Real-time game image optimization method based on intelligent layered rendering

Abstract

The invention discloses a real-time game image optimization method based on intelligent layered rendering, and particularly relates to the technical field of rendering, which comprises the following steps: carrying out frame rate monitoring on a real-time game image through a frame rate monitoring tool, and preliminarily identifying whether a rendering strategy is adaptive or not; extracting features of the images with the unmatched signs to obtain a feature group I and a feature group II; the first feature group generates a rendering complexity index through a pre-trained machine learning model, and the second feature group performs resource analysis to generate a resource load index; determining an optimization strategy type by using fuzzy logic in combination with the rendering complexity index, the resource load index and the target frame rate; the problem that the rendering strategy is not matched can be found in time, the rendering complexity and the resource consumption condition of the current game image can be known through the rendering complexity index and the resource load index, the rendering optimization strategy is intelligently determined, the capability of adapting to different hardware platforms is achieved, and the method and the device are convenient to use. And thus, players can obtain smoother experience in the game process.

Description

Technical Field

[0001] The present invention relates to the field of rendering technology. More specifically, the present invention relates to a real-time game image optimization method based on intelligent hierarchical rendering. Background Art

[0002] The core goal of game image rendering is to generate high-quality images and achieve fast feedback through real-time rendering technology. Traditional real-time rendering methods usually adopt pixel-based processing and rely on the computing power of the GPU. However, with the improvement of image quality requirements, the addition of effects such as physical lighting, detailed textures, and high dynamic range imaging (HDR) has led to a significant increase in the amount of computation required for each frame of rendering, thus affecting the frame rate and response time of the game.

[0003] In real-time rendering, there is a natural contradiction between image quality and rendering performance. To ensure the real-time response of the game, it is usually necessary to make certain compromises on image quality. For example, measures such as reducing texture details, reducing lighting effects, and reducing shadow quality can effectively reduce the amount of computation, but may also affect the visual effects of the game. How to determine whether the rendering strategy applied to the current scene has sufficient adaptability, and how to optimize the rendering strategy when the rendering strategy applied to the current scene does not have sufficient adaptability, these new technical challenges have emerged. Therefore, the present invention proposes a real-time game image optimization method based on intelligent hierarchical rendering in order to solve the above problems. Summary of the Invention

[0004] To achieve the above object, the present invention provides the following technical solutions: A real-time game image optimization method based on intelligent hierarchical rendering, comprising the following steps: Monitor the frame rate of the real-time game image through a frame rate monitoring tool, and perform frame rate analysis on the optimization effect of the real-time game image according to the monitored frame rate data to initially identify whether there are early signs of inappropriate rendering strategy for the real-time game image; Based on the initial identification result, for the game image with early signs of inappropriate rendering strategy, feature acquisition will be performed, and then Feature Group One and Feature Group Two will be obtained; Substitute Feature Group One into the pre-trained Machine Learning Model One to generate a rendering complexity index, and perform resource analysis on Feature Group Two to generate a resource load index; Use fuzzy logic to determine the strategy type for optimizing the real-time game image based on the rendering complexity index, the resource load index, and the target frame rate.

[0005] In a preferred embodiment, performing frame rate analysis on the optimization effect of the real-time game image means: Under a fixed time window, in obtaining the real-time game image optimization effect, the frame rate data corresponding to each time point is acquired to obtain a time series of frame rate data, and then the absolute value of the difference between the corresponding data of two adjacent time points is calculated, and then the average value and standard deviation of all absolute values are calculated.

[0006] In a preferred embodiment, the preliminary identification of whether there are early signs of inappropriate rendering strategies in real-time game images refers to: The average value of all absolute values is compared with a preset threshold one, and at the same time, the standard deviation of all absolute values is compared with a preset threshold two. If it satisfies that the average value of all absolute values is less than or equal to the preset threshold one and the standard deviation of all absolute values is less than or equal to the preset threshold two, a normal signal is generated. If it does not satisfy that the average value of all absolute values is less than or equal to the preset threshold one and the standard deviation of all absolute values is less than or equal to the preset threshold two, an abnormal signal is generated. When the abnormal signal is generated, it indicates that there are early signs of operating faults during the operation of the conveyor idler.

[0007] In a preferred embodiment, for game images with early signs of inappropriate rendering strategies, feature acquisition will be performed, and then obtaining Feature Group One and Feature Group Two refers to: For game images with early signs of inappropriate rendering strategies, a first type of feature data for evaluating the complexity of the rendering task and a second type of feature data for evaluating the amount of resources consumed during rendering are extracted, and then Feature Group One composed of the first type of feature data and Feature Group Two composed of the second type of feature data are obtained.

