Simulator real-time rendering scene generation method and system

Abstract

The invention relates to the field of computer graphic rendering, in particular to a simulation simulator real-time rendering scene generation method and system. The method comprises the following steps: acquiring peripheral terrain information and scene information through a detection unmanned aerial vehicle, and transmitting the terrain information and the scene information to an information processing terminal; the terrain data and the scene configuration file required by the training task are generated through the information processing terminal, the problem of deviation between the training effect and the actual battlefield condition is solved, meanwhile, terrain simulation and scene simulation are rendered separately, rendering details are dynamically adjusted through an algorithm, under the condition that high quality is guaranteed, computing resources and time are saved, and the training efficiency is improved. The rendering quality is improved, the quality and effect of simulated military training by training personnel are guaranteed, and the training efficiency is improved.

Description

Technical Field

[0001] The invention relates to the field of computer graphics rendering, and in particular to a method and system for generating a real-time rendering scene of a simulation simulator. Background Art

[0002] In the modern military training system, early simulated military training plays a vital role. With the rapid development of technology, training methods are constantly innovating to adapt to the complex and changing battlefield environment. In the early stages of simulated training, in order to efficiently utilize limited training resources, virtual reality (VR) technology is usually used more to construct simulated training content. This training method can not only save a lot of material and manpower costs, but also allow soldiers to familiarize themselves with the battlefield environment and combat procedures in advance through a highly simulated virtual environment. However, in conventional simulated training, direct satellite maps or high-altitude reconnaissance aerial photography are generally used, resulting in the simulation accuracy of the training simulation environment for the real environment is often not high enough, resulting in a deviation between the training effect and the actual battlefield situation. Secondly, real-time rendering requires a lot of computing resources and time. If low-standard real-time rendering is adopted, the real-time rendering effect will be poor, which will seriously reduce the quality and effect of simulated military training for trainees and reduce training efficiency. In view of this, this application is specially proposed. Summary of the invention

[0003] The technical problem to be solved by the present invention is that the use of direct satellite maps or high-altitude reconnaissance aerial photography often results in the simulation accuracy of the training simulation environment for the real environment being not high enough, resulting in a deviation between the training effect and the actual battlefield situation, and the use of low-standard real-time rendering will lead to poor real-time rendering effects, seriously reducing the quality and effect of simulated military training for trainees.

[0004] In order to solve the above technical problems, the present invention is implemented by the following technical solution: a method for generating a real-time rendering scene of a simulation simulator, the method comprising the following steps: Step S1: collecting surrounding terrain information and scene information through a detection drone, and transmitting the terrain information and the scene information to an information processing terminal; Step S2: Generate terrain data and scene configuration files required for the training task in the information processing terminal through the terrain information and the scene information, and initialize the rendering engine; Step S3: constructing a terrain basic model according to the terrain data and the scene configuration file, calculating a first visible area in the terrain basic model through a machine vision analysis algorithm, and rendering the first visible area; the first visible area is a visible area under an initial default viewing angle in a scene loaded by a trainee; Step S4: using a first learning rendering algorithm to generate the texture and layout of scene elements in the first visible area; Step S5: reading the viewpoint position of the training user, using a second learning rendering algorithm to render the first visible area to obtain an image result, and dynamically adjusting the scene detail level; Step S6: Output the image result to the VR visual device, and update the rendering content in real time according to the user's interactive operation.

[0005] Furthermore, the terrain data and scene configuration file required for the training task are generated in the information processing terminal, including: the terrain data and scene configuration file required for the training task are generated in the information processing terminal by using the Mesh surface fitting technology, the Mesh surface fitting technology is optimized by a self-organizing neural network intelligent algorithm, and the mathematical expression of the self-organizing neural network intelligent algorithm is: ; in, is the weight change of the xth neuron at the tth iteration; is the neural network learning rate; is the neighborhood function, which represents the neighborhood relationship between the y-th winning neuron and the x-th neuron; is the current input sample; is the weight vector of the xth neuron at iteration number t; is the current iteration number; The neural network learning rate is: ; in, is the gradient information of the current parameter; is a constant to ensure numerical stability; is the terrain information weight, is the scene information weight; is an extraordinary parameter used to control the step size of parameter update; The terrain information weight is: ; in, is an exponential decay coefficient, and 0.45> >0.9; To detect the percentage of mountain terrain collected by drones, To detect the percentage of hilly terrain collected by drones, To detect the proportion of flat terrain collected by the drone, To detect the water terrain ratio collected by the drone, To detect the valley terrain percentage value collected by the drone, is the proportion of terrain information collected by the detection drone at the t-1th iteration; is an exponentially weighted moving average; The scene information weight is: ; in, To detect the proportion of urban scenes collected by drones, To detect the proportion of forest scenes collected by drones, To detect the percentage of grassland shrub scenes collected by drones, To detect the proportion of other scenes collected by the drone, It is the percentage of scene information collected by the detection drone at the t-1th iteration.

