Image-embossment conversion process method based on AI
Through the AI-based image-relief conversion process method, problems such as inaccurate depth information and complex calculations when image conversion into relief are solved, efficient and accurate relief production is achieved, ensuring the high quality and personalized effect of relief works.
Patent Information
- Application Number
- CN202510329866.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-20
- Publication Date
- 2025-06-20
AI Technical Summary
When the prior art converts images into relief, there are problems such as inaccurate depth information, complex calculations, inefficient efficiency and unsatisfactory results.
Using the AI-based image-relief conversion process method, efficient image-to-relief conversion is achieved through steps such as multimodal data initialization, deep integration of element data, visualization and grayscale image generation, automated flattening and manual optimization.
It greatly improves the efficiency of depth map production, reduces artificial calculation errors, improves the accuracy and consistency of relief production, and ensures the high quality and personalized effect of the final relief work.
Smart Images
Figure CN120182498A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of image conversion, and specifically relates to an image-relief conversion process method based on AI. Background Art
[0002] In the early days, the generation of depth maps mainly relied on manual drawing. This process is not only time-consuming and laborious, but also requires extremely high skills from professional draftsmen. In addition, it is greatly affected by subjective factors. There are large differences in depth maps drawn by different draftsmen, making it difficult to ensure the accuracy and consistency of depth information, which seriously limits the efficiency and quality of converting two-dimensional images into three-dimensional relief images.
[0003] In terms of relief size calculation and 3D model construction, the traditional method relies on manual conversion based on experience and simple mathematical formulas, which is prone to calculation errors, resulting in relief size specifications not meeting expectations. When using conventional modeling tools, the operation process is complicated and a large number of parameters need to be set manually. For complex relief shapes, modeling is extremely difficult and often requires a lot of time and effort, and the final model effect is difficult to achieve the ideal state.
[0004] Image segmentation and element processing also face challenges. Traditional methods are difficult to accurately identify various elements in an image, let alone automatically merge identical elements and count their number. The coordinate and size records of elements are not only low in precision but also inefficient, and cannot meet the needs of fine processing of complex images. This makes it extremely difficult to independently process individual elements in the future, making it difficult to achieve personalized design and optimization of relief effects.
[0005] When constructing a complete relief image, the process of stitching elements together using traditional techniques is cumbersome and lacks accuracy. When stitching elements together manually, it is difficult to ensure the accuracy of the size and position of each element, resulting in a lack of layering and three-dimensionality in the final relief work, and the inability to show a delicate and realistic effect. At the same time, due to the lack of effective visual auxiliary means, it is difficult to fully and intuitively grasp the relationship between elements during the stitching process, which is not conducive to timely discovery and resolution of problems.
[0006] To this end, the present invention provides an AI-based image-relief conversion process method. Summary of the invention
[0007] In order to make up for the deficiencies of the prior art, at least one technical problem raised in the background technology is solved.
[0008] The technical solution adopted by the present invention to solve the technical problem is: the AI-based image-relief conversion process method of the present invention is characterized in that the method comprises the following steps:
[0009] S1. Multi-modal data initialization: Import the original image into the AI to generate a rough depth map. Based on the conversion relationship between the relief size and the custom pixel, construct the prototype of the 3D map. At the same time, use the AI image recognition technology to segment the image, identify and number the elements, merge the same elements, and record the coordinate dimensions and pixel ratio relationships to prepare multi-dimensional basic data for subsequent operations;
[0010] S2. In-depth integration of element data: Based on the original image, the final image and the element segmentation information obtained in the early stage, accurately calculate the coordinate positions, magnification ratios of each element in the final relief, clarify the depth position relationship, integrate and form a detailed data table, and unify the data standards to provide the core basis for the subsequent processing of the elements;
[0011] S3. Visualization and grayscale map generation: On the one hand, use professional graphics software to generate a visualization schematic diagram according to the element contours and coordinates to intuitively display the element distribution and position relationship; on the other hand, for each element, based on its depth and material characteristics, use AI to draw a grayscale map with a relief effect to lay the foundation for the final splicing;
[0012] S4. Automatic splicing and manual optimization: The system automatically scales, crops the element grayscale maps according to the established data, and moves them to the specified coordinate positions to complete the splicing. At the same time, a manual operation channel is reserved to allow users to fine-tune the element size and position according to actual needs to ensure that a high-quality relief work that meets expectations is finally presented.
