A night vision image generation method based on a real-time graphics engine

Abstract

The application discloses a night vision image generation method based on a real-time graphics engine, which comprises the following steps: establishing a first three-dimensional model of a terrain in a scene, establishing a second three-dimensional model of buildings of different heights in the terrain, adding a texture map and material parameters respectively, importing the two models into a real-time engine, matching the two models, creating a sky box according to actual requirements of the scene, establishing a faint light night vision coloring model for the two models, simultaneously generating a night vision green and night vision noise effect, defining a 850nm material interface for the faint light night vision coloring model, integrating the two models and the sky box together, superimposing the night vision green and night vision noise effect, and completing the construction of a night vision simulation scene; the application realizes the night vision effect of basic material under different night illuminations, and preliminarily explores the application of the faint light night vision image generation technology in practice.

Description

Technical Field

[0001] This application relates to the field of night view image generation technology, specifically to a night view image generation method based on a real-time graphics engine. Background Technology

[0002] Realistic image generation is an important branch of computer graphics research. Currently, visible light image generation technology is relatively mature, producing images with high realism and vivid contrast. Image generation technology based on the BRDF (Bidirectional Reflectance Distribution Function) shading model of microsurface theory can achieve photorealistic rendering effects. Low-light night vision image generation technology refers to generating images of night vision goggles through computer simulation based on computer graphics, the reflectivity of objects, and material databases. However, current domestic real-time graphics engine image generation technology can only simulate visible light images. This prevents the application of simulator systems developed based on real-time graphics engines in nighttime counter-terrorism training, flight training, border defense operations, and other fields. Nighttime scenes are typically low-light environments, and visible light image generation technology cannot accurately simulate low-light night vision images, thus failing to present realistic nighttime scene effects. Especially in military simulations or combat games, if real-time graphics engines cannot simulate low-light night vision images, it may affect the effectiveness of nighttime combat and reconnaissance. Summary of the Invention

[0003] The purpose of this application is to address the above problems by providing a night view image generation method based on a real-time image engine, comprising the following steps:

[0004] The terrain in the required scene is modeled in 3D to form a first 3D model. Texture maps and material parameters are added to the surface of the first 3D model. The texture maps are used to represent the basic properties of the surface of the first 3D model, such as the background color, bump map, and specular map. The material parameters refer to the roughness of the material surface.

[0005] Three-dimensional models are created for buildings of different heights in the terrain to form second three-dimensional models of different heights. The texture maps and material parameters are then added to the surface of the second three-dimensional models.

[0006] The first 3D model and multiple second 3D models are imported into a real-time graphics engine to achieve matching between the second 3D models of different heights and the first 3D model.

[0007] A dynamic, real-time skybox is created based on the actual needs of the scene. The skybox is used to simulate the changes in light throughout the day and the brightness of the scene under different lunar phases.

[0008] For the first three-dimensional model, a low-light night vision shading model is established with the second three-dimensional model, and night vision green and night vision noise effects are generated. For the low-light night vision shading model, an 850nm material interface is defined, and the material interface is used to input the 850nm material reflectivity.

[0009] The first 3D model, the second 3D model, and the skybox are integrated together. The shading mode of the first 3D model and multiple second 3D models is adjusted to the low-light night vision shading model. The night vision green and the night vision noise effect are superimposed to complete the construction of the night vision effect simulation scene.

[0010] According to the technical solution provided in the embodiments of this application, before the step of performing three-dimensional modeling of the terrain in the scene to form a first three-dimensional model, the method further includes:

[0011] According to actual needs, the basic geographic data of the terrain to be constructed is obtained from the Internet and converted into a different format. The basic geographic data includes satellite image data and elevation data. Both the satellite image data and the elevation data are converted into a different format using GlobalMapper software.

[0012] According to the technical solution provided in the embodiments of this application, the second three-dimensional model includes a general model and a site model. The general model refers to a model established without relying on vegetation on the terrain during the creation process, and it is constructed using SpeedTree modeling software. The site model refers to a three-dimensional building model that can adapt to the terrain height and is created based on the elevation data and the satellite image data. The three-dimensional building model includes at least houses and is created using the three-dimensional modeling software Creator.