[0006] Furthermore, the first visible area in the terrain basic model is calculated by a machine vision analysis algorithm, and the machine vision analysis algorithm uses a binocular matching algorithm, and its formula is: ; in, The area map needs to be rendered for the first visible area; The energy function corresponding to the area map to be rendered for the first visible area; is the left eye imaging area, is the imaging area of ​​the right eye; For the first visible area, other areas other than the left eye imaging area in the area map need to be rendered; The time cost and performance cost of rendering the area map for the first visible area, where the time cost is the unit time required to render the image area, and the performance cost is the amount of computing resources required to render the image area; is the disparity value of the left eye imaging area; is the disparity value of the right eye imaging area; is a penalty coefficient, which is applicable to the case where the disparity value of the left eye imaging area differs from the disparity value of the right eye imaging area by 1; It is also a penalty coefficient, which is applicable when the difference between the disparity value of the left eye imaging area and the disparity value of the right eye imaging area is greater than 1; It is an indicator function that returns 1 if the condition is true, otherwise it returns 0.

[0007] Furthermore, the first learning rendering algorithm includes: a texture rendering algorithm formula and a reflection algorithm formula, and the texture rendering algorithm formula is: ; in, is the color of the final pixel in the pixel area; is the weight of the i-th texture; is the color value of the i-th texture at the coordinate (j, k, l); is the horizontal coordinate in the coordinate (j, k, l), is the ordinate in the coordinates (j, k, l), is the height coordinate in the coordinates (j, k, l); is the total number of textures; The reflection algorithm formula is: ; in, is the brightness of the reflected light of the final pixel; is the ambient light intensity; is the ambient light reflection coefficient; is the main light intensity; is the diffuse reflectance; is the specular reflection coefficient; is the light source direction vector; is the surface normal vector; is the reflected light direction vector; is the viewing direction vector; is the specular highlight exponent.

[0008] Furthermore, the second learning rendering algorithm includes: a first rendering algorithm weight formula and a second rendering algorithm weight formula; the first rendering algorithm weight formula is used to calculate the rendering priority of the current terrain information, and the first rendering algorithm weight formula is: ; in, is the rendering weight of the corresponding terrain; is the complexity coefficient of the terrain; is the performance demand coefficient of the terrain; It is the rendering priority coefficient of the terrain; is the corresponding terrain, when =1 for mountainous areas, =2 for hills, =3 is flat ground, =4 is water area, =5 is a valley; The second rendering algorithm weight formula is used to calculate the rendering priority of the current scene information, and the second rendering algorithm weight formula is: ; in, is the rendering weight of the corresponding scene; is the complexity coefficient of the scene; is the performance requirement coefficient of the scenario; is the rendering priority coefficient of the scene; For the corresponding scene, when =1 for cities, =2 means forest, =3 is grassland shrubs, =4 means other types; Among them, when ≥ When , the details in the terrain information are rendered first. < When rendering the scene information, details in the scene information are given priority.

[0009] Further, the complexity coefficient of the terrain and the complexity coefficient of the scene are trained and optimized using a computer image training algorithm; the computer image training algorithm is trained using a computer image training algorithm formula, and the computer image training algorithm formula is: ; in, To optimize the weight ratio; is a first-order matrix; is a second-order matrix; is a constant, and 1≥ >0.

[0010] Based on the same inventive concept, on the other hand, the present invention also provides a simulation simulator real-time rendering scene generation system, the system comprising: a detection drone, an information processing terminal, an image simulation simulator, and a VR visual device; the detection drone transmits information to the information processing terminal, the information processing terminal transmits data and configuration files to the image simulation simulator, and the image simulation simulator transmits the generated image results to the VR visual device; the detection drone is used to obtain surrounding terrain information and scene information, the information processing terminal is used to process the terrain information and scene information obtained by the detection drone, and generate terrain data and scene configuration files required for training tasks through the terrain information and the scene information, the image simulation simulator is used to render the terrain data and the scene configuration files to generate the image results, and the VR visual device is used to present the rendered image results.