[0013] Preferably, the step S1 includes the following steps:
[0014] S101. Import the original image to generate a depth map: Import the original image into a specific AI tool or software. With the help of an algorithm model based on deep learning, after learning a large amount of image data, analyze the distance relationship of objects to generate a rough depth map; provide basic depth information for the two-dimensional to three-dimensional relief map to assist in constructing a three-dimensional model;
[0015] S102. Relief size calculation and 3D map conversion: Determine the relief depth of 100 mm. According to the custom pixel size conversion relationship of 1 px = 0.5 mm, calculate the pixel size of 10000 px × 3000 px from the actual relief size of 5000 mm × 1500 mm; Use 3D modeling software such as Blender and 3dsMax to import the image with the calculated size and the depth map generated in S101, set the grid resolution and material property parameters, and convert it into a three-dimensional model; Convert the planar image into a three-dimensional relief model with actual depth, and determine the relief size specifications and basic shapes;
[0016] S103. Image Recognition and Segmentation: Use AI image recognition technology based on the convolutional neural network object detection algorithm to identify all elements in the original image and number them; the AI automatically identifies and counts the number of identical elements; support users to manually merge elements according to design requirements; calculate the boundary range of elements in units of individual pixels of the original image, record the coordinates and dimensions; confirm and record the pixel size ratio relationship between the original image and the target relief image; decompose the original image into independent elements, optimize element management, and accurately record the key information of elements to prepare for subsequent individual processing of elements.
[0017] Preferably, the step S3 includes the following steps:
[0018] S301. Confirm the Final Size and Coordinate Position: Recheck the final size and coordinate position information of each element in S2, compare it with the original image, 3D model, and design requirements, check the rationality of the relative position of elements and whether the size meets the expected relief effect. If the data is deviated, return to S2 for correction to ensure the accuracy of the element size and coordinate information used in subsequent operations and prevent the relief production from being affected by data errors.
[0019] S302. Generate a Visual Schematic Diagram: Select professional graphic design software such as Adobe Illustrator, CorelDRAW or visualization tools, import the contour and coordinate position data of the segmented elements in S2, set the graphic styles: line color, thickness, draw the element contours according to the coordinate positions, adjust the perspective, add annotations, and generate multiple views: top view, side view, to display the element contours and position relationships, visually show the element distribution and position relationships, help understand the relief structure, provide a visual reference for element fitting and adjustment, and facilitate problem discovery.
[0020] S303. Draw the Relief Grayscale Image: For each element determined in S2, use a dedicated AI drawing tool or software with grayscale image drawing function, based on the depth position relationship and surface characteristics of the elements, use the AI algorithm to generate the relief grayscale image, with the depth change corresponding to the gradual change of grayscale values. According to the material properties of the elements such as metal, wood, adjust the details of the grayscale image to simulate different material textures, generate grayscale images with relief effects for each element, as the basis for subsequent fitting, and construct a complete relief image to make the relief work present rich levels and three-dimensional senses.
[0021] Preferably, the step S4 includes the following steps:
[0022] S401. Automatic Scaling: Combine the final size confirmed in S3, the magnification ratio information in S2, and the relief grayscale image generated in S303, and use an image transformation algorithm such as affine transformation to scale the grayscale image of each element as required. For example, if the element size in the final relief needs to be magnified several times, the length and width of the grayscale image will be magnified accordingly.
[0023] S402. Cropping process: For the outline and pattern content of the scaled grayscale image of the element, use an image segmentation algorithm, such as a threshold-based segmentation algorithm, to identify the part of the grayscale image that belongs to the embossed content of the element itself, remove the surrounding blank area, and only retain the embossed pattern part;
[0024] S403. Automatic positioning: For the coordinate position of each element determined in S2, through a coordinate transformation algorithm, automatically move the cropped grayscale image of the element to the established coordinate position in the final embossed image to form a complete embossed layout;
[0025] S404. Manual fine-tuning: Provide a user interface and retain the manual operation channel. The user can manually drag the element edge to fine-tune the size or modify the coordinate value in the coordinate input box to adjust the position. During the fine-tuning process, the system will display the splicing effect in real time to facilitate the user to judge whether the adjustment is appropriate.
