Method for generating three-dimensional landscape character model

Abstract

The invention relates to the field of digital art, and discloses a method for generating a three-dimensional landscape character model, which comprises the following steps of: 1, providing input character style information; 2, analyzing the character style information through an artificial intelligence algorithm, and extracting character style features; step 3, automatically generating a three-dimensional matrix model based on the character style features; step 4, carrying out preliminary modeling on the three-dimensional base body by using a polygon modeling technology; 5, applying a curved surface modeling method to the polygon modeling result to generate a smooth surface; and step 6, mapping and enhancing detail textures of the three-dimensional matrix model through a deep neural network. The input character style information is automatically analyzed through the artificial intelligence algorithm, the deep convolutional neural network and the generative adversarial network, and the three-dimensional character model is generated, so that the requirement of manual intervention is greatly reduced, the generation efficiency and accuracy of the three-dimensional character model are improved, and a large amount of time and cost are saved.

Description

Technical Field

[0001] The invention relates to the technical field of digital art, in particular to a method for generating a three-dimensional landscape character model. Background Art

[0002] With the continuous development of computer graphics and virtual reality technology, 3D modeling has become an indispensable technology in many industries (such as advertising design, film and television special effects, game development, virtual reality, etc.). Especially in the field of advertising and brand promotion, 3D landscape word models, as an important visual presentation method, have a high visual impact and communication effect. Traditional 3D word model production often requires manual design and fine adjustment, which is not only time-consuming and labor-intensive, but also difficult to achieve automated production. Especially when a large amount of customized content is required, it is particularly urgent to improve production efficiency and accuracy.

[0003] At present, the generation method of three-dimensional character models mainly relies on traditional computer graphics technology and manual modeling processes. Traditional methods usually use three-dimensional modeling software (such as AutodeskMaya, Blender, etc.) for manual modeling, and designers manually adjust the shape, texture, lighting effects, etc. of the letters according to the input text style. Although these methods can generate relatively fine three-dimensional character models, they often have disadvantages such as long production cycle, cumbersome design process, and excessive manual intervention. In recent years, with the advancement of deep learning technology, three-dimensional model generation methods based on artificial intelligence algorithms have gradually attracted attention. Some studies have tried to use convolutional neural networks (CNN) and generative adversarial networks (GAN) to automatically generate and optimize three-dimensional images and models.

[0004] Although the existing technology has improved the accuracy and efficiency of 3D word model generation to a certain extent, the following problems still exist: the existing methods lack effective technical support in the automatic generation and detail expression of 3D word models, especially in texture mapping, lighting adaptability, environmental adaptation and other aspects. For example, traditional texture mapping technology is prone to seams and defects, and the performance is unstable under changing lighting conditions, which makes it difficult to meet the demand for high-quality visual effects in modern advertising and virtual environments. In addition, the visual consistency of existing methods under different viewing angles cannot be guaranteed, especially when the lighting changes or the scene switches, the effect is often inconsistent. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention provides a method for generating a three-dimensional landscape word model, which solves the problem that traditional texture mapping technology is prone to produce seams and defects, and has unstable performance under changing lighting conditions, making it difficult to meet the requirements of high-quality visual effects in modern advertising and virtual environments.

[0006] To achieve the above purpose, the present invention is implemented by the following technical scheme: a method for generating a three-dimensional landscape word model, the method comprising the following steps: Step 1: Provide input text style information; Step 2: parse the text style information through an artificial intelligence algorithm to extract text style features; Step 3: Automatically generate a three-dimensional matrix model based on the text style features; Step 4: Using polygonal modeling technology to perform preliminary modeling on the three-dimensional matrix; Step 5: Applying a surface modeling method to the polygonal modeling result to generate a smooth surface; Step 6: Mapping and enhancing the detail texture of the three-dimensional matrix model through a deep neural network; Step 7: Using ray tracing technology to simulate the light and shadow of the three-dimensional matrix model under different lighting conditions; Step 8: Automatically adjust the color, material and reflection properties of the three-dimensional character model through environmental perception technology to make it blend with the lighting conditions and environment in the virtual scene.

[0007] Preferably, the text style information includes font, thickness, curve and appearance features, and the text style features are parsed by a deep convolutional neural network, and the obtained text style features are used to construct a three-dimensional matrix model.