[0013] According to the technical solution provided in the embodiments of this application, in the step of performing three-dimensional modeling of the terrain in the required scene to form a first three-dimensional model, texture maps and material parameters are added to the surface of the first three-dimensional model, wherein the texture map is used to characterize the background color, bump map, and specular map of the basic properties of the surface of the first three-dimensional model; the material parameters refer to the roughness of the material surface. The step also includes:

[0014] The texture map is added to the surfaces of the first 3D model and the second 3D model using texture mapping technology.

[0015] According to the technical solution provided in the embodiments of this application, the sun and moon in the skybox are simulated by calling two direct light sources in the real-time graphics engine.

[0016] According to the technical solution provided in the embodiments of this application, the step of importing the first three-dimensional model and multiple second three-dimensional models into a real-time graphics engine to achieve matching between the second three-dimensional models of different heights and the first three-dimensional model includes:

[0017] Based on satellite imagery data, the spatial layout of the second 3D model is determined to ensure that the orientation, relative distribution, and relative height difference of each second 3D model are the same as those of the buildings in the satellite imagery data.

[0018] According to the technical solution provided in the embodiments of this application, after the step of laying out the space of the second three-dimensional model based on satellite image data to ensure that the orientation, relative distribution, and relative height difference of each second three-dimensional model are the same as the buildings in the satellite image data, the method further includes:

[0019] Perform node design, geometric shape design, and texture design.

[0020] Compared with existing technologies, the beneficial effects of this application are as follows: First, the terrain in the desired scene is 3D modeled to form a first 3D model. Then, texture maps and material parameters are added to the surface of the first 3D model. The texture map is used to represent the background color of the basic surface properties of the first 3D model, and the texture map includes at least the background color, bump map, and specular map. The material parameters refer to the roughness of the material surface. Then, buildings of different heights in the terrain are 3D modeled to form second 3D models of different heights. Texture maps and material parameters are also added to the surface of the second 3D models. Finally, the first 3D model and multiple second 3D models are imported into a real-time graphics engine to realize different heights. The second 3D model is matched with the first 3D model, and a dynamic real-time skybox is created according to actual needs to simulate the changes in light throughout the day and the brightness of the scene under different lunar phases. A low-light night vision shading model is established for the first 3D model and the second 3D model to generate night vision green and night vision noise effects. An 850nm material interface is defined for the low-light night vision shading model to input the material reflectivity of 850nm. Then, the first 3D model, the second 3D model and the skybox are integrated together, the shading mode of the first 3D model and multiple second 3D models is adjusted to the low-light night vision shading model, and night vision green and night vision noise effects are superimposed to complete the construction of the night vision effect simulation scene.

[0021] In the process of use, this application establishes a first three-dimensional model and second three-dimensional models at different heights. Texture maps and material parameters are added to the first and second three-dimensional models respectively. The texture maps and material parameters can represent the basic properties of the first and second three-dimensional models. The first and second three-dimensional models with added texture maps and material parameters are imported into a real-time graphics engine to ensure that the second three-dimensional models at different heights can match the first three-dimensional model. A dynamic real-time skybox is created according to the actual needs of the scene to simulate the changes in light throughout the day and the brightness of the scene under different lunar phases. A corresponding low-light night vision shading model is established to generate night vision green and night vision noise effects. The two models and the skybox are integrated together. Then, the shading mode of the first and second three-dimensional models is adjusted to the night vision shading model, and night vision green and noise effects are superimposed to complete the construction of the entire night vision effect simulation scene.

[0022] This application establishes a model, adds texture maps and material parameters to the model surface, and imports them into a real-time graphics engine for matching. It also creates a skybox and builds a low-light night vision shading model. During the model building process, a material interface is defined for inputting the material reflectivity of 850nm to simulate the material's reflection characteristics of near-infrared light and to simulate the spectral response gain of the low-light night vision goggles. This generates night vision green and night vision noise effects, which are then integrated to complete the construction of a night vision simulation scene. The low-light night vision imaging technology in this application can convert weak light into visible images, presenting realistic nighttime scene effects. This allows users to clearly see surrounding objects and scenes at night or in low-light environments, improving nighttime vision capabilities and significantly enhancing efficiency and safety in nighttime operations and security. Furthermore, in the military field, it facilitates soldiers' reconnaissance, search, and rescue missions in the dark. Attached Figure Description

[0023] Figure 1 A flowchart illustrating a night view image generation method based on a real-time image engine, provided in an embodiment of this application. Detailed Implementation

[0024] To enable those skilled in the art to better understand the technical solution of this application, the application will be described in detail below with reference to the accompanying drawings. The description in this section is only exemplary and explanatory, and should not be used to limit the scope of protection of this application in any way.