[0008] In a preferred embodiment, Feature Group One composed of the first type of feature data includes the following: A unit polygon quantity P is set, and the geometric complexity is quantified according to the total number of polygons in the scene: ; represents the number of polygons of the i-th object in the current game image scene, n represents the number of objects in the scene, represents the geometric complexity of the current game image scene; The material complexity is quantified by the type of material and the texture quality: ; represents the number of textures used for the k-th material in the current game image scene, represents the preset quality weight corresponding to the k-th material, m represents the total number of material types in the current game image scene, represents the material complexity of the current game image scene; The light source complexity is quantified according to the light source type and quantity: ; represents the number of the j-th light source type in the current game image scene represents the preset influence weight corresponding to the j-th light source type, p represents the total number of light source types in the current game image scene represents the light source complexity of the current game image scene; Quantify the perspective complexity of the current game image scene according to the ratio of the number of objects in the viewing distance to the number of objects within the viewing range of the camera in the current game image scene ; Quantify the post-processing complexity by the number of post-processing effects: ; represents the preset effect weight corresponding to the u-th post-processing effect, q represents the number of post-processing effects adopted in the current game image scene represents the post-processing complexity corresponding to the current game image scene; Quantify the special effect complexity according to the number and types of special effects in the scene: ; represents the number of the h-th special effect type in the current game image scene represents the preset contribution weight corresponding to the h-th special effect type, H represents the total number of special effect types in the current game image scene represents the special effect complexity of the current game image scene.

[0009] In a preferred embodiment, the machine learning model one is a convolutional neural network model.

[0010] In a preferred embodiment, the acquisition logic of the resource load index is as follows: Extract multiple preset resource load metrics from the feature group two, and then substitute them into the following formula for calculation: ; is the time smoothing term, obtained by multiplying the resource load index of the game image scene corresponding to the previous moment by a preset time smoothing factor is the g-th resource load metric corresponds to a preset non-zero proportionality coefficient, and all resource load metrics correspond to a preset non-zero proportionality coefficient, and the sum of which is one, Z represents the total number of resource load metrics is a preset non-zero influence coefficient is the g-th resource load metric corresponds to a non-linear conversion function ; is the g-th resource load metric The corresponding sensitive adjustment value, is a non-zero constant; is an interaction term, which is obtained by calculating the product of any two resource load metrics (Product 1), then multiplying Product 1 by the correlation coefficient of the two resource load metrics to obtain Product 2, and finally summing up all Product 2s; represents the resource load index of the current game image scene.

[0011] In a preferred embodiment, the usage logic of fuzzy logic is as follows: Define the input variables as the rendering complexity index, the resource load index, and the target frame rate, define the output variable as the type of real-time game image optimization strategy, perform fuzzy processing on the input variables to convert the values of the input variables into fuzzy sets, perform fuzzy processing on the output variable to convert the value of the output variable into a fuzzy set, formulate fuzzy rules to describe the optimization requirements under different combinations of data types, and infer the fuzzy input variables through the fuzzy rules to determine the type of real-time game image optimization strategy required.

[0012] The technical effects and advantages of the present invention: By monitoring the real-time game frame rate, the present invention can promptly detect the problem of mismatched rendering strategies. When the frame rate is lower than expected or fluctuates, it indicates that the rendering calculation amount may be too high or the rendering strategy does not meet the requirements of the current game. This early warning mechanism can help developers or system optimization tools promptly identify problems and make adjustments. By detecting these signs of rendering mismatch as early as possible, adjustments can be made before the problems affect the game experience, thereby reducing performance problems caused by excessive or overly simple rendering and ensuring the smoothness and visual effects of the game.

[0013] By automatically extracting different features of the game image, Feature Group 1 and Feature Group 2 respectively conduct refined analysis on the rendering complexity and resource load, helping to understand the rendering complexity and resource consumption of the current game image. Through such feature acquisition and analysis, it is possible to more accurately evaluate which scenes require more resources and which can be optimized by simplification, avoiding unnecessary resource waste.

[0014] By inputting Feature Group 1 into a pre-trained machine learning model, a rendering complexity index can be generated. This is a complexity evaluation method learned based on a large amount of data, which can automatically predict the rendering complexity according to the characteristics of the real-time scene. This index can help the system identify which scenes or moments require a high image rendering quality and which can reduce the rendering precision, thereby dynamically adjusting the rendering details and optimizing the rendering efficiency. Through this method, the complexity and resource consumption of game image rendering can intelligently adapt to the capabilities of different hardware platforms, enabling the best experience on both low-end and high-end devices.

[0015] After analyzing the resources of Feature Group Two, a resource load index can be generated to evaluate the load of current image rendering on system hardware (especially GPUs, CPUs, and memory). By monitoring the resource load index, dynamic adjustment can be achieved. For example, when resources are scarce, the system will preferentially reduce complex rendering calculations, lower resource requirements, avoid system overload, and improve the stability of the game. Fuzzy logic intelligently determines the rendering optimization strategy based on the input parameters of the rendering complexity index, resource load index, and target frame rate. This decision-making mechanism can handle complex and ambiguous real-world situations, such as frame rate fluctuations and resource limitations, and provide more flexible optimization strategies.

