.

[0098] 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.

[0099] 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. An artificial intelligence-based three-dimensional model intelligent color matching method, characterized in that, The method comprises: obtaining an original three-dimensional model and user demand information, and performing rendering processing and feature extraction processing on the original three-dimensional model to obtain a model high-resolution snapshot and model feature information; performing background filtering processing on the model high-resolution snapshot based on a semantic segmentation deep learning model, removing the non-model area in the model high-resolution snapshot at the pixel level to obtain an input image, and performing illumination normalization processing and distortion detection processing on the input image to determine a target input image; sending the target input image, the model feature information and the user demand information to a preset intelligent color matching model for intelligent color matching processing to obtain a target color matching scheme; wherein the step of performing rendering processing and feature extraction processing on the original three-dimensional model to obtain a model high-resolution snapshot and model feature information comprises: performing orthogonal projection processing and perspective projection processing on the original three-dimensional model to obtain the model high-resolution snapshot, and performing feature extraction processing on the original three-dimensional model to obtain the model feature information; wherein, before the step of performing orthogonal projection processing and perspective projection processing on the original three-dimensional model to obtain the model high-resolution snapshot, the method comprises: driving a virtual camera to orbitally rotate around a model reference point of the original three-dimensional model based on interaction information between a user terminal and a perspective cube component to switch a standard perspective, and determining a target perspective by performing multi-angle observation of the model to obtain the model high-resolution snapshot under the target perspective; wherein the step of performing illumination normalization processing and distortion detection processing on the input image to determine a target input image comprises: performing intelligent light and shadow processing on the input image based on an improved illumination model, recoloring the main light source and the ambient light of the input image through light effect corpus control and physical illumination rendering engine to standardize the illumination of the input image, and performing distortion detection processing on the input image after light and shadow processing to determine the target input image; wherein, when performing background filtering and distortion detection processing, a two-dimensional projection result of the three-dimensional model snapshot is used to construct input parameters of an artificial intelligence large model together with geometric feature data of the three-dimensional model itself, wherein the geometric feature data of the three-dimensional model itself includes a volume feature set, a surface feature set, a normal vector set and material mapping information, wherein a high-resolution snapshot is generated during the model rendering stage, and model feature information is extracted synchronously, and the model feature information is packaged in Json format to utilize geometric information for morphological constraint during model inference. 2.The AI-based three-dimensional model intelligent color matching method according to claim 1, characterized in that, The step of performing distortion detection processing on the input image after light and shadow processing to determine the target input image comprises: performing distortion detection processing on the input image after light and shadow processing by a Hausdorff distance calculation model to determine a bidirectional Hausdorff distance; if the bidirectional Hausdorff distance exceeds a preset distance threshold, re-performing rendering processing; If the bidirectional Hausdorff distance does not exceed the preset distance threshold, it is determined that the input image is consistent with the geometric features of the original three-dimensional model, and the input image is determined as the target input image. 3.The AI-based three-dimensional model intelligent color matching method according to claim 1, characterized in that, The step of sending the target input image, the model feature information and the user demand information to a preset intelligent color matching model for intelligent color matching processing to obtain a target color matching scheme, comprises: Obtaining brand corpus information, and performing color feature extraction processing on the brand style graph in the brand corpus information to obtain target color information; After locking the target color information, the target input image, the model feature information and the user demand information are processed by a preset intelligent color matching model to obtain the target color matching scheme. 4.The AI-based three-dimensional model intelligent color matching method according to claim 3, characterized in that, The step of obtaining target color information by performing color feature extraction processing on the brand style graph in the brand corpus information, comprises: Performing clustering analysis processing on each pixel point in the brand style graph by a K-means clustering algorithm to obtain a pixel proportion of each cluster; Determine the main color and each auxiliary color according to the pixel proportion of each cluster, and the proportion information corresponding to the main color and each auxiliary color, and determine the main color, each auxiliary color and the proportion information as the target color information.

5. An artificial intelligence-based three-dimensional model intelligent color matching device, characterized in that, The device comprises: An information acquisition module acquires an original three-dimensional model and user demand information, and performs rendering processing and feature extraction processing on the original three-dimensional model to obtain a model high-resolution snapshot and model feature information; An image processing module performs background filtering processing on the model high-resolution snapshot by a deep learning model based on semantic segmentation, removes non-model areas in the model high-resolution snapshot at a pixel level to obtain an input image, and performs illumination standardization processing and distortion detection processing on the input image to determine a target input image; An intelligent color matching module sends the target input image, the model feature information and the user demand information to a preset intelligent color matching model for intelligent color matching processing to obtain a target color matching scheme; The step of performing rendering processing and feature extraction processing on the original three-dimensional model to obtain a model high-resolution snapshot and model feature information, comprises: performing orthogonal projection processing and perspective projection processing on the original three-dimensional model to obtain the model high-resolution snapshot, and performing feature extraction processing on the original three-dimensional model to obtain the model feature information; Before the step of obtaining the model high-resolution snapshot by performing orthogonal projection processing and perspective projection processing on the original three-dimensional model, it comprises: driving a virtual camera to perform orbital rotation around a model reference point of the original three-dimensional model to switch a standard view angle according to interactive information between a user terminal and a view angle cube component, and determining a target view angle by performing multi-angle observation of the model to obtain the model high-resolution snapshot under the target view angle; The step of performing illumination normalization processing and distortion detection processing on the input image to determine a target input image comprises: performing intelligent light and shadow processing on the input image based on an improved illumination model, recoloring the main light source and ambient light of the input image through light effect corpus control and physical illumination rendering engine, normalizing the illumination of the input image, and performing distortion detection processing on the input image after light and shadow processing to determine the target input image. In the background filtering and distortion detection processing, a two-dimensional projection result of a three-dimensional model snapshot is used to construct input parameters of an artificial intelligence large model together with geometric feature data of the three-dimensional model, wherein the geometric feature data of the three-dimensional model includes a volume feature set, a surface feature set, a normal vector set, and material mapping information, and in the model rendering stage, a high-resolution snapshot is generated and model feature information is extracted synchronously, and the model feature information is packaged in a Json format to constrain the morphology by using the geometric information during model inference.

6. A server, characterized by The processor executes 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, and when the computer executable instructions are called and executed by the processor, the computer executable instructions cause the processor to implement the method of any one of claims 1 to 4.

Images