[0025] like Figure 1 As shown, Figure 1 A flowchart illustrating a night view image generation method based on a real-time graphics engine provided in this application;

[0026] The application environment of this example is the development of a night vision flight simulation system for night vision training. The flight simulation equipment developed based on this low-light night vision image generation technology, combined with a dual-channel projector for visible light and near-infrared light, can support pilots' needs for flight simulation training in a dark room using either naked-eye or low-light night vision goggles. Low-light night vision image generation technology is the core technology of a night vision flight simulation system, and its rendering effect significantly impacts the pilot's immersive experience and the effectiveness of night vision training. Based on this method, the simulation of night vision effects in simple scenarios can be achieved. Specific steps are as follows:

[0027] S1. Perform 3D modeling on the terrain in the required scene to form a first 3D model. Add texture maps and material parameters to the surface of the first 3D model. The texture map is used to characterize the basic properties of the surface of the first 3D model. The texture map includes at least a background color, a bump map, and a specular map. The material parameters refer to the roughness of the material surface.

[0028] S2. Perform 3D modeling on buildings of different heights in the terrain to form second 3D models of different heights, and add texture maps and material parameters to the surface of the second 3D models;

[0029] S3. Import the first 3D model and multiple second 3D models into the real-time graphics engine to achieve matching between the second 3D models of different heights and the first 3D model;

[0030] S4. Create a dynamic skybox according to the actual needs of the scene. The skybox is used to simulate the changes in light throughout the day and the brightness of the scene under different lunar phases.

[0031] S5. For the first three-dimensional model, establish a low-light night vision shading model with the second three-dimensional model, and generate night vision green and night vision noise effects;

[0032] S6. Integrate the first 3D model, the second 3D model, and the skybox together, adjust the shading mode of the first 3D model and multiple second 3D models to a night vision shading model, and overlay the night vision green and the night vision noise effect to complete the construction of the night vision effect simulation scene.

[0033] S1. Perform 3D modeling on the terrain in the required scene to form a first 3D model. Add texture maps and material parameters to the surface of the first 3D model. The texture map is used to characterize the basic properties of the surface of the first 3D model. The texture map includes at least a background color, a bump map, and a specular map. The material parameters refer to the roughness of the material surface.

[0034] Specifically, in this embodiment, the basic geographic data of the terrain to be constructed is obtained from the Internet according to actual needs and the format is converted. The basic geographic data selected in this embodiment is the Ledong Airport area, with elevation data of 150m-377m and satellite image resolution of 0.5m.

[0035] Import the downloaded satellite imagery data into the GIS map processing software Global Mapper in PNG format, and configure it according to the maximum terrain size of 1531*1531 (m) supported by the UE4 level flow. 2 Starting from the bottom left corner, the satellite imagery is cropped, segmented, and exported. Any satellite images with excessive white borders are deleted after cropping. The qualified satellite images are then imported into the "Content" directory of the corresponding project resource management in UE4. The UVs of the satellite images are scaled and offset accordingly to create material spheres. Elevation data is converted using Global Mapper software, cropped, segmented, and height-set using Word Machine software, and exported in PNG format supported by UE4. The exported elevation data is imported into UE4 using a level flow approach. During the import process, the imported height values ​​are calculated based on the elevation data to ensure that the generated terrain height matches the actual height. Simultaneously, the created satellite image material spheres are added to the elevation data in a specific order to complete the terrain modeling in the scene, thus forming the first 3D model.

