An ultra-high voltage station rendering method based on real-time global light and rasterization mixed acceleration

Abstract

The application discloses a kind of based on real-time global illumination and grating hybrid acceleration ultra-high voltage station rendering method, it is related to ultra-high voltage station inspection simulation technical field, including the following steps: in strong electromagnetic interference environment, by adaptive electromagnetic interference filtering algorithm is combined with global illumination probe dynamic interpolation compensation algorithm, dynamic adjustment illumination sampling frequency and insert pre-computed global illumination data.The application is by adaptive electromagnetic interference filtering and illumination sampling dynamic compensation mechanism, effectively suppresses the high-frequency jitter of illumination data caused by strong electromagnetic noise, can according to the noise spectrum characteristics of real-time acquisition intelligent adjustment global illumination probe sampling frequency, and is combined with material perception illumination interpolation algorithm, ensure that metal and insulator surface under different interference intensity all present continuous, smooth illumination performance, so that ultra-high voltage station still can maintain stable visual output under strong electromagnetic interference environment, guarantee the reliability and authenticity of inspection simulation process.

Description

Technical Field

[0001] This invention relates to the field of ultra-high voltage power station inspection simulation technology, specifically to an ultra-high voltage power station rendering method based on real-time global illumination and rasterization hybrid acceleration. Background Technology

[0002] With the continuous growth of electricity demand, the power industry has increased its research and development and application of ultra-high voltage (UHV) transmission technology to improve power transmission efficiency and stability. In this process, the value of site rendering technology has become increasingly prominent. In UHV projects, site rendering can simulate equipment layout, spatial configuration and environmental impact in the preliminary design stage, helping engineers and designers to identify potential problems in a timely manner. In addition, rendered images can effectively assist in communication with various departments, the public and stakeholders, provide intuitive information, and help obtain project approval and public understanding.

[0003] In existing technologies, in UHV power station inspection simulation scenarios, strong electromagnetic environments can easily cause frequent jitter and distortion of real-time global illumination data on the surface of power equipment, forming flickering light spots and affecting the realism of visual interpretation. Furthermore, inconsistencies in the switching of illumination models between near and far equipment in the scenario cause visual tearing and depth perception confusion, affecting the accuracy of spatial positioning and path planning. Moreover, under different weather conditions, global illumination and rasterization rendering cause conflicts in refraction and scattering calculations on the surfaces of transparent and semi-transparent objects, resulting in abnormal light transmission of materials and weakening the credibility of simulations under complex weather conditions. To address these issues, a UHV power station rendering method based on real-time global illumination and rasterization hybrid acceleration is proposed. Summary of the Invention

[0004] To solve the above-mentioned technical problems, the present invention is implemented through the following technical solution: a rendering method for ultra-high voltage power stations based on real-time global illumination and rasterization hybrid acceleration, comprising the following steps: Step 1: In a strong electromagnetic interference environment, the adaptive electromagnetic interference filtering algorithm is combined with the global illumination (GI) probe dynamic interpolation compensation algorithm to dynamically adjust the illumination sampling frequency and insert pre-calculated global illumination data to ensure that the illumination on the surface of metal and insulating equipment is stable and flicker-free, effectively suppressing the illumination flicker and noise on the surface of the equipment caused by electromagnetic interference, and improving visual stability and interpretation accuracy. Step 2: Divide the view frustum region according to the distance between the power equipment and the viewpoint. Dynamically allocate global illumination and rasterization rendering weights to different regions through the view frustum layered lighting fusion algorithm. This achieves a balance between high-fidelity lighting in the core rendering area and resource usage in the distant view. While ensuring high fidelity in key areas, it significantly reduces the overall rendering load and optimizes resource utilization efficiency. Step 3: Use depth map comparison technology to detect the consistency of light and shadow transition between near and far. Use rasterization depth calibration algorithm to dynamically adjust the edges of shadows and lighting, eliminate visual tearing and depth perception confusion caused by rendering model switching, eliminate visual tearing and depth confusion caused by layered rendering in the scene, and achieve a smooth and natural transition between near and far scenes. Step 4: For complex weather conditions, a weather physics refraction simulation algorithm is introduced to dynamically correct the propagation path and intensity of global light on transparent and semi-transparent surfaces, ensuring the realism of light propagation under complex weather conditions, making the optical performance of transparent objects such as refraction and reflection under rain and fog weather more in line with physical laws, and enhancing the realism of the environment. Step 5: In the rasterization rendering stage, a real-time scattering medium rasterization compensation algorithm is introduced to simulate the scattering effect of light in rain and fog under complex weather conditions, and to dynamically adjust the material light response of transparent and semi-transparent objects to improve visual consistency and physical credibility, accurately simulate atmospheric scattering effects, and achieve physically correct fog effects, volumetric light and other weather atmosphere representations. Step 6: Integrate multi-dimensional parameters of electromagnetic interference, distance layering, and weather factors to construct a dynamic rendering parameter optimization model. Through the pixel difference feedback between real-time rendering results and the benchmark template, the model achieves adaptive fine-tuning of parameters and closed-loop verification of system stability, which can maintain the optimal balance between visual quality and performance in a long-term, variable environment.

[0005] Preferably, step 1 specifically includes: The noise spectrum characteristics under strong electromagnetic environment are collected in real time by electromagnetic sensors, and the sampling time interval of the global illumination probe is dynamically adjusted according to the peak value of the spectrum to suppress high-frequency illumination jitter, effectively suppress high-frequency illumination flicker caused by strong electromagnetic interference, and ensure visual stability. Between two adjacent lighting samples, linear interpolation is performed based on historical sampling data and pre-calculated static global illumination map to generate a continuous and smooth lighting data stream, compensating for sampling loss caused by electromagnetic interference, ensuring smooth transition of lighting intensity and color, and avoiding abrupt changes or breaks in the image. Based on the material type of the equipment surface, different interpolation weights and filtering intensity thresholds are set for metal and insulating surfaces respectively, so as to realize adaptive lighting stability of material perception, achieve precise lighting stability processing for different materials, and improve the visual realism and anti-interference ability of various equipment surfaces.

[0006] Preferably, step 1 further includes: Based on the physical material properties of power equipment, we established illumination response models for high reflectivity of metal surfaces and diffuse reflectivity of insulator surfaces, and calculated the sensitivity coefficients to electromagnetic noise respectively, realizing the physical correlation model between materials and electromagnetic interference, and providing a scientific basis for subsequent differentiated processing. Based on the aforementioned sensitivity coefficient, a high interpolation weight and narrow filtering bandwidth are configured for the metal surface to preserve high-frequency illumination details, and a smooth interpolation curve is configured for the insulator surface to effectively preserve the mirror gloss of the metal surface and the natural softness of the insulator surface while suppressing interference. In the rendering pipeline, the corresponding lighting stabilization algorithm branch is dynamically selected through shader variants to perform differentiated anti-interference processing on the surfaces of power equipment made of different materials, ensuring that equipment made of different materials can present stable, realistic and physically consistent visual effects in strong electromagnetic environments.