[0016] The present invention can flexibly adjust the rendering strategy according to the real-time monitored frame rate data and resource load conditions, thereby ensuring high frame rates and high-quality images of the game on different devices. Players can obtain a smoother experience during the game, reducing the impact of stuttering, screen delay, etc. This optimization method can be adapted to different hardware platforms (such as PCs, mobile devices, game consoles, etc.), enabling the game to run smoothly on low-performance devices and achieving better rendering effects on high-performance devices, truly realizing cross-platform optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] For the convenience of those skilled in the art to understand, the present invention will be further described below in conjunction with the accompanying drawings; Figure 1 It is a schematic diagram of the real-time game image optimization method based on intelligent hierarchical rendering in the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0019] Refer to Figure 1 The following embodiments are obtained: Embodiment 1: A real-time game image optimization method based on intelligent hierarchical rendering includes the following steps: Monitor the frame rate of the real-time game image through a frame rate monitoring tool, and perform frame rate analysis on the optimization effect of the real-time game image according to the monitored frame rate data to initially identify whether there are early signs of inappropriate rendering strategies in the real-time game image; Based on the initial identification results, for game images with early signs of inappropriate rendering strategies, feature acquisition will be performed, and then Feature Group One and Feature Group Two will be obtained; Substitute Feature Group 1 into the pre-trained Machine Learning Model 1 to generate a rendering complexity index, and perform resource analysis on Feature Group 2 to generate a resource load index. Use fuzzy logic to determine the strategy type for real-time game image optimization based on the rendering complexity index, resource load index, and target frame rate.

[0020] Frame Rate Monitoring and Preliminary Frame Rate Analysis: Use a frame rate monitoring tool to monitor the frame rate (FPS) of game images in real time. This is one of the basic indicators of game image performance. Frame rate data can reflect the smoothness of game rendering. A low frame rate may cause game stuttering or latency, thus affecting the player experience. Through frame rate monitoring, preliminarily analyze whether there are early signs of inappropriate rendering strategies in the game. For example, a decreasing frame rate may indicate insufficient optimization of image rendering or that some rendering strategies cannot adapt to the hardware performance.

[0021] Feature Acquisition and Formation of Feature Groups: After identifying early signs of possible rendering mismatch, features related to rendering need to be extracted. These features usually include graphics processing unit (GPU) load, CPU usage, texture size, the number of rendered polygons, etc. By analyzing these features, the bottleneck of rendering performance can be better understood. Based on the results of the preliminary frame rate analysis, obtain a set of relevant image rendering features. These features can be divided into Feature Group 1 and Feature Group 2, where: Feature Group 1: Features related to rendering complexity (such as the number of rendering objects).

[0022] Feature Group 2: Features related to hardware resources (such as GPU / CPU load, memory consumption).

[0023] Generation of Rendering Complexity Index Based on Machine Learning Model: Use a pre-trained machine learning model to process Feature Group 1 and predict the rendering complexity of the image. Rendering complexity refers to the amount of computation required for image rendering under given hardware and rendering conditions. By generating a rendering complexity index, the complexity of the rendering task can be quantitatively represented. Through the machine learning model, combined with the rendering features of the image, predict the complexity of the rendering task, and thus help identify possible bottlenecks in the rendering process (such as overly complex lighting calculations, processing of a large number of polygons, etc.).

[0024] Generating a resource load index based on resource load analysis: Analyze Feature Group 2, with a focus on evaluating the usage of hardware resources (such as CPU, GPU, memory, etc.). By monitoring these resource loads, a resource load index can be generated to measure the degree of system resource consumption. This helps identify whether frame rate drops or poor game performance are caused by resource bottlenecks. Through the analysis of hardware resource usage, quantify the system load to further understand whether there are performance bottlenecks during the rendering process, such as excessive GPU load or excessive memory usage.

[0025] Determination of fuzzy logic application and optimization strategy: Fuzzy logic is a reasoning method based on fuzzy set theory that can handle uncertainty and ambiguity. Here, fuzzy logic determines the most suitable optimization strategy by comprehensively considering multiple factors such as the rendering complexity index, resource load index, and target frame rate. The rendering complexity index reflects the complexity of image rendering; the resource load index reflects the usage of system resources; the target frame rate is the ultimate optimization goal, that is, to ensure that the game achieves the expected smoothness as much as possible during operation. Through fuzzy logic, combining these factors, different optimization strategies can be obtained, such as adjusting rendering details (reducing texture quality, reducing lighting calculations, etc.), dynamically adjusting the resolution, optimizing resource scheduling, etc., thereby improving the frame rate performance of the game.

[0026] Frame rate analysis of the optimization effect of real-time game images refers to: Under a fixed time window, obtain the frame rate data corresponding to each time point in the optimization effect of real-time game images to obtain a time series of frame rate data, then calculate the absolute value of the difference between the corresponding data of adjacent two time points, and then calculate the average value and standard deviation of all absolute values.