[0036] S2. Perform 3D modeling on buildings of different heights in the terrain to form second 3D models of different heights, and add texture maps and material parameters to the surface of the second 3D models;

[0037] Specifically, based on the different functions of the 3D model of the terrain surface, the 3D model is divided into general model and site model. The general model refers to the model that does not depend on the terrain during the creation process. It usually contains a single object and can be placed in a large number of places in the scene, such as trees and grass. The site model refers to the model that can adapt to the terrain height based on elevation and satellite imagery. It is usually represented by a model group. Its layout is consistent with the satellite imagery, and the height difference between objects also conforms to the terrain trend, such as houses, airports, and ports.

[0038] The general-purpose model of trees, grass, and other plants is built using the 3D modeling software SpeedTree. The process begins by creating the trunk, adjusting its curvature, height, and thickness to ensure its overall shape resembles a real tree. Large branches are then added to the trunk, small branches to the large branches, twigs to the small branches, and leaves to the twigs. The thickness, curvature, and number of these elements are adjusted sequentially to complete the basic geometric model of the tree. Next, texture maps are created for the trunk, branches, and leaves. These texture maps include a base color, bump maps, and specular maps, which are then applied to the trunk, branches, and leaves to complete the overall tree model.

[0039] The site model was created using the 3D modeling tool Creator. The 3D terrain with height information generated by the terrain modeling software TerraVista was exported in .flt format. Opening the terrain file in Creator revealed a 3D terrain model with consistent height differences and terrain textures (satellite imagery). The spatial layout of the 3D model was designed based on the satellite imagery of the terrain surface, ensuring that the orientation, relative distribution, and relative height differences of the model matched the buildings in the satellite imagery. Then, node design was performed on the entire model. Node design refers to representing the relationships between model entities using a tree structure during the modeling process. The relationships between root nodes, group nodes, volume nodes, and face nodes in the model were represented using a tree structure for easy subsequent management. The root node represents the entire model group. The creation process begins with group nodes, which categorize models within a scene according to certain rules. Volume nodes represent individual objects within a group node, while face nodes comprise all the basic geometric primitives that make up a single 3D model. Subsequently, the geometric appearance of the 3D model is designed based on the points, lines, and faces of these basic primitives. Geometric shape design refers to the design of the 3D model's geometric appearance based on the building shapes at corresponding locations in satellite imagery, ensuring that the designed model's geometric shape closely resembles the shapes of buildings in the real world. Finally, the base color texture, specular texture, bump map, transparency map, and material parameters of the model's surface are designed according to local cultural characteristics, and texture mapping technology is used to attach the textures to the 3D model's surface. Texture mapping technology refers to adding textures and RGB parameter values ​​that represent the appearance and material of the 3D model to the surface of the 3D geometric model, incorporating local cultural styles.

[0040] After the above operations, the modeling of the second 3D model is completed, and the second 3D model is exported in .FBX format supported by the real-time graphics engine.

[0041] S3. Import the first 3D model and multiple second 3D models into the real-time graphics engine to achieve matching between the second 3D models of different heights and the first 3D model;

[0042] Specifically, in this example, when the first 3D model is created, it has been imported into the real-time graphics engine. When the second 3D model is imported, multiple second 3D models created by the 3D modeling software Creator are imported into the real-time graphics engine UE4 in FBX format. The position, height, and orientation of the second 3D models are adjusted to match the first 3D model.

[0043] S4. Create a dynamic, real-time skybox based on the actual needs of the scene. The skybox is used to simulate the changes in light throughout the day and the brightness of the scene under different lunar phases.

[0044] Specifically, a dynamic, real-time skybox is created based on the actual needs of the scene. In this embodiment, the dynamic skybox is implemented using the Ultra Dynamic Sky plugin downloaded from the Unreal Engine official store (https: / / www.unrealengine.com / marketplace / zh-CN / product / ultra-dynamic-sky). The Ultra Dynamic Sky plugin can simulate the changes in lighting throughout a 24-hour day and the brightness of the scene under typical lunar phases.

[0045] In practice, the lighting of the sun and moon in the skybox is simulated by calling two direct light sources in UE4. The corresponding parameter interfaces in the plugin are adjusted to simulate the changes in the entire sky environment throughout 24 hours. The brightness of the night sky background and the intensity of the moon's light are adjusted to simulate the lighting effects under the conditions of no moon, half moon, and full moon.