[0007] Preferably, step 2 specifically includes: Centered on the observer's viewpoint, the three-dimensional view frustum is divided into a core rendering area, a transition rendering area, and a distant rendering area based on a preset distance threshold, thereby realizing intelligent layering of the scene's visual focus and resource allocation by area; Allocate high-priority real-time global illumination computing resources to devices in the core rendering area and use full-resolution rasterization rendering to maintain high fidelity of lighting and shadows in this area, ensuring that the rendering details and realism in the center of the operator's field of vision are optimal. The system dynamically adjusts the blending ratio of global illumination and rasterization rendering between the transition rendering area and the distant rendering area, and gradually reduces the frequency of lighting updates and the resolution of shadow maps to balance rendering quality and system performance in the visual edge area and avoid resource waste.

[0008] Preferably, step 2 further includes: A hybrid weighting function based on device distance and screen space occupancy is established to calculate the global illumination contribution required for each rendered object in the transition rendering zone and the distant rendering zone in real time. This can accurately quantify the visual importance of objects, ensure that rendering resources are allocated on demand, and avoid waste. In the transition rendering area, the global illumination contribution is weighted and mixed with the pre-calculated illumination probe data, and the real-time global illumination data is updated once every fixed number of frames. This balances visual realism and performance overhead, achieving smooth illumination transitions without obvious inter-frame flicker. In the distant rendering area, pre-computed global illumination maps are used entirely, and real-time rasterization shadow calculation is turned off. Rendering resources are concentrated on the core rendering area, which greatly reduces the rendering load of the distant area. The freed-up GPU resources effectively improve the rendering quality and frame rate stability of the core rendering area.

[0009] Preferably, step 3 specifically includes: In each rendering frame, the first depth map generated by global illumination and the second depth map generated by rasterization rendering are acquired synchronously. The geometric information of the two rendering paths is compared in real time to provide a data basis for accurately detecting visually inconsistent areas. By comparing two depth maps pixel by pixel, regions with depth value differences exceeding a preset threshold are identified and marked as potential visual tearing areas. The system automatically locates pixels with depth jumps caused by rendering model switching, thus achieving preliminary screening of problem areas. For pixels within the potential visual tearing zone, a depth edge smoothing algorithm based on bilateral filtering is used to perform transition calibration on the depth value and adjacent illumination intensity, eliminating discontinuous visual jumps, smoothing geometric and illumination transitions, significantly eliminating the sense of visual tearing, and improving the visual coherence of the scene.

[0010] Preferably, step 4 specifically includes: Based on real-time input meteorological parameters, including rain intensity, fog density, and air refractive index, the theoretical refraction offset of light on transparent and semi-transparent media surfaces under the current environment is calculated, effectively simulating the refraction of light on glass and insulator surfaces in rainy weather and improving visual realism. In the calculation of the propagation path of global illumination, the coordinates of the intersection point of the light and the object surface and the normal direction are dynamically corrected according to the refraction offset to eliminate the illumination misalignment caused by weather conditions and ensure that the light and shadow are accurately aligned with the object outline. Based on the corrected lighting path, the intensity attenuation of transmitted and reflected light is recalculated, and the shader lighting parameters of the affected area are updated to achieve a natural transition between the brightness and highlights of the translucent material under complex weather conditions, thereby enhancing physical reliability.

[0011] Preferably, step 5 specifically includes: In the rasterization rendering pipeline, a physically based scattering medium simulation layer is added. This layer receives the meteorological density and lighting direction of the current scene as input, which significantly enhances the atmosphere and spatial volume of rain and fog scenes and improves the integration of the environment. The scattering medium simulation layer simulates the multiple scattering and absorption processes of light in media such as rain and fog based on the light stepping algorithm, calculates the contribution value to the final pixel color, so that the light is naturally diffused in the medium, producing a soft scattering halo and a more realistic visual effect. The calculated scattering contribution value is superimposed and fused with the conventional rasterization coloring result of the object surface to achieve physical superposition of the medium's scattered light and the surface material, thereby improving the visual consistency and naturalness of the overall image.

[0012] Preferably, step 5 further includes: Based on the material properties of transparent and translucent objects, the light transmittance and scattering coefficient in the scattering medium are dynamically adjusted to ensure that the light transmission and scattering response of different materials conforms to the real physical behavior and improves the visual distinction of materials in the medium. For power equipment structures including insulators and glass, a surface dirt or water accumulation model is established to further correct the mixing ratio of light refraction, reflection and internal scattering on the surface, simulate the complex influence of surface attachments on light, and enhance the realism of the equipment appearance under wet or polluted conditions. The corrected material light response parameters are fed back to the scattering medium simulation layer and the surface shader to realize the coupled calculation of medium scattering and object material response, realize bidirectional light interaction between material and medium, and improve the lighting consistency and physical credibility of the overall scene.

[0013] Preferably, step 6 specifically includes: A multidimensional dynamic parameter vector is constructed with electromagnetic interference intensity coefficient, apparent cone region weight and weather influencing factors as core inputs to achieve full-dimensional quantitative perception of environmental status and provide accurate and unified input basis for intelligent decision-making. By inputting a multidimensional dynamic parameter vector into a pre-trained linear regression optimization model, the model outputs a set of real-time rendering parameters to control lighting, shadows, refraction, and scattering, achieving millisecond-level intelligent prediction of optimal rendering parameters from complex environments, replacing tedious manual parameter tuning. After each rendering frame, the rendered output image is compared with a baseline template image generated under ideal conditions at the pixel level. If the difference exceeds the tolerance, the difference data is fed back to the linear regression optimization model to fine-tune the model weights, forming a closed-loop adaptive system. This gives the system the ability to learn and continuously optimize itself, ensuring the stability and improvement of visual quality under long-term operation.

[0014] This invention provides a rendering method for ultra-high voltage power stations based on a hybrid acceleration method of real-time global illumination and rasterization. It has the following beneficial effects: (I) This ultra-high voltage power station rendering method based on real-time global illumination and rasterization hybrid acceleration effectively suppresses high-frequency jitter of illumination data caused by strong electromagnetic noise through adaptive electromagnetic interference filtering and dynamic compensation mechanism for illumination sampling. It can intelligently adjust the sampling frequency of global illumination probe according to the noise spectrum characteristics collected in real time, and combined with material-aware illumination interpolation algorithm, ensure that the surfaces of metals and insulators present continuous and smooth illumination under different interference intensities. This enables the ultra-high voltage power station to maintain stable visual output under strong electromagnetic interference environment, eliminates the interference of flickering light spots on equipment status interpretation, and ensures the reliability and authenticity of the inspection simulation process.