[0011] Furthermore, the image simulation simulator also includes: a first image information processing module and a second image information processing module, the terrain data and the scene configuration file obtain the first visible area through the first image information processing module, and the first visible area obtains the image result through the second image information processing module.

[0012] Compared with the prior art, the present invention has the following advantages and beneficial effects: using drones for low-altitude scanning to collect terrain information and scene information, a large amount of rich information data is collected, and terrain data and scene configuration files required for training tasks are generated through information processing terminals, thereby solving the problem of deviation between training effects and actual battlefield conditions. At the same time, by rendering terrain simulation and scene simulation separately, dynamically adjusting rendering details through algorithms, it is ensured that under high-quality conditions, computing resources and time are saved, rendering quality is improved, and the quality and effect of simulated military training for trainees is guaranteed, thereby improving training efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0014] Figure 1 A flowchart of a method for generating a real-time rendering scene of a simulation simulator provided in an embodiment of the present invention.

[0015] Figure 2 A module block diagram of a simulation simulator real-time rendering scene generation system provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0016] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with embodiments and drawings. The exemplary embodiments of the present invention and their description are only used to explain the present invention and are not intended to limit the present invention.

[0017] In the following description, a large number of specific details are set forth in order to provide a thorough understanding of the present invention. However, it is apparent to one of ordinary skill in the art that these specific details are not necessarily employed to practice the present invention. In other embodiments, in order to avoid obscuring the present invention, well-known structures, circuits, materials, or methods are not specifically described.

[0018] Throughout the specification, references to "one embodiment," "an embodiment," "an example," or "an example" mean that a particular feature, structure, or characteristic described in conjunction with the embodiment or example is included in at least one embodiment of the present invention. Therefore, the phrases "one embodiment," "an embodiment," "an example," or "an example" appearing in various places throughout the specification do not necessarily all refer to the same embodiment or example. In addition, particular features, structures, or characteristics may be combined in one or more embodiments or examples in any suitable combination and / or sub-combination. In addition, it will be appreciated by those of ordinary skill in the art that the figures provided herein are for illustrative purposes and that the figures are not necessarily drawn to scale. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0019] Example 1: Figure 1 As shown, this embodiment provides a method for generating a real-time rendering scene of a simulation simulator, and the method comprises the following steps: Step S1: collecting surrounding terrain information and scene information through a detection drone, and transmitting the terrain information and the scene information to an information processing terminal; Step S2: Generate terrain data and scene configuration files required for the training task in the information processing terminal through the terrain information and the scene information, and initialize the rendering engine; Step S3: constructing a terrain basic model according to the terrain data and the scene configuration file, calculating a first visible area in the terrain basic model through a machine vision analysis algorithm, and rendering the first visible area; the first visible area is a visible area under an initial default viewing angle in a scene loaded by a trainee; Step S4: using a first learning rendering algorithm to generate the texture and layout of scene elements in the first visible area; Step S5: reading the viewpoint position of the training user, using a second learning rendering algorithm to render the first visible area to obtain an image result, and dynamically adjusting the scene detail level; Step S6: Output the image result to the VR visual device, and update the rendering content in real time according to the user's interactive operation.

[0020] Furthermore, the terrain data and scene configuration files required for generating the training task in the information processing terminal are generated by Mesh surface fitting technology, and the Mesh surface fitting technology is optimized by a self-organizing neural network intelligent algorithm, and the mathematical expression of the self-organizing neural network intelligent algorithm is: ; in, is the weight change of the xth neuron at the tth iteration; is the neural network learning rate; is the neighborhood function, which represents the neighborhood relationship between the y-th winning neuron and the x-th neuron; is the current input sample; is the weight vector of the xth neuron at iteration number t; is the current iteration number; The neural network learning rate is: ; in, is the gradient information of the current parameter; is a constant to ensure numerical stability; is the terrain information weight, is the scene information weight; is an extraordinary parameter used to control the step size of parameter update; The terrain information weight is: ; in, is an exponential decay coefficient, and 0.45> >0.9; To detect the percentage of mountain terrain collected by drones, To detect the percentage of hilly terrain collected by drones, To detect the proportion of flat terrain collected by the drone, To detect the water terrain ratio collected by the drone, To detect the valley terrain percentage value collected by the drone, is the proportion of terrain information collected by the detection drone at the t-1th iteration; is an exponentially weighted moving average; The scene information weight is: ; in, To detect the proportion of urban scenes collected by drones, To detect the proportion of forest scenes collected by drones, To detect the percentage of grassland shrub scenes collected by drones, To detect the proportion of other scenes collected by the drone, For iterative The percentage of scene information collected by the UAV in the t-1th generation.