[0046] S5. For the first three-dimensional model, establish a low-light night vision shading model with the second three-dimensional model, and generate night vision green and night vision noise effects;

[0047] Specifically, the low-light night vision shading model refers to the mathematical modeling of the process during rendering where light emitted from a light source is attenuated to reach the object's surface, interacts physically with the object's material, and then leaves the object according to certain reflection laws, before being magnified by a low-light night vision lens. The design of the low-light night vision shading model allows for the correlation between the object's surface texture map and material parameters and the ambient lighting within the skybox, and can express the light reflection characteristics of the object's surface material. The implementation of the entire low-light night vision shading model includes five modules: light color, light attenuation rate, light reflection characteristics of the object's surface material, low-light night vision lens response gain, and quantization of radiometric values ​​into grayscale.

[0048] Among them, the lighting color is set to the color of the light emitted by the light source. In this example, the lighting color is set to the default value of 1. The lighting attenuation rate refers to the light attenuation law that the light follows from the light source to the object surface. The lighting attenuation rate follows the inverse square law of light attenuation when the light propagates uniformly in all directions. In order to avoid infinite light near the light source, 1 is added to the denominator of the attenuation rate. The light reflection characteristics of the object surface material refer to the simulation of the light reflection and reflection energy when the object surface material interacts with near-infrared light. By calculating the reflection characteristics, the lighting data can be linked with the texture map and material of the 3D model surface. When calculating the reflection characteristics of the night vision shading model, it is necessary to determine the material parameters such as the normal vector, background color, roughness, specular highlight, and near-infrared reflectivity of the material surface, as well as the BRDF shading model.

[0049] Material parameters such as normal vector, background color, roughness, and specular highlights are stored and retrieved through five channels: GBufferA, GBufferB, GBufferC, GBufferD, and GBufferE. Near-infrared reflectance requires a custom-defined material interface to input and store parameters for the material's reflectance in the 850nm band. Diffuse color in the night vision shading model is represented by multiplying the background color (characterizing the object's inherent color) by the material's reflectance coefficient in the 850nm band. The interaction between near-infrared light and the material's surface roughness and normal vector is handled by superimposing Lambert diffuse reflection onto a graph composed of the GGX normal distribution function and an approximate SmithJoint shadow. A specular reflection model composed of a shadow occlusion function and the Schlick Fresnel function is implemented. In implementing specular reflection, the GGX, approximate SmithJoin, and Schlick functions are calculated separately, and then superimposed to calculate the resulting specular highlight effect. GGX is a micro-surface-based reflection model that can be used to achieve specular reflection; the specific steps for implementing specular reflection are existing technology and will not be elaborated here. The SmithJoint shadow occlusion function is a function used to calculate shadow occlusion; it is based on a combination of the Smith occlusion function and the Joint occlusion function. The Smith occlusion function calculates the degree of occlusion by an object, while the Joint occlusion function considers the interaction between multiple objects; both are existing technology. The Schlick Fresnel function is a formula for calculating the intensity of reflected light, used to calculate the intensity of reflected light when light travels from one medium to another, especially when the light is perpendicular or nearly perpendicular to the surface; it is also existing technology.

[0050] The low-light night vision goggle response gain refers to the simulated value of the low-light night vision goggle response gain determined by analyzing the brightness of the display, the difference in brightness between pixels in the entire scene at night and during the day after engine rendering, and the brightness values. The final simulated value is set to 200. Finally, the diffuse reflection lighting data, specular reflection lighting data, and data related to material properties such as normal vectors and roughness are returned to the parallel lighting function. The parallel lighting function is a lighting model that simulates the effect of light from infinity illuminating an object. It is a prior art technology. The parallel lighting function calculates the lighting of a point in the camera viewport based on the object's position coordinates, camera position, Gbuffer data, and light attenuation rate, completing the lighting calculation in the night vision rendering scene. For example, the brightness of a leaf surface on a 2D screen is the result of the parallel lighting function calculating the lighting at that point. The location of the point, its position in the camera viewport, the material data corresponding to the point, and the propagation of light are used to calculate the presentation of the point on the 2D screen. Quantization grayscale refers to the desaturation of the lighting in the scene, retaining only the intensity value of the lighting. In implementation, a brightness extraction formula is used to desaturate the colors in the RGB color space, retaining only the brightness value. The brightness extraction formula usually uses a weighted average value for calculation. Considering all aspects of the response gain of the low-light night vision goggles, the simulated value is set to 200. According to radiometric conversion into grayscale, the brightness value is extracted by performing the operation R1 = 0.299R, G1 = 0.587G, B1 = 0.114B, Y = R1 + B1 + C1 in the RGB color space. R, G, and B refer to the values ​​of the three primary colors of the color space: red, green, and blue, respectively. Y refers to the extracted brightness value (the formula comes from color space conversion).