[0008] Preferably, the three-dimensional matrix model is generated by polygonal modeling technology, which specifically includes meshing the text contour information and using surface fitting technology to smoothly optimize the generated polygonal model, and finally generating a three-dimensional glyph model.

[0009] Preferably, the three-dimensional matrix model is generated by polygonal modeling technology, which specifically includes meshing the text contour information and using surface fitting technology to smoothly optimize the generated polygonal model, and finally generating a three-dimensional glyph model.

[0010] Preferably, the ray tracing technique is used to simulate reflection, refraction and diffuse reflection phenomena under lighting conditions.

[0011] Preferably, the environment perception technology automatically adjusts the color, reflectivity and glossiness of the three-dimensional character model to make it consistent with the surrounding environment through real-time perception of the light distribution, color distribution and material properties in the virtual scene.

[0012] Preferably, the detail enhancement of the three-dimensional character model is performed through a deep convolutional neural network, specifically including multi-scale learning of letter surface textures, thereby enhancing the performance of letter edges, surface gloss and material.

[0013] Preferably, the method further comprises performing multi-view rendering on the generated three-dimensional landscape character model, and using mean square error to verify the consistency of rendering results at different viewpoints.

[0014] Preferably, the method is applicable to advertising design, virtual scene modeling, augmented reality and virtual reality application scenarios, in which the three-dimensional character model can be adaptively adjusted according to changes in different environments.

[0015] A computer-readable storage medium stores program codes, which, when executed, can implement the three-dimensional landscape character model generation method.

[0016] The present invention provides a method for generating a three-dimensional landscape character model, which has the following beneficial effects: 1. The present invention automatically parses the input text style information and generates a three-dimensional character model through artificial intelligence algorithm deep convolutional neural network and generative adversarial network, which greatly reduces the need for manual intervention and improves the generation efficiency and accuracy of the three-dimensional character model. The automated process simplifies the design process and saves a lot of time and cost.

[0017] 2. The present invention uses deep learning technology to accurately map the texture of the three-dimensional character model, and combines it with a generative adversarial network to optimize the texture effect, which can effectively eliminate the common seams and defects in traditional texture mapping. In addition, the use of a deep convolutional neural network for detail enhancement makes the model surface texture, letter edges, gloss and other details more refined, improving the realism and visual effect of the three-dimensional model.

[0018] 3. The present invention uses ray tracing technology to simulate reflection, refraction and diffuse reflection under real lighting conditions, ensuring the consistency of the performance of the three-dimensional character model in different lighting environments. Whether in strong light, weak light or dynamic light source changing environments, the model can adapt to lighting changes and maintain the stability of the visual effect.

[0019] 4. The present invention uses environmental perception technology to perceive the properties of light, color, material, etc. in the virtual scene in real time, and intelligently adjusts the color, reflectivity, glossiness and other characteristics of the three-dimensional word model according to changes in the environment. This dynamic adjustment mechanism enables the three-dimensional word model to blend harmoniously with the surrounding environment, improves the overall expressiveness of the model, and avoids the problem of poor environmental adaptation in traditional three-dimensional modeling technology.

[0020] 5. The generation method of the present invention includes multi-perspective rendering and visual consistency verification steps. By using the mean square error to compare the rendering results of different perspectives, it is ensured that the generated three-dimensional character model always maintains a consistent visual effect under multiple perspectives, avoiding the problem of inconsistent effects caused by lighting changes or texture distortion under multiple perspectives. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 is a flow chart of the method of the present invention; Figure 2 It is a schematic diagram of the storage medium structure of the present invention. DETAILED DESCRIPTION

[0022] The following will be combined with the drawings in the specification of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0023] Please see attached Figure 1 ,The embodiment of the present invention provides a method for generating a three-dimensional landscape character model, the method comprising the following steps: step 1, providing input text style information; Step 2: Analyze text style information through artificial intelligence algorithms and extract text style features; Step 3: Automatically generate a three-dimensional matrix model based on the text style features; Step 4: Use polygonal modeling technology to perform preliminary modeling of the three-dimensional matrix; Step 5: Apply the surface modeling method to the polygon modeling result to generate a smooth surface; Step 6: Map and enhance the detail texture of the three-dimensional base model through a deep neural network; Step 7: Use ray tracing technology to simulate the light and shadow of the three-dimensional matrix model under different lighting conditions; Step 8: Automatically adjust the color, material and reflection properties of the three-dimensional character model through environmental perception technology to make it blend with the lighting conditions and environment in the virtual scene.