[0015] (II) This ultra-high voltage power station rendering method based on real-time global illumination and rasterization hybrid acceleration effectively solves the problems of visual tearing and depth perception confusion caused by using different rendering models for equipment at different distances through view frustum layered rendering and depth map comparison and calibration technology. It dynamically allocates global illumination and rasterization rendering weights according to the distance of the equipment, and achieves natural transition between different rendering areas through depth difference detection and edge smoothing algorithms in post-processing. This makes the scene have consistent geometric expression and light and shadow connection between far, middle and near scenes, improving the operator's spatial positioning accuracy and path planning reliability.

[0016] (III) This ultra-high voltage power station rendering method based on real-time global illumination and rasterization hybrid acceleration introduces physical weather optical simulation, including refraction path correction and real-time calculation of scattering medium, so that the visual performance of transparent and semi-transparent objects under different meteorological conditions is more in line with natural laws. By dynamically correcting the propagation path and intensity attenuation of light in rain and fog media, and simulating the influence of medium scattering on scene color, it can realistically reproduce the light transmission, reflection and scattering effects of equipment such as insulators and glass covers in humid and foggy environments, significantly improving the visual credibility and immersion of simulation under complex meteorological environments. Attached Figure Description

[0017] Figure 1 This is a schematic diagram illustrating the workflow of an ultra-high voltage power station rendering method based on real-time global illumination and rasterization hybrid acceleration according to the present invention. Figure 2 This is a schematic diagram of the process flow of an ultra-high voltage power station rendering method based on real-time global illumination and rasterization hybrid acceleration according to the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Example 1, please refer to Figure 1 , Figure 2 This invention provides a technical solution: a rendering method for ultra-high voltage power stations based on real-time global illumination and rasterization hybrid acceleration, comprising the following steps: Step 1: In a strong electromagnetic interference environment, an adaptive electromagnetic interference filtering algorithm combined with a global illumination (GI) probe dynamic interpolation compensation algorithm is used to dynamically adjust the illumination sampling frequency and insert pre-calculated global illumination data to ensure stable and flicker-free illumination on the surfaces of metal and insulating equipment. This effectively suppresses illumination flicker and noise on the equipment surface caused by electromagnetic interference, improving visual stability and interpretation accuracy. The electromagnetic sensor collects the noise spectrum characteristics in a strong electromagnetic environment in real time, and dynamically adjusts the sampling time interval of the global illumination probe according to the spectrum peak to suppress high-frequency illumination jitter, effectively suppressing high-frequency illumination flicker caused by strong electromagnetic interference and ensuring visual stability. Between two adjacent illumination samples, linear interpolation is performed based on historical sampling data and pre-calculated static global illumination maps to generate a continuous and smooth illumination data stream, compensating for sampling loss caused by electromagnetic interference, ensuring smooth transition of illumination intensity and color, and avoiding jumps or breaks in the image. Different interpolation weights and filtering intensity thresholds are set for metal and insulating surfaces according to the material type of the equipment surface to achieve material-aware adaptive illumination stabilization. Precise illumination stabilization processing is achieved for different materials, improving the visual realism and anti-interference ability of various equipment surfaces. Furthermore, broadband electromagnetic sensors deployed in the simulation scenario are used to collect real-time noise data from the simulated strong electromagnetic environment, obtaining a noise spectrum ranging from 0.1MHz to 1GHz. Spectrum analysis is performed at a frequency of 10 times per second to identify the main peak frequency and amplitude. Based on a preset spectrum-sampling mapping rule, when the main peak amplitude exceeds a threshold, the sampling interval of the global illumination probe is dynamically extended from once per frame to once every 2 to 3 frames to suppress illumination sampling jitter caused by high-frequency electromagnetic noise coupling. Simultaneously, the stability of the noise spectrum is continuously monitored; if the main peak amplitude is continuous... If all five samples are below the threshold, the sampling frequency is gradually restored to the default frequency to achieve adaptive anti-interference sampling control. Within the adjusted sampling interval, to ensure the continuity of lighting data, a linear interpolation algorithm is used to compensate for missing real-time samples. Specifically, a lighting buffer containing the five most recent valid samples is maintained, and a pre-generated static global illumination map (typically 1024×1024 resolution, HDR format) is loaded. During interpolation, linear weighting is applied between the two most recent valid sampling points based on the current timestamp, and the weighting coefficients are dynamically calculated based on the time distance. The baseline lighting values ​​corresponding to the scene coordinates are extracted from the pre-calculated lightmap and mixed into the interpolation result with a weight of 20%. This smooths out sampling gaps or abnormal jumps caused by electromagnetic interference, generating a continuous and physically reasonable lighting data stream to ensure no abrupt changes in lighting intensity and color during rendering. A parameter set is pre-configured based on the physical material properties of the power equipment to achieve material-adaptive lighting stability. For metal surfaces (reflectivity ≥ 0.6), a higher interpolation weight (0.7) and a narrower filtering bandwidth (cutoff frequency of 50% of normal value) are set to preserve high-frequency specular reflection details. For insulator surfaces… For surfaces (primarily diffuse reflection), a lower interpolation weight (0.3) and a wider filtering bandwidth (cutoff frequency of 150% of normal value) are used to enhance lighting smoothness. In the shader stage of the rendering pipeline, the corresponding lighting stabilization algorithm variant is dynamically selected according to the object material ID. Through the interpolation weight and filtering parameters passed in in real time, the metal and insulator surfaces are processed differently. At the same time, a filtering intensity threshold is set. When the calculated lighting change rate exceeds the threshold (5% per frame), additional smoothing filtering is activated to ensure that all types of materials can present a stable and natural visual performance in strong electromagnetic environments. In addition, step 1 also includes: based on the physical material properties of power equipment, establishing a lighting response model for high reflectivity of metal surfaces and diffuse reflectivity of insulator surfaces, calculating the sensitivity coefficients to electromagnetic noise respectively, realizing the physical correlation model between materials and electromagnetic interference, providing a scientific basis for subsequent differentiated processing, configuring high interpolation weights and narrow filtering bandwidths for metal surfaces based on the sensitivity coefficients to retain high-frequency lighting details, and configuring smooth interpolation curves for insulator surfaces to effectively retain the specular gloss of metal surfaces and the natural softness of insulator surfaces while suppressing interference, and dynamically selecting the corresponding lighting stabilization algorithm branch through shader variants in the rendering pipeline to perform differentiated anti-interference processing on the surfaces of power equipment of different materials, ensuring that equipment of different materials can present stable, realistic and physically consistent visual effects in strong electromagnetic environments; Furthermore, based on the electromagnetic optical properties of the physical materials of power equipment, differentiated surface illumination response models were established. For metallic surfaces, a micro-surface reflection model was used to calculate the sensitivity of specular reflection components to electromagnetic noise, based on high reflectivity (reflectivity ≥ 0.6). For insulator surfaces, a Lambertian reflection model was used to analyze the response of diffuse reflection components to electromagnetic noise, based on the dominant diffuse reflection characteristics. Through laboratory calibration and field data acquisition, the noise coupling coefficients of the two materials in specific electromagnetic frequency bands were obtained: the noise sensitivity coefficient for metallic surfaces was set at 0.7 ± 0.0. 