[0026] The beneficial effects of the present invention are as follows:
[0027] 1. For the AI-based image-emboss conversion process method described in the present invention, by importing the original image into the AI tool and using the deep learning algorithm to quickly generate a rough depth map, it provides key basic information for the conversion of a two-dimensional image into a three-dimensional embossed map, greatly shortening the production cycle of the depth map and significantly improving the efficiency compared with the traditional manual drawing of the depth map. According to the custom pixel conversion relationship, accurately calculate the embossed pixel size, and with the help of a professional 3D modeling software, quickly construct the prototype of the 3D model, determine the embossed size specifications and basic shape, reduce human calculation errors, and improve the modeling efficiency and accuracy. Using the AI image recognition technology, it can quickly identify all elements in the original image, automatically merge the same elements and count them, and at the same time support the user to manually merge elements, and can accurately record the element coordinates, size and pixel ratio relationship, efficiently decompose the complex image into independent and orderly managed element units, laying a solid foundation for subsequent refined processing.
[0028] 2. For the AI-based image-emboss conversion process method described in the present invention, by deeply integrating the original image, the final image and the element segmentation information, accurately calculate the coordinate position, magnification ratio and depth position relationship of each element in the final emboss, form a detailed data table, and unify the data standard. This enables accurate and unified basis for positioning, size adjustment and depth setting of elements in subsequent operations, avoiding production deviations caused by data chaos, and improving the accuracy and consistency of emboss production.
[0029] 3. The AI-based image-relief conversion process method described in the present invention uses professional graphic software to generate visualization diagrams, intuitively showing the element distribution and positional relationships, and presenting the element layout from different angles through multiple views (such as top view, side view). This helps users to more clearly and comprehensively understand the entire relief structure, quickly discover potential problems and unreasonable points during the element fitting and adjustment process, and improve work efficiency and decision-making accuracy.
[0030] 4. The AI-based image-relief conversion process method described in the present invention, for each element, according to its depth and material characteristics, generates a grayscale image with a realistic relief effect through AI algorithms. The grayscale value gradually changes with the depth and can adjust details according to different material properties (such as metal, wood) to simulate the corresponding texture. These high-quality grayscale images provide high-quality materials for constructing a complete relief image with rich levels and three-dimensional senses, enhancing the visual effect of the final relief work.
[0031] 5. The AI-based image-relief conversion process method described in the present invention, the system automatically scales, crops and locates the element grayscale images according to the established data, completes efficient fitting, ensures the efficiency and accuracy of the fitting process, reduces the tediousness of manual operations, and reduces the error probability. The manual operation channel is reserved, and users can manually fine-tune the element size and position according to the actual effect and design requirements. The system displays the fitting effect in real time for users to judge. This combination of automation and manual optimization not only improves production efficiency but also fully meets the personalized and refined design requirements, ensuring that the final high-quality relief work meets the user's expectations. BRIEF DESCRIPTION OF THE DRAWINGS
[0032] The present invention will be further described below with reference to the accompanying drawings.
[0033] Figure 1 is the process flow chart in the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0034] In order to make the technical means, creative features, achieved purposes and effects of the present invention easy to understand, the present invention will be further described below in conjunction with specific embodiments.
[0035] As Figure 1 shown, the AI-based image-relief conversion process method of the embodiment of the present invention is characterized in that the method includes the following steps:
[0036] S1. Multi-modal data initialization: Import the original image into the AI to generate a rough depth map. Based on the conversion relationship between the relief size and the custom pixel, construct the prototype of the 3D map. At the same time, use the AI image recognition technology to segment the image, identify and number the elements, merge the same elements, and record the coordinate dimensions and pixel ratio relationships to prepare multi-dimensional basic data for subsequent operations;
[0037] S2. In-depth integration of element data: Based on the original image, the final image and the element segmentation information obtained in the early stage, accurately calculate the coordinate positions, magnification ratios of each element in the final relief, clarify the depth position relationship, integrate and form a detailed data table, and unify the data standard to provide the core basis for the subsequent processing of the elements;
[0038] S3. Visualization and grayscale map generation: On the one hand, use professional graphics software to generate a visualization schematic diagram according to the element contours and coordinates to intuitively display the element distribution and position relationship; on the other hand, for each element, based on its depth and material characteristics, use AI to draw a grayscale map with a relief effect to lay the foundation for the final splicing;
[0039] S4. Automatic splicing and manual optimization: The system automatically scales, crops the element grayscale maps according to the established data, and moves them to the specified coordinate positions to complete the splicing. At the same time, a manual operation channel is reserved to allow users to fine-tune the element size and position according to actual needs to ensure that the final high-quality relief work meets the expectations.