[0051] In this embodiment, night vision green and night vision noise are designed and implemented by obtaining Gbuffer data during the rendering process and cached data during the deferred rendering process through UE4's post-processing. When implementing night vision green, a new material is first created using Blueprint, and the material's rendering mode is selected as post-processing material. The SceneTexture: Post-processing Input 0 node is used to obtain the cached values ​​of the four channels R, G, B, and Alpha during the scene rendering process. The Alpha channel value is occluded by the Mask node, leaving only the R, G, and B channel values. The value processed by the Mask node is multiplied with the green texture and connected to the self-illuminating color interface, so that the output scene color is converted into the display effect of a single green image based on the brightness value of the material surface, thereby simulating the imaging effect of night vision green.

[0052] When implementing night vision noise reduction, a new material is created using a blueprint, and the rendering mode is set to post-processing. The ScreenAlignedUVs node is used for X and Y scaling to ensure the output texture UVs fit the screen, allowing the noise simulation effect to be displayed only across the entire window. The Scalar parameter node can access and modify material instances and dynamic code. Calling this node and defining its attribute as Grain_tiling allows the Grain_tiling node to be multiplied by the ScreenAligenedUVs node. Adjusting the Grain_tiling value controls the amount of noise texture on the entire screen. Noise under low-light night vision goggles exhibits three-dimensional characteristics, adding a time dimension to the original noise. Therefore, during the design process, the Scalar parameter node is called, and its attribute is defined as Refresh_Rate. Multiplying this node by the time node controls the noise response frequency on the screen. The Append node integrates the two channels from the above four nodes into one channel, and the channel value is passed to the Vector. The Noise node is used, followed by the SceneTexture: post-processing input 0 node and Mask node to obtain the scene rendering effect. The output values ​​of the Mask node and Desaturation node are linearly interpolated with the noise gain node to control the amount of noise. Finally, the processed noise effect is passed to the self-illuminating color interface to complete the simulation of night vision noise.

[0053] S6. Integrate the first three-dimensional model, the second three-dimensional model, and the skybox together, adjust the shading mode of the first three-dimensional model and multiple second three-dimensional models to a low-light night vision shading model, and overlay the night vision green and the night vision noise effect to complete the construction of the night vision effect simulation scene;

[0054] Specifically, the first 3D model, the second 3D model, and the skybox are placed in the same level, arranged from bottom to top in space. The shading method of the first and second 3D models in the scene is modified to a low-light night vision shading model. The reflectance coefficient of the object surface material in the near-infrared band is added and set. The reflectance coefficient is derived from the reflectance of the material in the 850nm band in the material database. The designed night vision green and night vision noise materials are added to the post-processing materials and the application range is set to no boundary. At the same time, the light intensity and time parameters in the skybox are adjusted to simulate the night vision effect of the scene under different lunar phases and time nodes, so as to achieve a complete simulation of the night vision effect.

[0055] Furthermore, prior to the step of performing three-dimensional modeling of the terrain in the scene to form the first three-dimensional model, the method further includes:

[0056] According to actual needs, the basic geographic data of the terrain to be constructed is obtained from the Internet and converted into a different format. The basic geographic data includes satellite image data and elevation data. Both the satellite image data and the elevation data are converted into a different format using GlobalMapper software.