5. The surface noise sensitivity coefficient of the insulator is set to 0.3±0.05. This coefficient serves as the benchmark parameter for subsequent anti-interference processing, ensuring a quantitative correlation between the material's optical properties and electromagnetic interference response. Based on the material noise sensitivity coefficient, a differentiated anti-interference processing parameter configuration system is established. For metal surfaces, a high interpolation weight of 0.7±0.02 and a narrow filter bandwidth (cutoff frequency of 50%±5% of the benchmark value) are configured to suppress low-frequency noise while preserving the high-frequency detail components of specular reflection. For insulator surfaces, a low interpolation weight of 0.3±0.02 and a wide filter bandwidth (cutoff frequency of 50%±5% of the benchmark value) are configured. The system achieves smooth transition and noise suppression of diffuse components by stopping the frequency at 150%±5% of the baseline value. It automatically matches processing parameters based on the material identifier of the real-time rendering object and ensures, through a parameter verification module, that the interpolation weight is within the effective range of 0.2-0.8 and the filter bandwidth adjustment range is controlled between 30%-180% of the baseline value, preventing rendering anomalies caused by parameters exceeding limits. A dynamic scheduling mechanism for shader variants based on material identifiers is implemented in the Unity rendering pipeline. A shader variant library containing metal and insulator processing branches is pre-compiled. Each variant integrates a corresponding interpolation algorithm, filter kernel function, and noise suppression module. During rendering, the corresponding variant is activated in real-time via the material ID, and the calculated interpolation weight (0.3 / 0.7) and filter bandwidth parameters are dynamically passed through the Shader.PropertyToID interface. During processing, the pixel-level illumination change rate is monitored in real-time. When the change rate exceeds the threshold of 5% / frame, an auxiliary smoothing filter is automatically activated for secondary optimization. All processing is completed in the fragment shader stage, ensuring that the rendering latency per frame increases by no more than 1.5ms, maintaining real-time rendering performance requirements. Step 2: Divide the view frustum region based on the distance between the power equipment and the viewpoint. Dynamically allocate global illumination and rasterization rendering weights to different regions using a view frustum layered lighting fusion algorithm. This achieves a balanced optimization of high-fidelity lighting in the core rendering area and resource usage in the distant view. While ensuring high fidelity in key areas, it significantly reduces the overall rendering load and optimizes resource utilization efficiency. Centered on the observer's viewpoint, the 3D view frustum is divided into a core rendering area, a transition rendering area, and a distant view rendering area based on a preset distance threshold. This achieves intelligent layering of the scene's visual focus and resource allocation by area. High-priority real-time global illumination computing resources are allocated to the equipment in the core rendering area, and full-resolution rasterization rendering is used to maintain high fidelity of lighting and shadows in this area. This ensures that the rendering details and realism in the center of the operator's field of vision are optimal. The mixing ratio of global illumination and rasterization rendering is dynamically adjusted for the transition rendering area and the distant view rendering area. The frequency of lighting updates and the resolution of shadow maps are gradually reduced. This balances rendering quality and system performance in the visual edge area and avoids resource waste. Furthermore, based on the position and orientation of the real-time rendering camera (observer's viewpoint), the 3D view frustum space is divided into three continuous rendering regions according to pre-set distance thresholds. Specific thresholds are set based on typical inspection viewing distances and display resolutions: the core rendering region is 0 to 5 meters in front of the viewpoint; the transition rendering region is 5 to 15 meters; and the distant rendering region is 15 meters to the far plane of the view frustum clipping. This region division is achieved by calculating the Euclidean distance between the rendered object and the viewpoint for each frame. For all power equipment models falling within the core rendering region, the highest priority real-time global illumination calculation resources are allocated to ensure that each frame in this region performs complete lighting probe data updates and lighting propagation calculations. Simultaneously, the core rendering region is forced to use the same full-resolution rasterization rendering pipeline as the display output, with the shadow map resolution set to 2048x2048 and PCSS soft shadow technology enabled. This ensures that the lighting, shadow details, and surface material reflections of the equipment within the operator's key observation area have the highest fidelity, meeting the needs for close-range fine-tuning. For models falling within the core rendering region... For device objects within the transition rendering zone (5-15 meters), a hybrid calculation mode of global illumination and rasterization rendering is adopted. A hybrid weight coefficient α (0≤α≤1) is dynamically calculated for each object. This coefficient is determined by the object distance, screen space occupancy, and motion state. The calculation formula is α=(15-d) / 10*S, where d is the distance (meters) and S is the normalized screen occupancy. During rendering, the final color of the object is linearly mixed from α parts of the real-time global illumination shading result and (1-α) parts of the pre-calculated lightmap shading result. At the same time, the update frequency of global illumination data in this area is reduced to once every 3 frames, and the shadow map resolution is dynamically reduced to 1024x1024. For objects in the distant rendering zone (>15 meters), pre-calculated lightmaps are used for shading, and real-time shadow calculation is completely turned off. Through a scheduler based on frame period and region identifier, GPU computing resources are dynamically allocated to ensure that the rendering load of the transition rendering zone and the distant rendering zone is significantly lower than that of the core rendering zone, thereby achieving an optimized balance of overall performance.A dynamic rendering parameter table is maintained to manage real-time resource allocation in different regions. This table is linked with the view frustum layering module and the performance monitoring module. The performance monitoring module collects GPU frame time and memory usage in real time. When the frame time exceeds the 33ms threshold (corresponding to 30FPS), the resource scheduler automatically fine-tunes the partitioning strategy, temporarily shrinking the core rendering area to 4 meters, or further reducing the lighting update frequency of the transition rendering area to once every 4 frames. At the same time, each rendering object is bound to a priority score calculated based on material complexity, number of geometric faces, and the region it is located in, to guide asynchronous loading and unloading. All adjustment parameters are designed with reasonable boundary values ​​(the lower limit of the core rendering area distance is 1 meter, and the lower limit of the shadow resolution is 512x512). A feedback verification loop with a two-frame delay is used to ensure that the visual quality degradation after adjustment is within the preset tolerance, so as to achieve adaptive stability of rendering performance and effect while ensuring visual consistency. In addition, step 2 also includes: establishing a hybrid weight function based on device distance and screen space occupancy to calculate the global illumination contribution required for each rendered object in the transition rendering zone and the distant rendering zone in real time. This can accurately quantify the visual importance of objects, ensure that rendering resources are allocated on demand, and avoid waste. In the transition rendering zone, the global illumination contribution is weighted and mixed with pre-computed lighting probe data, and the real-time global illumination data is updated every fixed number of frames. This balances visual realism and performance overhead, achieving smooth lighting transitions without obvious inter-frame flicker. In the distant rendering zone, the pre-computed global illumination map is used completely, and real-time rasterization shadow calculation is turned off. The rendering resources are concentrated on the core rendering zone, which greatly reduces the load on distant rendering. The released GPU resources effectively improve the rendering quality and frame rate stability of the core rendering zone. Furthermore, to precisely control rendering resources, based on the Euclidean distance between the device and the observer's viewpoint, and the pixel occupancy of the object in the current frame's screen space, the real-time global illumination contribution distance required for each rendered object in the scene is dynamically calculated. This is determined according to preset partition thresholds (0-5 meters for the core rendering area, 5-15 meters for the transition rendering area, and >15 meters for the distant rendering area). By projecting the object onto the normalized screen coordinate system, the ratio of the pixel area covered by the bounding box to the total screen area is calculated to obtain the screen space occupancy rate. This value is estimated during the geometric processing stage before rasterization in each frame. Finally, the product of the distance factor and the screen occupancy rate factor is defined as a hybrid weighting function, where the distance factor is used in the transition... Within the rendering zone, the weight decreases linearly from 1.0 to 0 with increasing distance, and is forced to 0 in the distant rendering zone. The weight calculation result is limited to the interval [0, 1] and stored