[0027] Preliminarily identifying early signs of inappropriate rendering strategies in real-time game images refers to: Comparing the average value of all absolute values with a preset threshold one, and at the same time comparing the standard deviation of all absolute values with a preset threshold two. If it satisfies that the average value of all absolute values is less than or equal to the preset threshold one and the standard deviation of all absolute values is less than or equal to the preset threshold two, a normal signal is generated. If it does not satisfy that the average value of all absolute values is less than or equal to the preset threshold one and the standard deviation of all absolute values is less than or equal to the preset threshold two, an abnormal signal is generated. When the abnormal signal is generated, it indicates that there are early signs of operating faults during the operation of the conveyor idler.

[0028] By collecting and analyzing the rendering frame rate of real-time game images, the performance of the system at different time points can be monitored. By calculating the absolute value of the difference between adjacent frame rates, the magnitude of frame rate fluctuations can be revealed, thus reflecting the performance stability during image rendering. By quickly calculating the frame rate differences in the time series, the changes in system performance can be quickly captured. Calculating the average value and standard deviation of frame rate fluctuations can effectively identify whether the image rendering is stable and avoid sudden performance drops or stuttering phenomena.

[0029] By setting Threshold 1 (average value threshold) and Threshold 2 (standard deviation threshold), the system can be helped to automatically determine the stability of image rendering and preliminarily identify the adaptability of the rendering strategy: Threshold 1 (average value threshold): Used to measure the overall degree of frame rate fluctuations. If the average value of the differences between adjacent frame rates is too large, it may mean that there are large fluctuations in the system during the rendering process, and there may be performance bottlenecks or insufficient rendering optimizations.

[0030] Threshold 2 (standard deviation threshold): Used to measure the dispersion degree of frame rate fluctuations. If the standard deviation of the frame rate differences is too large, it indicates that the system frame rate fluctuations are unstable, which may lead to a decline in the user experience. Especially during games or graphics rendering, it will be manifested as stuttering or screen tearing and other phenomena.

[0031] Normal signal: If the average value and standard deviation of all frame rate differences are less than the preset thresholds, it is considered that the rendering strategy is well adapted, the system is in a stable state, and a normal signal is generated.

[0032] Abnormal signal: If the average value or standard deviation exceeds the preset threshold, it indicates that the frame rate fluctuations during the rendering process are too large, and there may be problems such as an inappropriate rendering strategy or uneven system resource allocation. At this time, an abnormal signal is generated to prompt that there may be performance problems or rendering failures in the system, and further optimization or adjustment is required.

[0033] By real-time monitoring of frame rate fluctuations, performance problems during the rendering process can be discovered in a timely manner, so as to take measures to optimize the rendering algorithm, resource allocation or graphics processing, and improve the game experience and the fluency of image rendering. By timely discovering potential rendering mismatch problems, phenomena such as stuttering and frame drops in the game can be reduced, and the user's immersion and game experience can be improved. By discovering problems with the adaptation of the rendering strategy or hardware resources at an early stage, optimization and adjustment can be carried out in advance to avoid more serious performance bottlenecks or system crashes.

[0034] Feature acquisition will be performed on game images with early signs of rendering strategy mismatch, and then what Feature Group 1 and Feature Group 2 refer to: For game images showing early signs of rendering strategy mismatch, extract a first type of feature data for evaluating the complexity of the rendering task and a second type of feature data for evaluating the amount of resources consumed during rendering. Then, obtain Feature Group One composed of the first type of feature data and Feature Group Two composed of the second type of feature data. The data contained in Feature Group One is intended to evaluate the complexity of the rendering task itself. The complexity of the rendering task typically involves the geometric structure of the scene, the complexity of the materials, the computational requirements for lighting and shadows, etc. By extracting these features, areas where the rendering tasks are more complex at the image or scene level and may cause rendering performance bottlenecks can be identified in advance. Feature Group Two focuses on the resources consumed during rendering, especially computational resources (such as GPU and CPU loads), memory usage, bandwidth consumption, etc. These features reflect the performance overhead during rendering, helping developers understand which parts of the actual rendering process may consume excessive resources, leading to a decrease in frame rate or an increase in latency. By extracting these two types of feature data and organizing them into Feature Group One and Feature Group Two respectively, early prediction of game image rendering can be carried out, understanding the potential complexity of the rendering task and the resource consumption situation. This information provides the necessary basis for optimizing the rendering process and can help achieve a smoother and more efficient rendering experience in different hardware environments. Specifically: Feature Group One focuses on the theoretical complexity of the rendering task (such as scene details, lighting requirements, etc.), and Feature Group Two focuses on the actual resource consumption during rendering (such as GPU / CPU loads, memory occupancy, etc.).

[0035] Feature Group One composed of the first type of feature data includes the following: Set a unit polygon quantity P and quantify the geometric complexity according to the total number of polygons in the scene: ; represents the number of polygons of the i-th object in the current game image scene, n represents the number of objects in the scene, represents the geometric complexity of the current game image scene; the geometric complexity measures the number of polygons in the scene, thereby evaluating the complexity of the geometric shape of the scene. This metric takes into account the number of objects in the scene and the number of polygons of each object. The more polygons, the greater the rendering workload and the higher the complexity.