[0057] Specifically, in this embodiment, the basic geographic data selected is the Ledong Airport area. The basic geographic data includes elevation data and satellite imagery data. The elevation data is 150m-377m, and the resolution of the satellite imagery is 0.5m. Both the satellite imagery data and the elevation data are converted using Global Mapper software.

[0058] Furthermore, the second three-dimensional model includes a general model and a site model. The general model refers to a model established without relying on vegetation on the terrain during the creation process, and it is constructed using SpeedTree modeling software. The site model refers to a three-dimensional building model that can adapt to the terrain height, created based on the elevation data and the satellite image data. The three-dimensional building model includes at least buildings and is created using the three-dimensional modeling software Creator.

[0059] Specifically, based on the different functions of the second 3D model, it is divided into general models and site models. General models refer to models that do not depend on terrain during the creation process. They usually contain a single object and can be placed in large numbers in the scene, such as trees, grass, and special effects water. General models are constructed using SpeedTree modeling software. Site models refer to models that can adapt to terrain height based on elevation and satellite imagery. They are usually represented by a model group, whose layout is consistent with the satellite imagery. The height difference between objects also conforms to the terrain trend, such as houses, airports, and ports. Site models are created using 3D modeling software Creator.

[0060] Furthermore, in the step of performing 3D modeling of the terrain in the desired scene to form a first 3D model, texture maps and material parameters are added to the surface of the first 3D model. The texture maps are used to characterize the basic properties of the first 3D model's surface, including background color, bump mapping, and specular mapping. The material parameters refer to the surface roughness of the material. The step also includes:

[0061] Texture mapping technology is used to add textures to the surfaces of the first 3D model and the second 3D model.

[0062] Specifically, texture mapping technology is used to add textures to the surfaces of the first 3D model and the second 3D model. Texture mapping technology refers to adding textures and RGB parameter values ​​that represent the appearance and material of the 3D model to the surface of the 3D geometric model in combination with local cultural style.

[0063] Furthermore, the sun and moon in the skybox are simulated by invoking two direct light sources in the real-time graphics engine.

[0064] Specifically, a dynamic, real-time skybox is created based on the actual needs of the scene. In this embodiment, the dynamic skybox is implemented by using a super-dynamic sky plugin downloaded from the internet. The dynamic, real-time skybox is used to simulate 24-hour lighting changes and scene brightness under different lunar phases. The sunlight and moonlight in the skybox are simulated by calling two direct light sources in UE4. By adjusting the corresponding parameter interfaces in the plugin, the entire weather environment changes within a 24-hour period can be simulated. Adjusting the brightness of the night sky background and the intensity of the moonlight can simulate the lighting environment under conditions of no moon, half moon, and full moon.

[0065] Furthermore, the step of importing the first 3D model and multiple second 3D models into a real-time graphics engine to achieve matching between the second 3D models of different heights and the first 3D model includes:

[0066] Based on satellite imagery data, the spatial layout of the second 3D model is determined to ensure that the orientation, relative distribution, and relative height difference of each second 3D model are the same as those of the buildings in the satellite imagery data.

[0067] It also includes node design, geometric shape design, and texture design.

[0068] Specifically, the spatial layout of the 3D model is designed based on satellite imagery of the terrain surface, ensuring that the orientation, relative distribution, and relative height difference of the model are the same as the buildings in the satellite imagery. Then, node design is performed on the entire model. Node design refers to representing the relationships between model entities using a tree structure during the modeling process. The relationships between root nodes, group nodes, volume nodes, and face nodes in the model are represented in a tree structure for easy subsequent management. The root node is the starting point for creating the entire group of models; group nodes are the beginning of classifying models within the scene according to certain rules; volume nodes are individual objects under group nodes; and face nodes are all the basic elements that make up a single 3D model. This geometric primitive is then used to design the geometric appearance of the 3D model based on the basic primitive points, lines, and faces. The geometric shape design refers to the design of the 3D model's geometric appearance based on the building shape at the corresponding location in satellite imagery, ensuring that the designed model's geometric shape is similar to the building shape in the real world. Finally, the base color map, specular map, bump map, transparency map, and material parameters of the model's surface are designed according to local cultural characteristics, and texture mapping technology is used to attach the textures to the 3D model's surface. Texture mapping technology refers to adding textures and RGB parameter values ​​that represent the appearance and material of the 3D model to the surface of the 3D geometric model in combination with local cultural style.