[0024] The text style information includes font, thickness, curve and appearance features. The text style features are analyzed through a deep convolutional neural network, and the obtained text style features are used to construct a three-dimensional matrix model.

[0025] The three-dimensional matrix model is generated through polygonal modeling technology, which specifically includes meshing the text contour information and using surface fitting technology to smooth and optimize the generated polygonal model, and finally generating a three-dimensional glyph model.

[0026] The three-dimensional matrix model is generated through polygonal modeling technology, which specifically includes meshing the text contour information and using surface fitting technology to smooth and optimize the generated polygonal model, and finally generating a three-dimensional glyph model.

[0027] Ray tracing technology is used to simulate reflection, refraction and diffuse reflection phenomena under lighting conditions.

[0028] Environmental perception technology automatically adjusts the color, reflectivity and glossiness of the three-dimensional font model to make it consistent with the surrounding environment through real-time perception of the light distribution, color distribution and material properties in the virtual scene.

[0029] The detail enhancement of the three-dimensional letter model is performed through a deep convolutional neural network, which specifically includes multi-scale learning of the surface texture of the letters, thereby enhancing the performance of the letter edges, surface gloss and material.

[0030] The method further includes performing multi-view rendering on the generated three-dimensional landscape character model, and using mean square error to verify the consistency of rendering results at different viewpoints.

[0031] The method is suitable for advertising design, virtual scene modeling, augmented reality and virtual reality application scenarios, in which the three-dimensional character model can be adaptively adjusted according to changes in different environments.

[0032] In one embodiment, step 1: input text style information The user inputs the text content to be generated (such as brand name, slogan, etc.) through the computer interface and selects the corresponding font style and design elements. For example, the user inputs the text "BRAND" and selects the font ArialBold, bold style, and sets parameters such as font size and spacing.

[0033] The system obtains the input text, and combines it with the style information set by the user to generate corresponding digitized text content, and passes this content to the subsequent processing module.

[0034] Step 2: Text style information analysis The selected text style information (such as font, thickness, curve, corner, etc.) is input into the text parsing module and processed using a deep convolutional neural network (CNN). The CNN network performs multi-level feature extraction on the text, identifies various geometric features of the text, and converts them into a digital text style feature vector F. s .

[0035] The input "B" letter is converted into a set of geometric features through convolutional and pooling layers. These features include information such as the edges, line thickness, angles, and curvatures of the letters.

[0036] Step 3: 3D base model construction According to the parsed text style feature F sThe system automatically uses polygon modeling technology to mesh the text outline. Specifically, the Delaunay triangulation algorithm is used to convert the text outline into a preliminary three-dimensional polygon mesh, and then the surface fitting algorithm is used to smooth the mesh to generate a smooth three-dimensional font model.

[0037] Each edge of the letter is meshed. For example, the surface of the letter "B" is processed to generate a smooth rounded curve boundary, removing the corners and edges that may appear in the traditional modeling process.

[0038] Step 4: Texture mapping and detail enhancement The generated 3D font model is texture mapped using a deep convolutional neural network (CNN). The system first generates a preliminary texture using the preset textures in the texture library and maps it to the surface of the 3D font model. Then, the generated texture is optimized using a generative adversarial network (GAN) to automatically repair texture seams and fine lines. Assuming that the user requires the generated "B" letter to have a metallic luster effect, the system uses a GAN network to optimize the metal texture, automatically repair the seams in the texture map, and make the texture fit the 3D font more naturally.

[0039] Step 5: Ray tracing and lighting effects simulation Ray tracing technology is applied to the generated 3D word model to simulate the lighting effects in the real world, including reflection, refraction and diffuse reflection. By calculating the interaction between light and the surface of the object, the system can adjust the surface lighting and shadow effects of the 3D word model in real time to ensure the realism of the model under different lighting conditions.

[0040] Assuming the model is in a virtual scene, the light source consists of a fixed point light source and a variable directional light source. The ray tracing algorithm simulates the details of the metal surface reflection and the shadow effect of the letters by calculating the interaction between the light source and the surface of the object.

[0041] Step 6: Environmental Perception and Adaptive Adjustment Environmental perception technology automatically perceives the environmental characteristics of the current scene by analyzing the lighting distribution, color distribution and material properties in the virtual scene in real time. The system dynamically adjusts the color, material and glossiness of the 3D word model based on the perceived environmental characteristics to ensure that the model blends naturally into the scene.