[0036] Quantify the material complexity by the type of material and the texture quality: ; represents the number of texture maps used for the k-th material in the current game image scene, represents the preset quality weight corresponding to the k-th material, m represents the total number of material types in the current game image scene, Represents the material complexity of the current game image scene; the material complexity measures the complexity of the materials and textures in the scene. The number of materials, the number of textures used for each material, and the texture quality all affect the rendering complexity. High-quality textures and a greater variety of materials increase the rendering burden.

[0037] Quantify the light source complexity according to the type and number of light sources: ; Represents the number of the j-th light source type in the current game image scene, Represents the preset influence weight corresponding to the j-th light source type, and p represents the total number of light source types in the current game image scene. Represents the light source complexity of the current game image scene; the light source complexity measures the impact of the light source type and its number in the scene on rendering. The number and type of light sources determine the amount of lighting calculations in the scene. The calculation overhead of different light sources is different. For example, the calculation complexities of point light sources, parallel light sources, and spotlights may be different.

[0038] Quantify the perspective complexity of the current game image scene according to the ratio of the number of objects in the viewing distance to the number of objects within the camera's field of view in the current game image scene ; The perspective complexity measures the ratio of the number of objects within the camera's field of view to the number of objects in the viewing distance. This metric takes into account the number of visible objects that need to be rendered in the scene. If there are more objects within the field of view or the camera's viewing distance is large, this increases the rendering workload and results in higher complexity.

[0039] Quantify the post-processing complexity by the number of post-processing effects: ; Represents the preset effect weight corresponding to the u-th post-processing effect, and q represents the number of post-processing effects adopted in the current game image scene. Represents the post-processing complexity corresponding to the current game image scene; the post-processing complexity measures the number of post-processing effects in the scene and their impact. Post-processing effects (such as blur, halo, color adjustment, etc.) can significantly increase the rendering computational burden. Therefore, complex post-processing effects increase the complexity of the scene.

[0040] Quantify the special effect complexity according to the number and types of special effects in the scene: ; Represents the number of the h-th special effect type in the current game image scene, Represents the preset contribution weight corresponding to the h-th special effect type, and H represents the total number of special effect types in the current game image scene. Represents the special effect complexity of the current game image scene. The types and quantities of special effects (such as explosions, particle effects, etc.) directly affect the rendering complexity. Especially during real-time rendering, these special effects may require a large amount of computing resources.

[0041] Machine learning model one is a convolutional neural network model. All the first-class feature data in feature group one are jointly input into the convolutional neural network model, and the output result of the convolutional neural network model is the rendering complexity index. Convolutional Neural Network (CNN) is a deep learning model, especially suitable for processing data with grid structures, such as images, videos, or structured data similar to images. CNN extracts local features through convolutional layers and combines these features layer by layer to identify higher-level patterns. In the present invention, the role of CNN is to effectively extract and learn the input feature data. Through convolutional operations, CNN can extract useful information from complex features and then output a meaningful result.

[0042] Feature group one contains various rendering-related features (such as texture, lighting conditions, geometric shapes, etc.) of the image, and these features reflect some important factors in the rendering process. Different features may represent different dimensions or aspects of rendering, such as rendering speed, quality, complexity, etc. Inputting these features into the CNN together means that you hope to find associations and patterns from all these features through the model to predict the rendering complexity.

[0043] The rendering complexity index is a measure of the rendering process, which may refer to the computing resources, time, complexity, or difficulty of quality requirements required for rendering. The level of this index reflects the difficulty of rendering an image or scene. Generally, the higher the complexity, the more computing resources and time are required.

[0044] Input features: All the first-class feature data in feature group one (such as texture, lighting, object complexity, scene depth, etc.) will be used as input. Through multiple processes of convolutional layers, CNN can capture the spatial and hierarchical structural relationships between these features.

[0045] Output of the convolutional network: After being processed by multiple layers of networks such as convolutional layers and pooling layers, the CNN model will output the rendering complexity index. This index represents the rendering complexity inferred based on the input features.

[0046] Predicting Rendering Complexity: The advantage of this method is that the CNN can automatically learn the important information in the input features without the need for manually designing complex feature extraction rules. Through training, the model can predict the rendering complexity based on different combinations of input features. The application of this model can be used in computer graphics and rendering optimization. For example, when rendering a 3D scene or image, knowing the rendering complexity in advance can help developers optimize the rendering algorithm and allocate resources. In real-time rendering or game development, accurately predicting the rendering complexity is crucial for balancing image quality and performance and improving the user experience. The significance of this convolutional neural network model lies in that it automatically predicts the rendering complexity index for a given scene or task by learning a set of rendering-related features. This process can not only help quantify the rendering difficulty but also provide data support for optimizing the rendering algorithm and adjusting resource allocation, thereby enhancing the rendering efficiency and quality.