[0040] The step S1 includes the following steps:
[0041] S101. Import the original image to generate a depth map: Import the original image into a specific AI tool or software. With the help of an algorithm model based on deep learning, after learning a large amount of image data, analyze the object's distance relationship to generate a rough depth map; provide basic depth information for the two-dimensional to three-dimensional relief map to help construct a three-dimensional model;
[0042] S102. Relief size calculation and 3D map conversion: Determine the relief depth of 100 mm. According to the custom pixel size conversion relationship of 1 px = 0.5 mm, calculate the pixel size of 10000 px × 3000 px from the actual relief size of 5000 mm × 1500 mm; use 3D modeling software such as Blender and 3ds Max to import the image with the calculated size and the depth map generated in S101, set the grid resolution and material property parameters, and convert it into a three-dimensional model; convert the planar image into a three-dimensional relief model with actual depth, and determine the relief size specifications and basic shapes;
[0043] S103. Image Recognition and Segmentation: Use AI image recognition technology, based on the convolutional neural network object detection algorithm, to identify all elements in the original image and number them; AI automatically identifies and counts the number of identical elements; supports users to manually merge elements according to design requirements; calculate the boundary range of elements in units of individual pixels of the original image, record the coordinates and dimensions; confirm and record the pixel size ratio relationship between the original image and the target relief image; decompose the original image into independent elements, optimize element management, and accurately record the key information of elements to prepare for subsequent individual processing of elements.
[0044] Specifically,
[0045] Element Recognition and Numbering: Use AI image recognition technology to analyze the original image. Through specific algorithms (such as the object detection algorithm based on convolutional neural network), all elements in the image are identified, and each element is given a unique identifier and number. For example, for an image containing multiple objects, AI can regard each object as an independent element and number them separately as 1, 2, 3, etc.
[0046] Counting and Merging of Identical Elements: AI continues to analyze the identified elements, automatically identifies the identical elements, and merges and counts them. For example, there may be multiple identical flower patterns in the image. AI can identify that these flowers belong to the same type of element and count the number of flower elements. For example, there are 8 flower elements numbered 10 and 2 elements of a certain shape numbered 11.
[0047] Element Merging Operation: Provide the function for users to manually merge any two or more elements. For example, users can, according to design requirements, specify the merging of elements numbered 9 and 13 into a new entity. This is very useful in practical applications when certain elements need to be presented as a whole in the relief effect.
[0048] Recording of Coordinates and Dimensions: Taking the individual pixel of the original image as the coordinate system unit, AI accurately records the coordinates and size of each element after segmentation by calculating the boundary range of each element in the image. For example, if the upper left corner coordinates of a certain element are (x1, y1) and the lower right corner coordinates are (x2, y2), then its size is (x2 - x1) × (y2 - y1) pixels.
[0049] Clarifying the Pixel Ratio Relationship: Reconfirm and record the pixel size ratio relationship between the original image and the target relief image. This relationship is crucial for subsequent size adjustment and position calculation of elements, ensuring accurate conversion of elements under different image sizes.
[0050] Objective: To decompose the original image into independent element units, providing a basis for subsequent individual processing of each element. Meanwhile, by merging the same or related elements, optimize element management, and accurately record the key information of each element, facilitating subsequent operations.
[0051] Step S3 includes the following steps:
[0052] S301. Confirm the final size and coordinate position: Review the final size and coordinate position information of each element in S2, compare it with the original image, 3D model, and design requirements, check the rationality of the relative position of the elements and whether the size meets the expected relief effect. If there are data deviations, return to S2 for correction to ensure the accuracy of the element size and coordinate information used in subsequent operations and prevent problems in relief production due to data errors.