[0069] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. The above descriptions are only preferred embodiments of this application. It should be noted that due to the limitations of written expression, while there are objectively infinite specific structures, those skilled in the art can make several improvements, modifications, or changes without departing from the principles of this invention, and can also combine the above technical features in an appropriate manner. These improvements, modifications, changes, or combinations, or the direct application of the inventive concept and technical solution to other situations without modification, should all be considered within the scope of protection of this application.

Claims

1. A method for generating night view images based on a real-time graphics engine, characterized in that, Includes the following steps: The terrain in the required scene is modeled in 3D to form a first 3D model. Texture maps and material parameters are added to the surface of the first 3D model. The texture map is used to represent the basic properties of the surface of the first 3D model and includes at least a background color, a bump map, and a specular map. The material parameters refer to the roughness of the material surface. Three-dimensional models are created for buildings of different heights in the terrain to form second three-dimensional models of different heights. The texture maps and material parameters are then added to the surface of the second three-dimensional models. The first 3D model and multiple second 3D models are imported into a real-time graphics engine to achieve matching between the second 3D models of different heights and the first 3D model. A dynamic, real-time skybox is created based on the actual needs of the scene. The skybox is used to simulate the changes in light throughout the day and the brightness of the scene under different lunar phases. For the first three-dimensional model, a low-light night vision shading model is established with the second three-dimensional model, and night vision green and night vision noise effects are generated. For the low-light night vision shading model, an 850nm material interface is defined, and the material interface is used to input the 850nm material reflectivity. The first 3D model, the second 3D model, and the skybox are integrated together. The shading mode of the first 3D model and multiple second 3D models is adjusted to the low-light night vision shading model. The night vision green and the night vision noise effect are superimposed to complete the construction of the night vision effect simulation scene.

2. The night view image generation method based on a real-time graphics engine according to claim 1, characterized in that, Before the step of performing three-dimensional modeling of the terrain in the scene to form a first three-dimensional model, the method further includes: According to actual needs, the basic geographic data of the terrain to be constructed is obtained from the Internet and converted into a different format. The basic geographic data includes satellite image data and elevation data. Both the satellite image data and the elevation data are converted into a different format using GlobalMapper software.

3. The night view image generation method based on a real-time graphics engine according to claim 2, characterized in that, The second three-dimensional model includes a general model and a site model. The general model refers to a model created without relying on vegetation on the terrain during the creation process, and it is constructed using SpeedTree modeling software. The site model refers to a three-dimensional building model that can adapt to the terrain height, created based on the elevation data and the satellite image data. The three-dimensional building model includes at least buildings and is created using the three-dimensional modeling software Creator.

4. The night view image generation method based on a real-time graphics engine according to claim 1, characterized in that, The process involves creating a first 3D model by 3D modeling the terrain in the desired scene. Texture maps and material parameters are then added to the surface of this first 3D model. The texture maps represent the basic surface properties of the first 3D model, including background color, bump mapping, and specular mapping. The material parameters refer to the surface roughness of the material. The process also includes: The texture map is added to the surfaces of the first 3D model and the second 3D model using texture mapping technology.

5. The night view image generation method based on a real-time graphics engine according to claim 1, characterized in that, The sun and moon in the skybox are simulated by invoking two direct light sources in the real-time graphics engine.

6. The night view image generation method based on a real-time graphics engine according to claim 1, characterized in that, The step of importing the first 3D model and multiple second 3D models into a real-time graphics engine to achieve matching between the second 3D models of different heights and the first 3D model includes: Based on satellite imagery data, the spatial layout of the second 3D model is determined to ensure that the orientation, relative distribution, and relative height difference of each second 3D model are the same as those of the buildings in the satellite imagery data.

7. The night view image generation method based on a real-time graphics engine according to claim 6, characterized in that, The step of laying out the space of the second 3D model based on satellite imagery data, ensuring that the orientation, relative distribution, and relative height difference of each second 3D model are the same as the buildings in the satellite imagery data, further includes: Perform node design, geometric shape design, and texture design.

Images