[0047] The acquisition logic of the resource load index is as follows: Extract multiple preset resource load metrics from Feature Group 2 and then substitute them into the following formula for calculation: ; is the time smoothing term, obtained by multiplying the resource load index of the game image scene corresponding to the previous moment by a preset time smoothing factor. The role of this term is to balance the influence of the resource load index of the previous moment on the current moment through the time smoothing factor. is the g-th resource load metric corresponds to a preset non-zero proportionality coefficient, and all resource load metrics The sum of the corresponding preset non-zero proportionality coefficients is one. Each resource load metric (usually different metrics related to system resource usage, such as CPU usage, memory usage, GPU load, etc.) has a corresponding proportionality coefficient, and these proportionality coefficients play different roles when weighted and summed. The sum of all the coefficients is 1, ensuring that the final weighted result is within a certain range.

[0048] Z represents the total number of resource load metrics is a preset non-zero influence coefficient. is the g-th resource load metric corresponds to a non-linear conversion function. ; is the g-th resource load metric corresponds to a sensitivity adjustment value. is a non-zero constant; the non-linear conversion function is to cope with the non-linear characteristics of different resource loads in different load intervals. The larger the sensitivity adjustment value, the more sensitive some resource load indicators may be to the system load, or these sensitivities can be adjusted according to specific business requirements. is an interaction term. By calculating the product of any two resource load indicators Product One, and then multiplying Product One by the correlation coefficient of these two resource load indicators to obtain Product Two, and finally summing up all Product Twos; the interaction term represents the mutual influence between different resource load indicators For example, CPU load and memory load may be related. When both are high, their impact on the system may be greater than when they are high separately. Through the interaction term, the model can capture the complex relationships between different resource loads, making the model not just a simple weighted sum, but also taking into account the dependencies and influences between various resources. represents the resource load index of the current game image scene. The resource load index model comprehensively evaluates the current resource load status through multiple different factors, considering the smoothing of historical loads, the weights of various resources, the non-linear relationships between various resource load indicators, and their interaction effects, so as to more accurately reflect the resource load status in the game image scene.

[0049] The usage logic of fuzzy logic is as follows: Define the input variables as the rendering complexity index, resource load index, and target frame rate, define the output variable as the strategy type for real-time game image optimization, perform fuzzy processing on the input variables to convert the values of the input variables into fuzzy sets, perform fuzzy processing on the output variable to convert the value of the output variable into a fuzzy set, formulate fuzzy rules to describe the optimization requirements under different combinations of data types, and infer the fuzzy input variables through the fuzzy rules to determine the strategy type for real-time game image optimization that needs to be performed.

[0050] Based on the rendering complexity index, resource load index, and the stable frame rate number expected by the user, select the intelligent hierarchical rendering strategy through fuzzy logic. The rendering can be divided into different levels (hierarchical) to ensure that the target frame rate (such as 60 FPS) is achieved under different load conditions. In this strategy, different rendering adjustment levels and strategies can be dynamically adjusted according to the current rendering load and resource status, so as to optimize the balance between performance and visual effects.

[0051] Input factors: Rendering complexity index: Reflects the complexity of the current scene, usually considering factors such as the number of objects, lighting, shadows, and post-processing effects in the scene.

[0052] Resource Load Index: Reflects the current hardware load of the system and determines whether to reduce the image quality to relieve the computing pressure.

[0053] Stable Frame Rate Expected by Users: Such as the target frame rate of 60 FPS, which serves as the benchmark for performance adjustment.

[0054] These input factors are evaluated through a fuzzy logic system to determine the rendering strategy that the current system should adopt.

[0055] Hierarchical Structure of the Intelligent Hierarchical Rendering Strategy: Level 1: Slight Optimization (Low Load, Target Frame Rate Close): When the resource load is low, or there is a certain margin between the load and the expected frame rate, the rendering strategy mainly focuses on maintaining a high visual quality and making slight optimizations.

[0056] Resolution Scaling: Slightly reduce the resolution (such as reducing it to 90% or 95%) to lower the GPU load while maintaining a high visual effect.

[0057] Anti-Aliasing: Use a lower level of anti-aliasing effect (such as FXAA, SMAA), or reduce some details as needed.

[0058] Reduce Post-Processing Effects: Some resource-intensive post-processing effects (such as tone mapping, depth of field effects) can be appropriately reduced to release resources without affecting the main visual effect.

[0059] Level 2: Medium Optimization (Medium Load, Low Target Frame Rate): When the system resource load is high, or the rendering complexity is large and the expected frame rate cannot be fully maintained, more obvious optimizations start to be carried out.

[0060] Resolution Scaling: Reduce the resolution to 80% or 70% to lower the GPU load.

[0061] Shadow Quality: Reduce the shadow quality to a low resolution, reduce shadow details, or disable some unimportant shadow effects.

[0062] Turn off Some Post-Processing Effects: Disable anti-aliasing, disable or reduce effects such as motion blur, screen space reflection (SSR), etc.