[0042] Assuming that the background in the virtual environment is a dark city night scene with weak lighting, the system will automatically adjust the color of the model based on the environmental perception results to make it brighter, so as to ensure that the text maintains high visibility in the virtual scene.

[0043] Step 7: Multi-view rendering and consistency verification After the 3D word model is generated, the system will render it from multiple perspectives. Specifically, the model will be rendered from multiple perspectives (such as from the top, side, oblique side, etc.), and then the consistency of the rendering results under different perspectives will be verified through the mean square error algorithm to ensure that the lighting and texture effects of the 3D word model are consistent under all perspectives.

[0044] Users can adjust the viewing angle in the virtual environment, and the system will automatically verify the rendering effects under multiple perspectives and ensure that there are no lighting inconsistencies or texture damage.

[0045] Step 8: Export the 3D font model Finally, the generated 3D font model can be rendered and displayed in real time through a computer graphics engine (such as OpenGL or DirectX), or exported to OBJ, FBX standard 3D file formats for use in other design software or virtual environments.

[0046] Please see attached Figure 2 The present invention also provides a computer-readable storage medium, wherein: The program code includes a text style information parsing module, which is used to receive input text style information, parse the text style information through an artificial intelligence algorithm, and extract text style features; The program code also includes a three-dimensional matrix generation module, which is used to generate a three-dimensional matrix model based on text style features using polygon modeling technology, and optimize the three-dimensional matrix model through a surface modeling method; The program code further includes a texture mapping and detail enhancement module for performing high-precision texture mapping on a three-dimensional base model through a deep convolutional neural network (CNN), optimizing texture effects using a generative adversarial network (GAN), and further performing detail enhancement through a deep convolutional neural network (DCNN); The program code further includes a ray tracing and lighting simulation module for applying ray tracing technology to simulate lighting effects, including reflection, refraction and diffuse reflection, according to the geometric shape of the three-dimensional character model and the position of the light source; The program code also includes an environment perception and adaptive adjustment module for real-time perception of the lighting, color distribution and material properties in the virtual scene, and automatically adjusting the color, material and reflectivity of the three-dimensional character model to make it consistent with the lighting and background characteristics of the virtual environment; The program code also includes a multi-view rendering and consistency verification module, which is used to render the three-dimensional character model from different angles and use the mean square error (MSE) to verify the consistency of the rendering results under different perspectives, ensuring that the visual effect of the three-dimensional character model remains consistent under all perspectives.

[0047] Computer-readable storage media include, but are not limited to, hard disks, solid-state drives, optical disks, USB flash drives, SD cards, cloud storage, or other media with storage capabilities.

[0048] The program code may also be executed by a programmable logic device, an integrated circuit (IC), or a processor (eg, a CPU, a GPU, an FPGA, etc.) to implement the method of the present invention.

[0049] The program code may include an application programming interface (API) and interface code for interacting with other program components to facilitate collaboration with other software and hardware devices.

[0050] The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory), erasable programmable read-only memory (Erasable Programmable Read-Only Memory), programmable read-only memory (Programmable Red-Only Memory), read-only memory (Read-Only Memory), magnetic storage, flash memory, magnetic disk or optical disk.

[0051] Experimental setup 1. Experimental environment Hardware configuration: Processor: Intel Core i9-11900K 8 cores 16 threads GPU: NVIDIA RTX3080 Memory: 64GB RAM Storage: 1TB SSD Operating system: Windows 10 Pro 64-bit Software Configuration: Programming language: Python 3.8 Deep learning framework: TensorFlow 2.6 Computer graphics engine: Blender 2.93 (for 3D rendering) Image processing library: OpenCV4.5 Generative Adversarial Networks (GAN): Implementation with TensorFlow and Keras Experimental platform: A three-dimensional landscape character model generation system is adopted based on the technical solution in the present invention.

[0052] The comparison system uses traditional 3D modeling tools (such as Autodesk Maya or Blender) and a 3D font model generation method based on manually set texture and lighting parameters.

[0053] 2. Experimental content and steps Experiment 1: Comparison of visual effects of generated models Experimental purpose: To compare the difference between the method of the present invention and the traditional method in the visual effect of the three-dimensional character model.