in real time in the dynamic parameter buffer associated with each rendering object. For rendering objects falling into the transition rendering zone (5-15 meters), the final lighting result is generated by mixing real-time global illumination and pre-calculated lighting data. During shading, the fragment shader reads the real-time global illumination contribution weight value of the object. This weight value is multiplied by the lighting information (including diffuse color, specular direction and intensity) calculated by the real-time global illumination system in the current frame or the most recent valid frame to obtain the real-time contribution part. At the same time, it is sampled from the pre-baked global illumination map. Pre-computed lighting data corresponding to world coordinates is retrieved, and the final pixel color is linearly blended based on weights. To balance performance and visual continuity, the real-time global illumination data in the transition rendering area adopts a fixed-interval periodic update strategy. The standard update cycle is a complete light probe resampling and propagation calculation every 3 frames. In non-updating frames, the shader uses the most recently updated cached data. For all objects in the distant rendering area (distance greater than 15 meters), the rendering pipeline relies entirely on the pre-computed lighting data. During the shading process, all real-time global illumination calculation processes are skipped, including light probe lookup, real-time shadow map generation, and light propagation solution. The fragment shader directly samples the pre-computed data. Global illumination maps are used to obtain diffuse and ambient occlusion information, and simplified basic shading models such as Blinn-Phong without real-time specular highlights are adopted. At the same time, real-time rasterized shadow calculation is completely turned off. Objects can only obtain static shadow information from pre-baked shadow maps, or have no dynamic shadows. Through this optimization, the GPU computing units and bandwidth resources originally allocated to real-time lighting and shadow calculation in the distant rendering area are completely released. The core rendering resources saved by the scheduler are redirected and prioritized for allocation to the core rendering area to support full-resolution shadow map rendering, high-frequency lighting updates, and more complex lighting model calculations, achieving maximum focus on visual fidelity within the overall performance budget. Step 3: Utilize depth map comparison technology to detect the consistency of light and shadow transitions between near and far views. Dynamically adjust shadow and lighting edges using a rasterization depth calibration algorithm to eliminate visual tearing and depth perception confusion caused by rendering model switching, and eliminate visual tearing and depth distortion caused by layered rendering in the scene. Achieve a smooth and natural transition between near and far views. In each rendering frame, simultaneously acquire the first depth map generated by global illumination and the second depth map generated by rasterization rendering. Compare the geometric information of the two rendering paths in real time to provide a data foundation for accurately detecting visually inconsistent areas. Perform pixel-by-pixel comparison on the two depth maps to identify areas where the depth value difference exceeds a preset threshold and mark them as potential visual tearing areas. Automatically locate pixels with depth jumps caused by rendering model switching to achieve preliminary screening of problem areas. For pixels in potential visual tearing areas, use a depth edge smoothing algorithm based on bilateral filtering to perform transition calibration on depth values ​​and adjacent lighting intensities, eliminate discontinuous visual jumps, smooth geometric and lighting transitions, significantly eliminate visual tearing, and improve scene visual coherence. Furthermore, in the post-processing stage of each rendered frame, after all lighting and shading calculations are completed and before final image compositing, depth information is extracted synchronously from two independent rendering pipelines. The first depth map originates from the rasterization output of the global illumination system, and the depth values ​​record the geometric distribution of the scene under the global illumination calculation view. The second depth map originates from the depth buffer of the rasterization rendering pipeline, representing the geometric information of the final visible surface. Both depth maps need to be unified to the same viewport resolution and the same normalized depth space (linear or non-linear mapping from the camera's near plane to the far plane). Before comparison, the two depth maps undergo necessary coordinate alignment and sampling filtering to ensure accurate pixel-level correspondence. Subsequently, the GPU compute shader performs pixel-by-pixel depth value difference calculations, dynamically setting the difference threshold based on the scene scale and camera parameters. When the depth difference at a certain pixel At that time, the pixel is marked as a potential visual tearing area, and its screen coordinates are stored in a temporary mask texture. This represents the depth value extracted from the rasterized output of the global illumination system. This represents the depth value extracted from the depth buffer of the main rasterization rendering pipeline. After initial difference comparison, morphological optimization is performed on the marked potential tear regions to eliminate misjudgments of isolated points caused by depth map noise or boundary sampling errors. Specifically, a 3×3 convolution kernel is used to dilate the mask texture, merging isolated points into adjacent regions to form connected regions to be processed. Simultaneously, to distinguish different types of discontinuous regions, the angle between the local depth gradient of the difference region and the normal direction of adjacent pixels is further calculated. If the gradient direction is consistent within the region and the normal angle is less than 15 degrees, it is determined to be a geometric edge tear caused by rendering model switching; otherwise, it may be surface details or shadow boundaries, requiring different processing. All optimized region information, including bounding boxes, center positions, and average depth difference values, is encapsulated into a structure array. For the marked and optimized potential tear regions, depth and lighting edge smoothing calibration based on bilateral filtering is implemented. A 5×5 neighborhood window is taken centered on the tear region pixel, and bilateral filtering is applied to the depth values ​​in the first and second depth maps respectively. The spatial weights of the bilateral filtering are... Set to 2.0 pixels, depth difference weight The depth values ​​are dynamically adjusted based on the current depth differences in the region, and then filtered to obtain a smooth transition depth value. Subsequently, based on the smooth transition ratio of the depth values, linear interpolation calibration is performed on the illumination intensity (sampled from the frame buffer or illumination texture) at the same pixel location, using the following formula: ,in , This indicates the final illumination intensity after smoothing calibration. This represents the light intensity obtained from the global illumination path. This represents the light intensity obtained from the main rasterization rendering path. Indicates the mixed weighting coefficient. This represents the depth value after bilateral filtering and smoothing, ensuring a smooth and synchronized transition in lighting intensity within depth transition rendering areas, eliminating brightness jumps caused by depth discontinuities. near hour, Approaching 0, the final illumination is Master, when near hour, Approaching 1, the final illumination is... The main focus is on the post-processing shader, where all calibration calculations are performed in a separate post-processing shader. The total processing time per frame is controlled within 1 millisecond to ensure real-time performance. After processing, the depth and lighting data are rewritten into the corresponding buffers for final image compositing, thereby achieving a visually seamless transition between foreground and background. Step 4: For complex weather conditions, a weather physics refraction simulation algorithm is introduced to dynamically correct the propagation path and intensity of global light on transparent and semi-transparent surfaces, ensuring the realism of light propagation under complex weather conditions, making the optical performance of transparent objects such as refraction and reflection under rain and fog weather more in line with physical laws, and enhancing the realism of the environment. Step 5: In the rasterization rendering stage, a real-time scattering medium rasterization compensation algorithm is introduced to simulate the scattering effect of light in rain and fog under complex weather conditions, and to dynamically adjust the material light response of transparent and semi-transparent objects to improve visual consistency and physical credibility, accurately simulate atmospheric scattering effects, and achieve physically correct fog effects, volumetric light and other weather atmosphere representations. Step 6: Integrate multi-dimensional parameters of electromagnetic interference, distance layering, and weather factors to construct a dynamic rendering parameter optimization model. Through the pixel difference feedback between real-time rendering results and the benchmark template, the model achieves adaptive fine-tuning of parameters and closed-loop verification of system stability, which can maintain the optimal balance between visual quality and performance in a long-term, variable environment.