[0053] S302. Generate a visualization schematic diagram: Select professional graphic design software such as Adobe Illustrator, CorelDRAW or visualization tools, import the contour and coordinate position data of the segmented elements in S2, set the graphic style: line color, thickness, draw the element contours according to the coordinate positions, adjust the perspective, add annotations, and generate multiple views: top view, side view, to display the element contours and positional relationships, presenting the element distribution and positional relationships in an intuitive graph, helping to understand the relief structure, providing a visual reference for element fitting and adjustment, and facilitating problem discovery.
[0054] S303. Draw the relief grayscale image: For each element determined in S2, use a specialized AI drawing tool or software with grayscale image drawing function, based on the depth position relationship and surface characteristics of the elements, use AI algorithms to generate the relief grayscale image, with the depth change corresponding to the gradual change of grayscale values, adjust the details of the grayscale image according to the element material properties such as metal, wood, simulate different material textures, generate grayscale images with relief effects for each element, as the basis for subsequent fitting, and construct a complete relief image, making the relief work present a rich sense of hierarchy and three-dimensionality.
[0055] Specifically,
[0056] Confirm the final size and coordinate position:
[0057] Operation: Check and confirm again the final size and coordinate position information of each element calculated in S2. By comparing with the original image, 3D model, and design requirements, ensure the accuracy of the data. This may involve checking whether the relative position between elements is reasonable and whether the size meets the expected relief effect. If data deviations are found, return to S2 in a timely manner for correction.
[0058] Objective: Ensure that in subsequent operations, the element size and coordinate information used are accurate and error-free, avoiding problems in relief production due to data errors.
[0059] Generate a visualization schematic diagram:
[0060] Operation: Use professional graphic design software (such as Adobe Illustrator, CorelDRAW, etc.) or specialized visualization tools to import the contour information and coordinate position data of the segmented elements obtained from S2. Set appropriate graphic styles in the software, such as line color, thickness, etc., and accurately draw the contours of each element on the canvas according to their coordinate positions. By adjusting the perspective, adding annotations, etc., generate a visualization schematic diagram that can clearly show the contours of the segmented elements and the coordinate position relationships between them. Multiple views can be generated, such as top view, side view, etc., to observe the layout of the elements from different angles.
[0061] Purpose: Display the distribution and position relationships of the elements in an intuitive graphical way, help users understand the structure of the entire relief more clearly, provide a visualization reference for subsequent element assembly and adjustment, and facilitate the discovery of potential problems and unreasonable points.
[0062] Draw a grayscale relief image:
[0063] Operation: For each element determined in S2, use a specialized AI drawing tool or software with grayscale image drawing function. According to the depth position relationship and surface characteristics of each element, use AI algorithms to generate corresponding grayscale relief images. For example, if the depth range of a certain element is 0 - 50 mm, the AI will generate gradually changing grayscale values from light to dark according to the depth change, so that the grayscale image can reflect the relief effect of the element. During the drawing process, the details of the grayscale image can be adjusted according to the material properties of the element (such as metal, wood, etc.) to simulate the texture of different materials.
[0064] Purpose: Generate grayscale images with relief effects for each element, which will be used as the basis for subsequent assembly to construct a complete relief image, so that the final relief work presents rich layers and three - dimensional sense visually.
[0065] The step S4 includes the following steps:
[0066] S401. Automatic scaling: Combining the final size confirmed in S3, the magnification ratio information in S2, and the grayscale relief image generated in S303, use image transformation algorithms such as affine transformation to scale the grayscale image of each element as required. For example, if the element size in the final relief needs to be magnified several times, the length and width of the grayscale image will be magnified accordingly.
[0067] S402. Cropping process: For the outline and pattern content of the scaled grayscale image of the element, use an image segmentation algorithm, such as a threshold-based segmentation algorithm, to identify the part of the grayscale image that belongs to the embossed content of the element itself, remove the surrounding blank area, and only retain the embossed pattern part;
[0068] S403. Automatic positioning: For the coordinate positions of each element determined in S2, through a coordinate transformation algorithm, automatically move the cropped grayscale image of the element to the established coordinate position in the final embossed image to form a complete embossed layout;
[0069] S404. Manual fine-tuning: Provide a user interface and retain the manual operation channel. The user can manually drag the element edge to fine-tune the size or modify the coordinate value in the coordinate input box to adjust the position. During the fine-tuning process, the system will display the splicing effect in real time to facilitate the user to judge whether the adjustment is appropriate.