[0021] Specifically, the mesh surface fitting technology uses a combination of self-organizing neural network intelligent algorithms to significantly improve the efficiency, accuracy and intelligence level of mesh generation. The mesh surface fitting technology can be used for automatic mesh generation. This method uses the self-learning ability of the neural network to automatically generate meshes that adapt to complex shapes based on the input geometry and topology information. For example, the Let-It-Grow neural network has been used to generate coarse meshes for overlapping unstructured multi-grid algorithms. Through the adjustment and optimization of the self-organizing neural network intelligent algorithm, the mesh can be dynamically adjusted and optimized according to different application scenarios and needs. For example, the mesh density is automatically predicted to generate meshes that adapt to the needs of different regions. This method can automatically generate high-quality meshes based on complex geometry and physical properties, reducing manual intervention. The combination of the self-organizing neural network intelligent algorithm used in the mesh surface fitting technology can not only improve the efficiency and quality of mesh generation, but also make the complex image rendering process more intelligent.

[0022] Furthermore, in the process of calculating the first visible area in the terrain basic model by using a machine vision analysis algorithm, the machine vision analysis algorithm adopts a binocular matching algorithm, and its formula is: ; in, The area map needs to be rendered for the first visible area; The energy function corresponding to the area map to be rendered for the first visible area; is the left eye imaging area, is the imaging area of ​​the right eye; For the first visible area, other areas other than the left eye imaging area in the area map need to be rendered; The time cost and performance cost of rendering the area map for the first visible area, where the time cost is the unit time required to render the image area, and the performance cost is the amount of computing resources required to render the image area; is the disparity value of the left eye imaging area; is the disparity value of the right eye imaging area; is a penalty coefficient, which is applicable to the case where the disparity value of the left eye imaging area differs from the disparity value of the right eye imaging area by 1; It is also a penalty coefficient, which is applicable when the difference between the disparity value of the left eye imaging area and the disparity value of the right eye imaging area is greater than 1; It is an indicator function that returns 1 if the condition is true, otherwise it returns 0; Specifically, visual analysis algorithms are core technologies in the field of computer vision, which aim to extract and understand visual information from images or videos through image processing and pattern recognition technologies. These algorithms are widely used in many fields such as image recognition, target detection, face recognition, and posture estimation. The binocular matching algorithm obtains three-dimensional information of an object through multiple images. Common stereo vision algorithms include methods based on binocular stereo vision and structured light. It draws on the "parallax" principle of human eyes, that is, there is a difference between the left and right eyes in observing an object in the real world. Our brain uses the difference between the left and right eyes to enable us to identify the distance of objects. In computer vision, the depth information of an object is calculated by the parallax between two or more images. Taking the SAD algorithm as an example, its basic process is as follows: 1. Input two images that have been calibrated to achieve row alignment: the left image and the right image.

[0023] 2. Scan the left view, select an anchor point and build a small window.

[0024] 3. Use this small window to cover the left view and the right view, and select all the pixels in the area covered by the small window.

[0025] 4. Calculate the sum of the absolute values ​​of the pixel differences between the left and right view coverage areas.

[0026] 5. Move the small window of the right view and repeat the above operation to find the small window with the smallest SAD value, which is the best matching pixel block.

[0027] Similarly, the binocular matching algorithm renders the terrain information and scene information in the field of view separately, separates the length of time and the amount of calculation required for rendering, prioritizes important rendering objects, and achieves different rendering accuracies for different objects by rendering different rendering objects in segments, thereby achieving initial and rapid rendering of the scene, allowing trainees to obtain better visual information before the start of training and avoid the discomfort caused by low-quality rendering.