[0054] Experimental steps: Sample selection: Select multiple glyph samples (such as "BRAND", "PRODUCT") for testing.

[0055] Method 1 (traditional method): Use traditional 3D modeling tools to manually set the font shape and texture, and generate a 3D font model by setting static lighting and simple texture mapping.

[0056] Method 2 (method of the present invention): Use the method of the present invention to automatically parse the text style and generate a three-dimensional character model through deep learning, and apply ray tracing and environmental perception to automatically adjust the lighting, color and material.

[0057] Comparison content: Compare the visual effects, detail performance (such as gloss, texture clarity) and consistency of the generated models under different lighting conditions (strong light, weak light, dynamic light source).

[0058] Evaluation indicators: Detail: Compare each model's gloss, surface texture, and reflections using a visual scoring criteria.

[0059] Consistency: Compare the consistency of models under multi-view rendering, especially under different lighting conditions.

[0060] Experimental results: Traditional method: When the lighting conditions change, the details of the surface texture of the generated three-dimensional character model are lost, the reflection effect is relatively stiff, and the display effect is inconsistent under strong light and weak light conditions.

[0061] The method of the present invention: when the lighting conditions change, the generated three-dimensional character model can adjust the gloss, reflection properties and color in real time, so that the display effect under strong light, weak light and dynamic light conditions is consistent, and the details are more rich and delicate.

[0062] Table 1 Lighting adaptability: measures the model's ability to adapt to different lighting conditions.

[0063] Texture clarity: The fineness of the texture, ranging from low (50%) to high (100%).

[0064] Reflection effect: the strength of the reflection effect on the model surface.

[0065] Multi-view consistency: The lighting and texture consistency of the model under multiple viewpoints.

[0066] Experiment 2: Generation speed comparison Experiment purpose: To evaluate the efficiency of the method of the present invention in the process of generating three-dimensional character models compared with the traditional method.

[0067] Experimental steps: Method 1 (traditional method): Use traditional 3D modeling tools to manually build models, perform texture mapping, light settings, and render.

[0068] Method 2 (method of the present invention): Use the automated three-dimensional character generation system of the present invention to automatically parse the text style and generate a three-dimensional character model, and perform ray tracing, texture mapping and detail enhancement.

[0069] Evaluation indicators: Time efficiency: The total time from text input to generating a 3D character model is calculated for each method.

[0070] Experimental results: Traditional method: In the process of generating the model, it takes a long time to manually set and adjust the lighting, texture, material and other steps. It takes about 20-30 minutes to generate a complex glyph model.

[0071] The method of the present invention: through an automated process, it only takes about 5-7 minutes to generate a complex three-dimensional character model, which is significantly faster. Experimental methods Generation time (one minute) Time boost Embodiment 1 20-30 minutes - Experiment 2 5-7 minutes Increase by about 70%

[0072] Table 2 Generation time: the time from inputting text to generating a complete 3D character model.

[0073] Time Improvement: The percentage by which the generation speed is improved compared to traditional methods.

[0074] Model complexity: The amount of computation required to generate the model based on the complexity of the input text.

[0075] Experiment 3: Environmental adaptability test of the model Experiment purpose: to test the environmental adaptability of the method of the present invention in generating three-dimensional character models under different virtual scenes and lighting conditions.

[0076] Experimental steps: Virtual scene design: Design different virtual scenes, including light intensity, material properties, and background environment. The scenes include bright urban environments, dim night scenes, and virtual rooms with dynamic light source changes.

[0077] Method 1 (traditional method): Use traditional 3D modeling tools to generate models and manually adjust the lighting and materials in the scene.

[0078] Method 2 (method of the present invention): Use the method of the present invention to generate a three-dimensional character model, and use environmental perception technology to automatically adjust the color and material of the character model to adapt it to different virtual environments.

[0079] Evaluation indicators: Adaptability: Compare the visual effects of the generated 3D font models in different scenes, including the adaptability of the model’s color, gloss, and reflection properties to environmental changes.

[0080] Experimental results: Traditional methods: In different scenarios, the three-dimensional character models generated by traditional methods have poor adaptability to changes in lighting and materials, and the generated models are prone to produce inconsistent effects with the environment under complex backgrounds.

[0081] The method of the present invention: the generated three-dimensional character model can automatically adjust its color, glossiness and reflection properties according to the lighting, color and background in the virtual environment, ensuring a high degree of integration between the model and the environment.