[0020] Example 2, as Figure 1 , Figure 2 As shown, based on Embodiment 1, the present invention provides a technical solution: Step 4 specifically includes: calculating the theoretical refraction offset of light on the surface of transparent and semi-transparent media under the current environment according to the real-time input meteorological parameters, including rain intensity, fog density and air refractive index, effectively simulating the refraction of light on the surface of glass and insulators in rainy weather, improving visual realism, dynamically correcting the intersection coordinates and normal direction of light and object surface according to the refraction offset in the global illumination propagation path calculation, eliminating the illumination misalignment caused by meteorological influence, ensuring that the light and shadow are accurately aligned with the object outline, recalculating the intensity attenuation of transmitted and reflected light according to the corrected illumination path, and updating the shader illumination parameters of the affected area, realizing the natural transition of brightness and highlights of translucent materials under complex weather conditions, and enhancing physical credibility; Furthermore, in the rendering system, meteorological parameters of the current scene are obtained in real time by connecting to the meteorological simulation module or external data interface. Rainfall intensity is expressed in millimeters per hour, with an input range of 0 to 50 mm / h, corresponding to light rain to heavy rain; fog density is represented by the visibility attenuation coefficient (β), ranging from 0.001 to 0.1m. -1This corresponds to light to dense fog; the air refractive index (n) is set to 1.000293 based on standard atmospheric conditions and finely adjusted with temperature and humidity. Based on Snell's law and Mie scattering theory, the path curvature of light propagating in rain and fog is calculated. For transparent or translucent surfaces (such as insulators or glass covers), the cumulative refractive angle offset Δθ per unit distance is calculated using light step integration, based on the ratio of the material's refractive index to the current air refractive index and the distribution density of raindrops or fog particles. Typical values ​​range from 0.1° to 5° and increase with meteorological parameters. The calculation results are expressed as the world space offset per pixel. Output in quantitative form; during the global illumination propagation calculation phase, pre-generated refraction offset data is read. For each ray emitted from the light source or camera, when it intersects with a surface marked as transparent or translucent, the corresponding refraction offset vector is queried based on the surface material type and current meteorological parameters at the intersection point. Based on this vector, two key geometric data are dynamically corrected: first, the coordinates of the intersection point between the ray and the surface are finely adjusted along the surface normal direction and the tangent plane direction, with the adjustment magnitude proportional to the magnitude of the offset vector; second, the effective normal direction at the intersection point is simulated by weighted mixing of the original normal and the offset direction to mimic the effect of adhesion at the medium interface. The actual orientation changes caused by water droplets or dirt are corrected simultaneously during the ray tracing or light probe interpolation stages to ensure that the propagation direction of light conforms to the physical laws under current meteorological conditions when entering, passing through, and leaving complex media layers. After completing the light path correction, the intensity attenuation of light is recalculated based on the new propagation path. For transmitted light, the exponential attenuation of light intensity with distance is calculated using the Beer-Lambert law combined with the current fog density coefficient, based on the corrected path length. For reflected light, the Fresnel reflection coefficient is recalculated based on the corrected normal direction and incident angle. All of the above calculations are performed within the G... The process is completed in real time in the PU's ComputeShader or ray tracing shader. Subsequently, the updated key lighting parameters, including the transmitted light intensity attenuation factor, the reflected light intensity coefficient, and the changes in indirect lighting contribution due to path curvature, are encapsulated into a structure and dynamically passed to the surface shader through the UniformBuffer. Based on the key lighting parameters, the shader uses the corrected values ​​to mix when calculating specular reflection, ambient light occlusion, and subsurface scattering. Ultimately, the visual performance of transparent and translucent objects under complex weather conditions such as rain and fog is consistent with physical laws in terms of brightness, color saturation, and specular shape. Step 5 specifically includes: adding a physically based scattering medium simulation layer in the rasterization rendering pipeline. This layer receives the meteorological density and lighting direction of the current scene as input, significantly enhancing the atmosphere and spatial volume of rain and fog scenes and improving environmental integration. The scattering medium simulation layer simulates the multiple scattering and absorption processes of light in rain, fog and other media based on the light stepping algorithm, calculates the contribution value to the final pixel color, so that the light is naturally diffused in the medium, producing a soft scattering halo, and the visual effect is more realistic. The calculated scattering contribution value is superimposed and fused with the conventional rasterization shading result of the object surface to achieve physical superposition of the scattering light of the medium and the surface material, improving the visual consistency and naturalness of the overall picture. In addition, step 5 also includes: dynamically adjusting the light transmittance and scattering coefficient in the scattering medium according to the material properties of transparent and translucent objects, ensuring that the light transmission and scattering response of different materials conforms to the real physical behavior, improving the visual distinction of materials in the medium, and establishing a surface dirt or water accumulation model for power equipment structures including insulators and glass, further correcting the mixing ratio of light refraction, reflection and internal scattering on the surface, simulating the complex influence of surface attachments on light, enhancing the realism of the equipment appearance in wet or polluted conditions, feeding back the corrected material light response parameters to the scattering medium simulation layer and surface shader, realizing the coupled calculation of medium scattering and object material response, realizing bidirectional light interaction between materials and medium, and improving the overall scene lighting consistency and physical credibility; Furthermore, within the Unity engine's general rendering pipeline framework, an independent, physically based scattering medium simulation layer is constructed before the post-processing stage. This layer serves as a computational channel within the programmable rendering pipeline, bound to the camera component of the current scene via scripts, and receives structured data input in real-time from the weather simulation module. The input data includes the current scene's weather density parameters: fog density coefficient, rain intensity parameters, and the direction vector of the scene's main light source. During initialization, this layer allocates a RenderTexture of appropriate size based on the camera's view frustum parameters and the current rendering resolution to store intermediate calculation results. The shader program is designed as a full-screen post-processing shader, called in the OnRenderImage callback, ensuring execution occurs after the main geometry rasterization and before final tone mapping and compositing. In the fragment shader of the scattering medium simulation layer, a volume rendering integral based on a ray stepping algorithm is implemented. Starting from the camera ray corresponding to the current pixel, it iterative sampling with a fixed step size is performed along the ray direction within the view frustum depth range. In each sampling step, based on the world coordinates of the current sampling point, a pre-calculated or real-time updated 3D weather density field is queried to obtain the local scattering coefficient at that point. (Linear mapping from fog density) and absorption coefficient (Typical value) The scattering probability between the current light direction and the light source direction is calculated based on the Henyey-Greenstein phase function (the anisotropy factor g is set according to the weather type: -0.2 to 0.2 for fog, and 0.6 to 0.8 for rain). This is done by accumulating the transmittance at each step (based on the Beer-Lambert law). With the intensity of emitted light ( ), Step size, The phase function is used for iterative calculation until the ray reaches its endpoint. Ultimately, two key values ​​are output: the cumulative contribution of scattered light along the path. (RGB vector) and transmittance reaching the camera After the calculation is completed, the contribution value of the scattered light is... Surface color output from a conventional rasterization channel Physically based overlay fusion is performed, and the fusion process is completed in the final output stage of the scattering medium simulation layer, sampling the current pixel from the main frame buffer. And the corresponding depth value, the fusion formula is: ,in, This represents the degree to which the scene medium attenuates the color of the background surface, thus achieving a sense of deep decay in the fog effect; The halo and brightness represent the light emitted by the scattering of light by the medium itself; to ensure energy conservation... The intensity of each channel is limited to a reasonable range, and normalization is performed based on the intensity of the main light source and the density of the medium in the scene to avoid oversaturation. The resulting fusion is... A new RenderTexture is written and used as input for subsequent post-processing. The execution time of the entire scattering layer is constrained by dynamically adjusting the step size and downsampling strategy to ensure that the GPU execution time does not exceed 2 milliseconds at the target frame rate. Step 6 specifically includes: constructing a multi-dimensional dynamic parameter vector with electromagnetic interference intensity coefficient, visual cone region weight, and weather influencing factors as core inputs to achieve full-dimensional quantitative perception of environmental status, providing accurate and unified input basis for intelligent decision-making; inputting the multi-dimensional dynamic parameter vector into a pre-trained linear regression optimization model, which outputs a set of real-time rendering parameters for controlling illumination, shadows, refraction, and scattering, achieving millisecond-level intelligent prediction from complex environments to optimal rendering parameters, replacing tedious manual parameter tuning; after each rendering frame, performing pixel-level difference analysis between the rendered output image and a baseline template image generated under ideal conditions; if the difference exceeds the tolerance, feeding the difference data back to the linear regression optimization model to fine-tune the model weights, forming a closed-loop adaptive system, giving the system the ability to learn and continuously optimize, ensuring the stability and improvement of visual quality under long-term operation; Furthermore, a set of precisely quantified multidimensional dynamic parameter vectors serves as the core input driver. This vector is constructed in real-time for each rendering frame. The dimensions and data source are defined as follows: electromagnetic interference intensity coefficients are collected by broadband electromagnetic sensors deployed in the simulation scene, and the amplitude of the main peak in the spectrum is analyzed in real-time. After normalization, these coefficients are mapped to a range of [0.0, 1.0], where 1.0 represents a preset peak interference threshold. The weights of the view frustum region are determined based on the three-dimensional vector, representing the core rendering area (0-5 meters), the transition rendering area (5-15 meters), and the distant rendering area (>15 meters), respectively. Real-time rendering resource allocation priority weights are dynamically calculated and normalized based on the sum of the screen pixel occupancy of visible devices in each area, ensuring the sum is 1.0. Weather influencing factors, including fog density attenuation coefficient and rainfall intensity coefficient, are determined based on a two-dimensional vector, provided by the meteorological simulation module at a frequency of 10 times per second. The three core parameters are synchronously sampled in each frame and together constitute the input vector. The constructed input vector is fed in real time into a pre-trained lightweight linear regression optimization model. This model uses over 100,000 sets of data covering different electromagnetic environments, observation angles, and meteorological conditions in the offline phase. The model is trained using rendering scene data and corresponding optimal manually adjusted parameters to learn the mapping relationship from complex environmental states to the optimal rendering parameter set. During online runtime, the model performs rapid inference via forward propagation, outputting a set of 8-dimensional real-time rendering control parameter vectors. These 8 parameters include: global illumination sampling interval multiplier, core rendering area shadow map resolution level, adjustment offset of the transition rendering area illumination blending weight coefficient, step scaling factor for scattering medium simulation, cardinality of bilateral filter depth difference weights, and fine-tuning values ​​for interpolation weights for metal and insulator materials. After model output, the real-time rendering control parameters are immediately read by the corresponding rendering subsystem and applied to the rendering calculation of the next frame, achieving frame-level dynamic adaptation of the rendering configuration. After each frame renders and outputs the final image, an asynchronous verification and feedback process is initiated. The rendered output image and a high-precision benchmark template image of the same scene pre-rendered under standard ideal conditions (no electromagnetic interference, clear weather, frontal view) are subjected to rapid pixel-level difference analysis on the GPU. The analysis uses a combination of weighted structural similarity index and absolute difference sum to calculate an overall difference score. Preset a difference tolerance threshold (0.05), if If the current rendering output deviates significantly from the ideal visual fidelity, then the input vector of this frame, the output rendering parameters, and the calculated overall difference score are encapsulated into a training sample and added to a circular buffer. When the number of accumulated samples in the buffer reaches 32, the weights of the linear regression optimization model are fine-tuned online using a small batch gradient descent with the new samples. The learning rate is set to a low 1e-4 to prevent oscillations, forming a closed loop of rendering-analysis-feedback-adjustment. This allows the model to continuously adapt to unforeseen environmental combinations, ensuring that the system maintains adaptive stability in visual quality and performance during long-term operation.