[0028] Furthermore, the first learning rendering algorithm includes: a texture rendering algorithm formula and a reflection algorithm formula, and the texture rendering algorithm formula is: ; in, is the color of the final pixel in the pixel area; is the weight of the i-th texture; is the color value of the i-th texture at the coordinate (j, k, l); is the horizontal coordinate in the coordinate (j, k, l), is the ordinate in the coordinates (j, k, l), is the height coordinate in the coordinates (j, k, l); The reflection algorithm formula is: ; in, is the brightness of the reflected light of the final pixel; is the ambient light intensity; is the ambient light reflection coefficient; is the main light intensity; is the diffuse reflectance; is the specular reflection coefficient; is the light source direction vector; is the surface normal vector; is the reflected light direction vector; is the viewing direction vector; is the specular highlight exponent.

[0029] Furthermore, the second learning rendering algorithm includes: a first rendering algorithm weight formula and a second rendering algorithm weight formula; the first rendering algorithm weight formula is used to calculate the rendering priority of the current terrain information, and the first rendering algorithm weight formula is: ; in, is the rendering weight of the corresponding terrain; is the complexity coefficient of the terrain; is the performance demand coefficient of the terrain; It is the rendering priority coefficient of the terrain; is the corresponding terrain, when =1 for mountainous areas, =2 for hills, =3 is flat ground, =4 is water area, =5 is a valley; The second rendering algorithm weight formula is used to calculate the rendering priority of the current scene information, and the second rendering algorithm weight formula is: ; in, is the rendering weight of the corresponding scene; is the complexity coefficient of the scene; is the performance requirement coefficient of the scenario; is the rendering priority coefficient of the scene; For the corresponding scene, when =1 for cities, =2 means forest, =3 is grassland shrubs, =4 means other types; Among them, when ≥ When , the details in the terrain information are rendered first. < When rendering the scene information, details in the scene information are given priority.

[0030] Furthermore, the complexity coefficient of the terrain and the complexity coefficient of the scene are trained and optimized using a computer image training algorithm; the computer image training algorithm is trained using a computer image training algorithm formula, and the computer image training algorithm formula is: ; in, To optimize the weight ratio; is a first-order matrix; is a second-order matrix; is a constant, and 1≥ >0.