[0070] Purpose: Accurately splice together the embossed grayscale images of each element after preprocessing to form a complete embossed image, ensure efficiency and accuracy through automatic operations, and at the same time meet personalized and refined design requirements with the help of manual fine-tuning, and finally obtain an embossed work that meets expectations.
[0071] Furthermore, based on the proportional relationship between the original image and the final image obtained after S102 processing (i.e., the image size and ratio corresponding to the three-dimensional embossed model), as well as the size and coordinate information of the individual elements obtained from S103 after segmentation.
[0072] Calculate the coordinate positions of each segmented element in the final embossed model. For example, through proportional conversion, convert the coordinates of the element in the original image to the coordinate system of the final embossed image. At the same time, calculate the magnification ratio of each element from the original size to the final embossed size. For example, if the size of a certain element in the original image is 100×100 pixels and becomes 200×200 pixels in the final embossed image, the magnification ratio is 2.
[0073] Combined with the depth map information generated by S101, determine the positional relationship of each element in the embossed depth direction. For example, some elements may be on the surface of the embossment (depth is 0), while some elements may be at a certain depth inside the embossment (such as a depth of 50 mm).
[0074] Finally, record the final size, depth range, and coordinates of each element (all coordinates are based on the lower left corner as the origin):
[0075] Element 1: 1150.6x1108.03, depth 0 - 100, coordinates 115.62, 0;
[0076] Element 2: ……;
[0077] Element 3: ……;
[0078] Element 4: ……;
[0079] Element 5: ……;
[0080] Element 6: ……;
[0081] Element 7-1: 452.67 x 273.63, depth 50.91 - 100, coordinates 860.67, 1167.36;
[0082] Element 7-2: 452.67 x 273.63, depth 50.91 - 100, coordinates 860.67, 1167.36;
[0083] Element 8: ……;
[0084] Element 9: …… (including Element 13);
[0085] Element 10-1: ……;
[0086] Element 10-2: ……;
[0087] Element 10-8: ……;
[0088] Element 11: ……;
[0089] Element 12: ……;
[0090] Element 13: (incorporated into Element 9, marked as empty);
[0091] Element 14: ……;
[0092] Note:
[0093] 1. For duplicate elements, the dimensions and coordinates of each element need to be marked separately;
[0094] 2. For merged elements, they are defined as a single text entity, and the overall dimensions and overall coordinates after merging are marked, such as the merging of Elements 9 and 13;
[0095] 3. For elements that have been attached and merged with other elements, such as Element 9 attached to Element 13, the mark of Element 13 is empty.
[0096] Purpose: To integrate various data obtained in the previous steps, match and convert the information of image elements with the requirements of the final relief model, and provide a unified data basis for subsequent accurate positioning, size adjustment, and depth setting of elements.
[0097] The foregoing has shown and described the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments, and what is described in the above embodiments and the specification is only to illustrate the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements fall within the scope of the present invention claimed. The scope of the present invention claimed is defined by the appended claims and their equivalents.
Claims
1. The AI-based image-relief conversion process method is characterized by: The method comprises the following steps: S1. Multimodal data initialization: Import the original image into AI to generate a rough depth map, build a 3D image prototype based on the relief size and custom pixel conversion relationship, and use AI image recognition technology to segment the image, identify and number elements, merge the same elements, record the coordinate size and pixel ratio relationship, and prepare multi-dimensional basic data for subsequent operations; S2. In-depth integration of element data: Based on the original image, final image and element segmentation information obtained in the early stage, the coordinate position and magnification ratio of each element in the final relief are accurately calculated, the depth position relationship is clarified, and a detailed data table is integrated to unify the data standards, providing a core basis for the subsequent processing of the elements; S3. Visualization and grayscale image generation: On the one hand, professional graphics software is used to generate a visualization diagram based on the element outline and coordinates to intuitively display the element distribution and position relationship; on the other hand, for each element, according to its depth and material characteristics, AI is used to draw a grayscale image with a relief effect, laying the foundation for the final splicing; S4. Automatic stitching and manual optimization: The system automatically scales and crops the element grayscale image based on the given data, and moves it to the specified coordinate position to complete the stitching. At the same time, it retains the manual operation channel, allowing the user to fine-tune the element size and position according to actual needs, ensuring that the final high-quality relief work that meets expectations is presented.