[0063] LOD Optimization: Use low-polygon models for distant objects, reduce the level of detail, thereby reducing the computing burden on the GPU.

[0064] Level 3: Significant Optimization (High Load, Difficult to Achieve Target Frame Rate): When the resource load is close to saturation and the expected frame rate is difficult to achieve, a more drastic optimization strategy is adopted, sacrificing some visual effects to ensure a smooth gaming experience.

[0065] Substantial Resolution Scaling: Reduce the resolution to 60% or 50%, significantly reducing GPU load, but may affect visual effects.

[0066] Disable Shadows: Completely disable the shadow effect, or reduce the resolution and quality of shadows.

[0067] Turn Off All Post-Processing Effects: Completely turn off post-processing effects such as anti-aliasing, motion blur, depth of field, etc., and even reduce the complexity of image post-processing.

[0068] Stronger LOD Optimization: Use low-detail LOD for objects even at close range, further reducing the computational burden.

[0069] Reduce Material Details: Lower the texture quality (e.g., reduce the texture from high resolution to medium / low resolution).

[0070] Level 4: Extreme Optimization (Ultra-High Load, Extreme Conditions): In extreme load situations, it is necessary to minimize the image quality to ensure smoothness and responsiveness, even sacrificing almost all visual effects.

[0071] Ultra-Low Resolution: Reduce the resolution to 40% or lower of the original resolution to ensure that the system can maintain a high frame rate as much as possible.

[0072] Completely Disable Shadows and All Special Effects: Completely disable all high-load special effects such as shadows, reflections, lighting effects, etc.

[0073] Large-Scale LOD Optimization: All objects use low-polygon models, reducing the level of detail for all objects, near and far.

[0074] Disable Dynamic Lighting: Disable dynamic light sources and use static light sources and ambient lighting instead, reducing real-time calculations.

[0075] Fuzzy Logic Decision: Through the fuzzy logic system, the corresponding intelligent hierarchical rendering strategy can be automatically selected based on the following input conditions: Input Rule Design: The fuzzy logic system makes inferences based on the following rules: When the resource load is low and the frame rate is close to the target, use a mild optimization strategy (Level 1).

[0076] When the resource load is medium and the frame rate is slightly low, apply a medium optimization strategy (Level 2).

[0077] When the resource load is high and the frame rate is difficult to achieve, apply a significant optimization strategy (Level 3).

[0078] When the resource load is close to saturation and the frame rate is extremely low, use an extreme optimization strategy (Level 4).

[0079] Rendering complexity: The rendering complexity index of the current scene (e.g., the number of objects, lighting complexity, etc.). The difference between the expected frame rate and the current frame rate: Calculate the difference between the current frame rate and the target frame rate as an important decision-making basis.

[0080] Fuzzy inference rules: For example: Rule 1: If the resource load is low and the rendering complexity is low, output "slight optimization". Rule 2: If the resource load is high and the rendering complexity is high, output "significant optimization". Rule 3: If the target frame rate cannot be maintained and the resource load is extremely high, output "extreme optimization". These rules usually use fuzzy operations (such as: minimum value, weighted average, etc.) to solve. Each rule will derive a result based on the fuzzy values (membership degrees) of the input variables. After applying all the inference rules, the outputs of each rule are integrated through a synthesis method. Common fuzzy inference methods include the min-max inference method and the weighted average method. The weighted average method means that for the output generated by each rule, a weighted average is performed. The output result of each rule will be weighted and calculated according to its corresponding membership degree, and finally a comprehensive fuzzy result is obtained.

[0081] The result of fuzzy inference is often a fuzzy set and cannot be directly used for decision-making. Therefore, it is necessary to convert the fuzzy set back to a specific output value, and this process is called defuzzification. Common methods of defuzzification include: Centroid method: By calculating the centroid of the output fuzzy set, a specific value is obtained. The centroid method is the most commonly used defuzzification method. It performs a weighted average on the membership degree of the fuzzy set and the corresponding output values to obtain an exact value.

[0082] Maximum membership degree method: Select the value with the largest membership degree in the output fuzzy set as the final output. It is suitable for some relatively simple situations, but may not be as accurate as the centroid method.

[0083] Weighted average method: Similar to the weighted average in fuzzy inference, the final output value is obtained by weighting the fuzzy output results.

[0084] After defuzzification processing, an exact output value will be obtained, that is, the type of real-time game image optimization strategy is slight optimization, medium optimization, significant optimization or extreme optimization; through the above hierarchical rendering strategy, the visual effect of the game can be dynamically adjusted according to the resource load index, rendering complexity and target frame rate. The corresponding types of different levels of rendering optimization strategies are intelligently selected through the fuzzy logic system to achieve the balance between rendering performance and visual effect.

[0085] The above formulas are all dimensionless and take their numerical calculations. The formulas are obtained by collecting a large amount of data for software simulation to get a formula closest to the real situation. The preset parameters in the formulas are set by technicians in this field according to the actual situation.