[0031] Specifically, multiple detection drones are dispatched to collect information about the simulated training area. The detection drones fly according to the preset routes, and use the position accuracy sensors they carry to collect terrain height data to form terrain information; and use the image recognition sensors they carry to collect scene environment data to form scene information. These data are transmitted to the information processing terminal for subsequent processing and analysis. After receiving the data transmitted by the drone, the information processing terminal begins to generate the terrain data and scene configuration files required for the training task. Through information comparison, the collected terrain information includes the proportion of each terrain, mountain terrain accounts for 40%; hilly terrain accounts for 10%; flat terrain accounts for 35%; water terrain accounts for 10%; valley terrain accounts for 5%. The collected scene information includes the proportion of each scene, urban area accounts for 15%; forest area accounts for 30%; grassland shrub area accounts for 35%; other areas account for 20%. Terrain data: contains grid point coordinates, terrain type (mountain, flat, etc.) and height information. Scene configuration file: contains the location, type and size information of scene elements (such as buildings, trees, etc.). The mesh surface fitting technology is combined with the self-organizing neural network intelligent algorithm for optimization. The collected terrain and scene data are meshed through the mesh surface fitting technology to generate a terrain mesh model and scene layout configuration. For example, for mountainous terrain, grid points with different altitudes are generated according to the height data to form a three-dimensional grid model of the mountain. At the same time, combined with the self-organizing neural network optimization, the mesh model is optimized through the neural network algorithm to adjust the density and distribution of the grid to better adapt to the complexity of the terrain and scene. For example, for complex urban areas, the neural network will increase the mesh density to more finely represent the outline of the building and the street layout. Calculate the first visible area and render the terrain base model based on the generated terrain data and scene configuration file. Use the machine vision analysis algorithm to calculate the first visible area under the initial default perspective of the trainer, and render the first visible area. The machine vision analysis algorithm uses the binocular matching algorithm to calculate the first visible area. Assuming that the trainer's initial perspective is straight ahead, the algorithm determines the range of the first visible area visible from this perspective by analyzing the grid points and scene elements in the terrain base model. For example, if the trainee is on flat ground with mountains and part of the city ahead, the algorithm will calculate the mountain slopes and part of the city buildings that can be seen from the current position. Render the first visible area. For example, render the green vegetation-covered area of ​​the mountain, the gray building outlines of the city, and the blue background of the sky. Generate the texture and layout of the scene elements, using the first learning rendering algorithm to generate the texture and layout for the scene elements in the first visible area. Use the texture rendering algorithm formula to generate the corresponding texture according to the type and position of the scene element. For example, for trees in the forest area, generate green leaf textures and brown trunk textures according to the type and distribution density of the trees.For buildings in urban areas, a gray wall texture and a blue glass window texture are generated. The reflection algorithm formula is used to calculate the reflection effect of scene elements. For example, for the glass windows of buildings, the brightness and color of the glass window reflection are calculated based on the ambient light intensity and the direction of the light source. Assuming that the ambient light intensity is 80 (full value 100), the main light intensity is 100, the diffuse reflection coefficient is 0.5, and the specular reflection coefficient is 0.3, the reflection brightness of the glass window is calculated to be 40 (ambient light reflection) + 30 (specular reflection) = 70. Read the viewpoint position of the training user and use the second learning rendering algorithm to dynamically adjust the scene detail level. According to the complexity coefficient, performance requirement coefficient, and rendering priority coefficient of the terrain and scene, the rendering priority of the current terrain and scene is calculated. For example, for mountainous terrain, the complexity coefficient is 0.8, the performance requirement coefficient is 0.7, and the rendering priority coefficient is 0.9. The rendering weight of the mountain is calculated to be 0.8×0.7×0.9=0.504. For urban areas, the complexity coefficient is 0.9, the performance requirement coefficient is 0.6, and the rendering priority coefficient is 0.8. The calculated rendering weight of the city is 0.9×0.6×0.8=0.432. The scene detail level is dynamically adjusted according to the user's viewpoint position and movement direction. For example, when the trainer is close to the urban area in the field of vision, the details of the urban buildings are rendered first; when far away from the urban area, the detail level of the urban area is reduced to improve the rendering efficiency of other areas. The rendered scene image is output to the display device, and the rendering content is updated according to the user's interactive operation. The terrain and scene elements of the image result are rendered according to the calculated texture, layout and detail level to generate a complete scene image. The rendered image is transmitted to the VR visual device, and the user observes and interacts immersively through the VR device. When the user moves in the scene or changes the perspective, the rendering content is updated in real time. For example, when the user turns from the urban area to the forest area, the rendering system dynamically adjusts the detail level of the forest area and re-renders the scene to ensure that the user always sees a more realistic scene.

[0032] Based on the same inventive concept, Figure 2As shown, this embodiment also provides a simulation simulator real-time rendering scene generation system, including: a detection drone, an information processing terminal, an image simulation simulator, and a VR visual device; the detection drone transmits information to the information processing terminal, the information processing terminal transmits data and configuration files to the image simulation simulator, and the image simulation simulator transmits the generated image results to the VR visual device; the detection drone is used to obtain surrounding terrain information and scene information, the information processing terminal is used to process the terrain information and scene information obtained by the detection drone, and generate terrain data and scene configuration files required for training tasks through the terrain information and the scene information, the image simulation simulator is used to render the terrain data and the scene configuration files to generate the image results, and the VR visual device is used to present the rendered image results.

[0033] Furthermore, the image simulation simulator also includes: a first image information processing module and a second image information processing module, the terrain data and the scene configuration file obtain the first visible area through the first image information processing module, and the first visible area obtains the image result through the second image information processing module.