2. The AI-based image-relief conversion process according to claim 1, characterized in that: The step S1 comprises the following steps: S101. Import the original image to generate a depth map: Import the original image into a specific AI tool or software, and use the algorithm model based on deep learning to analyze the distance relationship of the object after learning a large amount of image data to generate a rough depth map; provide basic depth information for converting 2D to 3D relief images, and help build a 3D model; S102, relief size calculation and 3D image conversion: clarify the relief depth of 100mm, according to the custom pixel size conversion relationship 1px = 0.5mm, the pixel size of 10000px × 3000px is calculated from the actual relief size of 5000mm × 1500mm; use 3D modeling software such as Blender and 3dsMax to import the image with calculated size and the depth map generated by S101, set the grid resolution and material attribute parameters, and convert it into a 3D model; convert the plane image into a 3D relief model with actual depth, and determine the relief size specifications and basic shape; S103, Image recognition and segmentation: Use AI image recognition technology, based on the convolutional neural network target detection algorithm, to identify and number all elements of the original image; AI automatically identifies and counts the number of identical elements; supports users to manually merge elements according to design requirements; calculates the element boundary range, records coordinates and dimensions based on a single pixel of the original image; confirms and records the pixel size ratio between the original image and the target relief image; decomposes the original image into independent elements, optimizes element management, accurately records key element information, and prepares for subsequent individual element processing.
3. The AI-based image-relief conversion process according to claim 2, characterized in that: The step S3 comprises the following steps: S301, confirm the final size and coordinate position: review the final size and coordinate position information of each element in S2, compare it with the original image, three-dimensional model, and design requirements, check the rationality of the relative position of the elements, and whether the size meets the expected relief effect. If the data is biased, return to S2 for correction to ensure that the element size and coordinate information used in subsequent operations are accurate, so as to prevent the relief production from being affected by data errors; S302, generate visualization diagram: use professional graphic design software such as Adobe Illustrator, CorelDRAW or visualization tools, import the outline and coordinate position data of the segmented elements in S2, set the graphic style: line color, thickness, draw the element outline according to the coordinate position, adjust the viewing angle, add annotations, generate multiple views: top view, side view, display the element outline and position relationship, display the element distribution and position relationship with intuitive graphics, help understand the relief structure, provide visual reference for element splicing and adjustment, and facilitate problem discovery; S303, draw relief grayscale image: for each element determined in S2, use a special AI drawing tool or software with grayscale drawing function, and generate a relief grayscale image using AI algorithm according to the depth position relationship and surface characteristics of the elements. The depth change corresponds to the grayscale value gradient. According to the material properties of the elements such as metal and wood, adjust the grayscale image details, simulate the texture of different materials, and generate a grayscale image with relief effect for each element. As the basis for subsequent stitching, construct a complete relief image, so that the relief work presents rich layering and three-dimensional sense.
4. The AI-based image-relief conversion process according to claim 3, characterized in that: The step S4 comprises the following steps: S401, automatic scaling: combining the final size confirmed in S3, the magnification ratio information in S2 and the relief grayscale image generated in S303, using an image transformation algorithm such as affine transformation, the grayscale image of each element is scaled as required. For example, if the size of the element in the final relief needs to be magnified several times, the length and width of the grayscale image are magnified several times accordingly; S402, cropping processing: using an image segmentation algorithm, such as a threshold-based segmentation algorithm, to identify the part of the grayscale image that has embossed content belonging to the element itself, remove the surrounding blank area, and only retain the embossed pattern part; S403, automatic positioning: the coordinate position of each element determined in S2 is automatically moved to the predetermined coordinate position in the final relief image by the coordinate transformation algorithm, so as to form a complete relief layout; S404, manual fine-tuning: Provide a user interface and retain a manual operation channel. Users can manually drag the edge of the element to fine-tune the size, or modify the coordinate value in the coordinate input box to adjust the position. During the fine-tuning process, the system displays the stitching effect in real time to facilitate users to judge whether the adjustment is appropriate.