[0082] Table 3 Through the comparison of experimental data, it can be clearly seen that the method of the present invention has obvious advantages over the traditional method in terms of environmental adaptability, adaptive adjustment speed and detail expression, especially in the adaptability to changes in lighting conditions, dynamic light sources and different virtual scenes. Based on environmental perception technology and automatic lighting adjustment, the method of the present invention can generate more realistic and natural three-dimensional word models, with stronger application flexibility and wide adaptability, and is particularly suitable for application in multiple fields such as advertising design, virtual reality, and augmented reality.

[0083] in conclusion: Lighting adaptability and detail expression: The method of the present invention ensures higher visual quality under different lighting conditions by dynamically adjusting texture, gloss and reflection effects, improves texture clarity and reflection effects, and has significant advantages in multi-perspective consistency.

[0084] Generation speed: The method of the present invention greatly improves the generation speed of three-dimensional character models. Compared with traditional methods, the efficiency is improved by about 70%-80%, and the generation of complex models can be completed more quickly.

[0085] Environmental adaptability: Through environmental perception and adaptive adjustment, the method of the present invention enables the three-dimensional character model to quickly respond to changes in lighting, color and material in the virtual environment, demonstrating stronger environmental adaptability.

[0086] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for generating a three-dimensional landscape character model, characterized in that: The method comprises the following steps: Step 1: Provide input text style information; Step 2: parse the text style information through an artificial intelligence algorithm to extract text style features; Step 3: Automatically generate a three-dimensional matrix model based on the text style features; Step 4: Using polygonal modeling technology to perform preliminary modeling on the three-dimensional matrix; Step 5: Applying a surface modeling method to the polygonal modeling result to generate a smooth surface; Step 6: Mapping and enhancing the detail texture of the three-dimensional matrix model through a deep neural network; Step 7: Using ray tracing technology to simulate the light and shadow of the three-dimensional matrix model under different lighting conditions; Step 8: Automatically adjust the color, material and reflection properties of the three-dimensional character model through environmental perception technology to make it blend with the lighting conditions and environment in the virtual scene.

2. The method for generating a three-dimensional landscape character model according to claim 1, characterized in that: The text style information includes font, thickness, curve and appearance features. The text style features are parsed through a deep convolutional neural network, and the obtained text style features are used to construct a three-dimensional matrix model.

3. The method for generating a three-dimensional landscape character model according to claim 1, characterized in that: The three-dimensional matrix model is generated by polygonal modeling technology, which specifically includes meshing the text contour information and using surface fitting technology to smoothly optimize the generated polygonal model, and finally generating a three-dimensional font model.

4. The method for generating a three-dimensional landscape character model according to claim 1, characterized in that: The three-dimensional matrix model is generated by polygonal modeling technology, which specifically includes meshing the text contour information and using surface fitting technology to smoothly optimize the generated polygonal model, and finally generating a three-dimensional font model.

5. The method for generating a three-dimensional landscape character model according to claim 1, characterized in that: The ray tracing technique is used to simulate reflection, refraction and diffuse reflection phenomena under lighting conditions.

6. The method for generating a three-dimensional landscape character model according to claim 1, characterized in that: The environmental perception technology automatically adjusts the color, reflectivity and glossiness of the three-dimensional character model to make it consistent with the surrounding environment through real-time perception of the light distribution, color distribution and material properties in the virtual scene.

7. The method for generating a three-dimensional landscape character model according to claim 1, characterized in that: The detail enhancement of the three-dimensional character model is performed through a deep convolutional neural network, specifically including multi-scale learning of the surface texture of the letters, thereby enhancing the performance of the letter edges, surface gloss and material.

8. The method for generating a three-dimensional landscape character model according to claim 1, characterized in that: The method further includes performing multi-view rendering on the generated three-dimensional landscape character model, and using mean square error to verify the consistency of rendering results at different viewpoints.

9. The method for generating a three-dimensional landscape character model according to claim 1, characterized in that: The method is applicable to advertising design, virtual scene modeling, augmented reality and virtual reality application scenarios, in which the three-dimensional character model can be adaptively adjusted according to changes in different environments.

10. A computer-readable storage medium having program code stored thereon, characterized in that: When the program code is executed, the three-dimensional landscape character model generation method according to any one of claims 1 to 9 can be implemented.

Images