[0034] Furthermore, In addition, detection drones have important applications in many fields, including but not limited to: Military field: improve battlefield situation awareness capabilities and provide real-time and accurate drone intelligence for command departments. Civilian field: applied to aviation, meteorology, environmental protection and other fields to provide support for economic and social development. Airport clear area: detect and track drones that invade airport clear areas to ensure flight safety. Important infrastructure protection: provide outer protection for bases and important infrastructure. The present invention uses detection drones mainly for scanning terrain data information and collecting environmental data information. Photoelectric detection technology uses optical and electronic principles to capture and analyze the light signals reflected or emitted by drones to achieve long-distance, non-contact detection of targets. The detection drone can collect information using a variety of methods including visible light imaging, infrared imaging and laser imaging. Photoelectric detection technology can provide intuitive image information, suitable for daytime and nighttime detection, especially in complex environments with high detection accuracy. Information processing terminal refers to a device that directly interacts with users or other devices in a computer network. These devices are interfaces for information processing and communication, and are usually used to input programs and data into computers, or receive processing results output by computers. For example, common information processing terminals include desktops, laptops, smart phones, tablet computers, etc., which interact with users through various input and output devices (such as keyboards, mice, displays, cameras, etc.). In the present invention, the information processing terminal is mainly used for information processing. Through high integration, a highly integrated SoC (System On Chip) design is adopted, which integrates multiple functions such as a central processing unit, a memory, and a communication module, thereby improving the performance and reliability of the device. An image simulation simulator is a device or software system for generating and processing image data, which can simulate real-world image scenes in a virtual environment. It generates realistic image data streams through computer graphics and image processing technology, which are used to test and verify image processing algorithms, computer vision systems, and machine learning models. For example, the CamSim™ camera simulator can generate high-quality image data streams and supports multiple interface standards. In the present invention, high-quality and high-resolution image data is generated through an image simulation simulator, which supports multiple image formats and sensor types, and supports real-time image generation and transmission, ensuring the synchronization and real-time nature of image data, ensuring the improvement of image transmission efficiency, reducing time costs, and ensuring the reliability and accuracy of the picture. VR visual devices, namely virtual reality (VR) devices, refer to devices that use computer technology to generate a three-dimensional virtual environment, allowing users to immerse themselves in and interact with it. These devices usually include head-mounted display devices (such as VR headsets), handles, sensors, etc., which can provide multi-sensory experiences of vision, hearing and even touch. In the present invention, VR visual devices are mainly used to provide training images so that trainers can train better.

[0035] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for generating a real-time rendering scene of a simulation simulator, characterized in that: The method comprises the following steps: Collecting surrounding terrain information and scene information through a detection drone, and transmitting the terrain information and scene information to an information processing terminal; Generate terrain data and scene configuration files required for the training task in the information processing terminal through the terrain information and the scene information, and initialize a rendering engine; A terrain basic model is constructed according to the terrain data and the scene configuration file, a first visible area in the terrain basic model is calculated by a machine vision analysis algorithm, and the first visible area is rendered; the first visible area is a visible area at an initial default viewing angle in a scene loaded by a trainee; Using a first learning rendering algorithm to generate texture and layout of scene elements in the first visible area; Reading the viewpoint position of the training user, rendering the first visible area using a second learning rendering algorithm to obtain an image result, and dynamically adjusting the scene detail level; The image result is output to a VR visual device, and the rendering content is updated in real time according to the user's interactive operation.

2. A method for generating a real-time rendering scene of a simulation simulator according to claim 1, characterized in that: Generating terrain data and scene configuration files required for the training task in the information processing terminal includes: In the information processing terminal, the terrain data and scene configuration files required for the training task are generated by the Mesh surface fitting technology. The Mesh surface fitting technology is optimized by a self-organizing neural network intelligent algorithm. The mathematical expression of the self-organizing neural network intelligent algorithm is: ； in, is the weight change of the xth neuron at the tth iteration; is the neural network learning rate; is the neighborhood function, which represents the neighborhood relationship between the y-th winning neuron and the x-th neuron; is the current input sample; is the weight vector of the xth neuron at iteration number t; is the current iteration number; The neural network learning rate is: ； in, is the gradient information of the current parameter; is a constant to ensure numerical stability; is the terrain information weight, is the scene information weight; is an extraordinary parameter used to control the step size of parameter update; The terrain information weight is: ； in, is an exponential decay coefficient, and 0.45> >0.9; To detect the percentage of mountain terrain collected by drones, To detect the percentage of hilly terrain collected by drones, To detect the proportion of flat terrain collected by the drone, To detect the water terrain ratio collected by the drone, To detect the valley terrain percentage value collected by the drone, is the proportion of terrain information collected by the detection drone at the t-1th iteration; is an exponentially weighted moving average; The scene information weight is: ； in, To detect the proportion of urban scenes collected by drones, To detect the proportion of forest scenes collected by drones, To detect the percentage of grassland shrub scenes collected by drones, To detect the proportion of other scenes collected by the drone, It is the percentage of scene information collected by the detection drone at the t-1th iteration.