[0021] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0022] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A rendering method for ultra-high voltage power stations based on real-time global illumination and rasterization hybrid acceleration, characterized in that, Includes the following steps: Step 1: In a strong electromagnetic interference environment, the illumination sampling frequency is dynamically adjusted and pre-calculated global illumination data is inserted by combining an adaptive electromagnetic interference filtering algorithm with a global illumination probe dynamic interpolation compensation algorithm. Step 2: Divide the view frustum region according to the distance between the power equipment and the viewpoint, and dynamically allocate global illumination and rasterization rendering weights to different regions through the view frustum layered lighting fusion algorithm to achieve a balance and optimization between high-fidelity lighting in the core rendering area and resource usage in the distant view. Step 3: Use depth map comparison technology to detect the consistency of light and shadow transitions between near and far distances, and dynamically adjust the edges of shadows and lighting through rasterization depth calibration algorithm to eliminate visual tearing and depth perception confusion caused by rendering model switching. Step 4: For complex meteorological conditions, a weather physics refraction simulation algorithm is introduced to dynamically correct the propagation path and intensity of global illumination on transparent and semi-transparent surfaces. Step 5: In the rasterization rendering stage, a real-time scattering medium rasterization compensation algorithm is introduced to simulate the scattering effect of light in rain and fog media under complex weather conditions, and to dynamically adjust the material light response of transparent and semi-transparent objects. Step 6: Integrate multi-dimensional parameters of electromagnetic interference, distance layering, and weather factors to construct a dynamic rendering parameter optimization model. Through the pixel difference feedback between real-time rendering results and the benchmark template, achieve adaptive fine-tuning of parameters and closed-loop verification of system stability.

2. The ultra-high voltage power station rendering method based on real-time global illumination and rasterization hybrid acceleration as described in claim 1, characterized in that: Step 1 specifically includes: The noise spectrum characteristics under strong electromagnetic environment are collected in real time by electromagnetic sensors, and the sampling time interval of the global illumination probe is dynamically adjusted according to the spectrum peak value to suppress high-frequency illumination jitter. Between two adjacent lighting samples, linear interpolation is performed based on historical sampling data and a pre-computed static global illumination map to generate a continuous and smooth lighting data stream; Based on the material type of the device surface, different interpolation weights and filtering intensity thresholds are set for metal and insulating surfaces respectively to achieve material-aware adaptive illumination stabilization.

3. The ultra-high voltage power station rendering method based on real-time global illumination and rasterization hybrid acceleration as described in claim 2, characterized in that: Step 1 further includes: Based on the physical material properties of power equipment, we established illumination response models for high reflectivity of metal surfaces and diffuse reflectivity of insulator surfaces, and calculated the sensitivity coefficients to electromagnetic noise respectively. Based on the aforementioned sensitivity coefficient, a high interpolation weight and a narrow filtering bandwidth are configured for the metal surface to preserve high-frequency illumination details, and a smooth interpolation curve is configured for the insulator surface. In the rendering pipeline, the corresponding lighting stabilization algorithm branch is dynamically selected through shader variants to perform differentiated anti-interference processing on the surfaces of power equipment made of different materials.

4. The ultra-high voltage power station rendering method based on real-time global illumination and rasterization hybrid acceleration as described in claim 1, characterized in that: Step 2 specifically includes: Centered on the observer's viewpoint, the three-dimensional view frustum is divided into a core rendering area, a transition rendering area, and a distant rendering area based on a preset distance threshold. Allocate high-priority real-time global illumination computing resources to devices within the core rendering area and employ full-resolution rasterization rendering to maintain high fidelity in lighting and shadows within the core rendering area. The system dynamically adjusts the blending ratio of global illumination and rasterization rendering between the transition rendering area and the distant rendering area, and gradually reduces the frequency of lighting updates and the resolution of shadow maps.

5. The ultra-high voltage power station rendering method based on real-time global illumination and rasterization hybrid acceleration as described in claim 4, characterized in that: Step 2 also includes: Establish a hybrid weighting function based on device distance and screen space occupancy to calculate the global illumination contribution required for each rendered object in the transition rendering zone and the distant rendering zone in real time; In the transition rendering area, the global illumination contribution is weighted and mixed with the pre-calculated illumination probe data, and the real-time global illumination data is updated once every fixed number of frames. In the distant rendering area, pre-computed global illumination maps are used entirely, and real-time rasterization shadow calculations are turned off, concentrating rendering resources on the core rendering area.

6. The ultra-high voltage power station rendering method based on real-time global illumination and rasterization hybrid acceleration as described in claim 1, characterized in that: Step 3 specifically includes: In each rendering frame, the first depth map generated by global illumination and the second depth map generated by rasterization rendering are acquired synchronously. The two depth maps are compared pixel by pixel to identify areas where the depth value difference exceeds a preset threshold, and these areas are marked as potential visual tearing areas. For pixels within the potential visual tear zone, a depth edge smoothing algorithm based on bilateral filtering is used to perform transition calibration on the depth value and adjacent illumination intensity to eliminate discontinuous visual jumps.

7. The ultra-high voltage power station rendering method based on real-time global illumination and rasterization hybrid acceleration as described in claim 1, characterized in that: Step 4 specifically includes: Based on real-time input meteorological parameters, including rainfall intensity, fog density, and air refractive index, calculate the theoretical refraction offset of light on transparent and semi-transparent media surfaces under the current environment; In the calculation of the propagation path of global illumination, the coordinates of the intersection point of the light ray and the object surface and the normal direction are dynamically corrected based on the refraction offset. Based on the corrected lighting path, the intensity attenuation of transmitted and reflected light is recalculated, and the shader lighting parameters of the affected area are updated.

8. The ultra-high voltage power station rendering method based on real-time global illumination and rasterization hybrid acceleration as described in claim 1, characterized in that: Step 5 specifically includes: In the rasterization rendering pipeline, a physically based scattering medium simulation layer is added, which receives the current scene's weather density and lighting direction as input; The scattering medium simulation layer simulates the multiple scattering and absorption processes of light in the medium based on the light stepping algorithm, and calculates the contribution value to the final pixel color. The calculated scattering contribution value is superimposed and fused with the conventional rasterization coloring result of the object surface.

9. The ultra-high voltage power station rendering method based on real-time global illumination and rasterization hybrid acceleration as described in claim 8, characterized in that: Step 5 further includes: Based on the material properties of transparent and translucent objects, the light transmittance and scattering coefficient in the scattering medium are dynamically adjusted. For power equipment structures that include insulators and glass, a surface contamination or water accumulation model is established to further correct the mixing ratio of light refraction, reflection and internal scattering on the surface. The corrected material light response parameters are fed back to the scattering medium simulation layer and the surface shader.

10. The ultra-high voltage power station rendering method based on real-time global illumination and rasterization hybrid acceleration as described in claim 1, characterized in that: Step 6 specifically includes: Construct a multidimensional dynamic parameter vector with electromagnetic interference intensity coefficient, apparent cone region weight and weather influence factors as core inputs; A multidimensional dynamic parameter vector is input into a pre-trained linear regression optimization model, which outputs a set of real-time rendering parameters to control lighting, shadows, refraction, and scattering. After each rendering frame, the rendered output image is compared with a baseline template image generated under ideal conditions at the pixel level. If the difference exceeds the tolerance, the difference data is fed back to the linear regression optimization model to fine-tune the model weights and form a closed-loop adaptive system.

Images