3. A method for generating a real-time rendering scene of a simulation simulator according to claim 1, characterized in that: The calculating the first visible area in the terrain basic model by using a machine vision analysis algorithm includes: The machine vision analysis algorithm adopts binocular matching algorithm, and its formula is: ； in, The area map needs to be rendered for the first visible area; The energy function corresponding to the area map to be rendered for the first visible area; is the left eye imaging area, is the imaging area of ​​the right eye; For the first visible area, other areas other than the left eye imaging area in the area map need to be rendered; The time cost and performance cost of rendering the area map for the first visible area, where the time cost is the unit time required to render the image area, and the performance cost is the amount of computing resources required to render the image area; is the disparity value of the left eye imaging area; is the disparity value of the right eye imaging area; is a penalty coefficient, which is applicable to the case where the disparity value of the left eye imaging area differs from the disparity value of the right eye imaging area by 1; It is also a penalty coefficient, which is applicable when the difference between the disparity value of the left eye imaging area and the disparity value of the right eye imaging area is greater than 1; It is an indicator function that returns 1 if the condition is true, otherwise it returns 0.

4. A method for generating a real-time rendering scene of a simulation simulator according to claim 1, characterized in that: The first learning rendering algorithm includes: a texture rendering algorithm formula and a reflection algorithm formula, and the texture rendering algorithm formula is: ； in, is the color of the final pixel in the pixel area; is the weight of the i-th texture; is the color value of the i-th texture at coordinate (j, k, l); is the horizontal coordinate in the coordinate (j, k, l), is the ordinate in the coordinates (j, k, l), is the height coordinate in the coordinates (j, k, l); is the total number of textures; The reflection algorithm formula is: ； in, is the brightness of the reflected light of the final pixel; is the ambient light intensity; is the ambient light reflection coefficient; is the main light intensity; is the diffuse reflectance; is the specular reflection coefficient; is the light source direction vector; is the surface normal vector; is the reflected light direction vector; is the viewing direction vector; is the specular highlight exponent.

5. The method for generating a real-time rendering scene of a simulation simulator according to claim 1, characterized in that: The second learning rendering algorithm includes: a first rendering algorithm weight formula and a second rendering algorithm weight formula; the first rendering algorithm weight formula is used to calculate the rendering priority of the current terrain information, and the first rendering algorithm weight formula is: ； in, is the rendering weight of the corresponding terrain; is the complexity coefficient of the terrain; is the performance demand coefficient of the terrain; The rendering priority coefficient of the terrain; is the corresponding terrain, when =1 for mountainous areas, =2 for hills, =3 is flat ground, =4 is water area, =5 is a valley; The second rendering algorithm weight formula is used to calculate the rendering priority of the current scene information, and the second rendering algorithm weight formula is: ； in, is the rendering weight of the corresponding scene; is the complexity coefficient of the scene; is the performance requirement coefficient of the scenario; is the rendering priority coefficient of the scene; For the corresponding scene, when =1 for cities, =2 means forest, =3 is grassland shrubs, =4 means other types; Among them, when ≥ When , the details in the terrain information are rendered first. < When rendering the scene information, details in the scene information are given priority.

6. A method for generating a real-time rendering scene of a simulation simulator according to claim 5, characterized in that: The complexity coefficient of the terrain and the complexity coefficient of the scene are trained and optimized using a computer image training algorithm; the computer image training algorithm is trained using a computer image training algorithm formula, and the computer image training algorithm formula is: ； in, To optimize the weight ratio; is a first-order matrix; is a second-order matrix; is a constant, and 1≥ >

0.

7. A real-time rendering scene generation system for a simulation simulator, based on the method according to any one of claims 1 to 6, characterized in that: The system includes: a detection drone, an information processing terminal, an image simulation simulator, and a VR visual device; the detection drone transmits information to the information processing terminal, the information processing terminal transmits data and configuration files to the image simulation simulator, and the image simulation simulator transmits the generated image results to the VR visual device; the detection drone is used to obtain surrounding terrain information and scene information, the information processing terminal is used to process the terrain information and scene information obtained by the detection drone, and generate terrain data and scene configuration files required for training tasks through the terrain information and the scene information, the image simulation simulator is used to render the terrain data and the scene configuration files to generate the image results, and the VR visual device is used to present the rendered image results.

8. A simulation simulator real-time rendering scene generation system according to claim 7, characterized in that: The image simulation simulator also includes: a first image information processing module and a second image information processing module. The terrain data and the scene configuration file obtain the first visible area through the first image information processing module, and the first visible area obtains the image result through the second image information processing module.

Images