[0086] It should be understood that in various embodiments of the present application, the magnitudes of the sequence numbers of the above processes do not imply the order of execution, and the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0087] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Skilled professionals can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present application.

[0088] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments, and will not be elaborated herein.

[0089] As described above, the above is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present application can easily think of changes or substitutions, and all should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A real-time game image optimization method based on intelligent hierarchical rendering, characterized in that It includes the following steps: Monitor the frame rate of real-time game images. According to the monitored frame rate data, conduct frame rate analysis on the optimization effect of real-time game images to preliminarily identify whether there are early signs of inappropriate rendering strategies in real-time game images; Based on the preliminary identification results, for game images with early signs of inappropriate rendering strategies, feature acquisition will be carried out, and then Feature Group 1 and Feature Group 2 will be obtained; Substitute Feature Group 1 into the pre-trained Machine Learning Model 1 to generate a rendering complexity index, and conduct resource analysis on Feature Group 2 to generate a resource load index; Use fuzzy logic to determine the type of optimization strategy for real-time game images based on the rendering complexity index, resource load index, and target frame rate.

2. The real-time game image optimization method based on intelligent hierarchical rendering according to claim 1, wherein, Conducting frame rate analysis on the optimization effect of real-time game images means: Under a fixed time window, obtain the frame rate data corresponding to each time point in the optimization effect of real-time game images to obtain a frame rate data time series, then calculate the absolute value of the difference between the corresponding data of adjacent two time points, and then calculate the average value and standard deviation of all absolute values.

3. The real-time game image optimization method based on intelligent hierarchical rendering according to claim 2, characterized in that Preliminarily identifying whether there are early signs of inappropriate rendering strategies in real-time game images means: Compare the average value of all absolute values with a preset Threshold 1, and at the same time compare the standard deviation of all absolute values with a preset Threshold 2. If it satisfies that the average value of all absolute values is less than or equal to the preset Threshold 1 and the standard deviation of all absolute values is less than or equal to the preset Threshold 2, a normal signal will be generated. If it does not satisfy that the average value of all absolute values is less than or equal to the preset Threshold 1 and the standard deviation of all absolute values is less than or equal to the preset Threshold 2, an abnormal signal will be generated. When the abnormal signal is generated, it indicates that there are early signs of operating faults during the operation of the conveyor idler.

4. The real-time game image optimization method based on intelligent hierarchical rendering according to claim 3, characterized in that, For game images with early signs of inappropriate rendering strategies, feature acquisition will be carried out, and then Feature Group 1 and Feature Group 2 will be obtained means: For game images with early signs of inappropriate rendering strategies, extract the first type of feature data used to evaluate the complexity of the rendering task and the second type of feature data used to evaluate the amount of resources consumed during rendering, and then obtain Feature Group 1 composed of the first type of feature data and Feature Group 2 composed of the second type of feature data.

5. The real-time game image optimization method based on intelligent hierarchical rendering according to claim 4, characterized in that, Feature Group 1 composed of the first type of feature data includes the following content: Set a unit polygon quantity P and quantify the geometric complexity according to the total number of polygons in the scene: ; represents the number of polygons of the i-th object in the current game image scene, and n represents the number of objects in the scene, represents the geometric complexity of the current game image scene; Quantify the material complexity through the type of material and texture quality: ; represents the number of texture maps used for the k-th material in the current game image scene, represents the preset quality weight corresponding to the k-th material, and m represents the total number of material types in the current game image scene, represents the material complexity of the current game image scene; Quantify the light source complexity according to the type and quantity of light sources: ; represents the number of the j-th light source type in the current game image scene, represents the preset influence weight corresponding to the j-th light source type, and p represents the total number of light source types in the current game image scene, represents the light source complexity of the current game image scene; Quantify the perspective complexity of the current game image scene according to the ratio of the number of objects in the viewing distance to the number of objects within the viewing range of the camera for the current game image scene ; Quantify the post - processing complexity by the number of post - processing effects: ; represents the preset effect weight corresponding to the \(u\) - th post - processing effect, \(q\) represents the number of post - processing effects adopted in the current game image scene, represents the post - processing complexity corresponding to the current game image scene; Quantify the special effect complexity according to the quantity and type of special effects in the scene: ; represents the number of the h-th special effect type in the current game image scene, represents the preset contribution weight corresponding to the h-th special effect type, and H represents the total number of special effect types in the current game image scene, represents the special effect complexity of the current game image scene.

6. The real-time game image optimization method based on intelligent hierarchical rendering according to claim 5, wherein Machine Learning Model 1 is a convolutional neural network model.

7. The real-time game image optimization method based on intelligent hierarchical rendering according to claim 6, wherein The acquisition logic of the resource load index is: Extract multiple preset resource load indicators from Feature Group 2 and then substitute them into the following formula for calculation: ； is the time smoothing term, obtained by multiplying the resource load index of the game image scene corresponding to the previous moment by a preset time smoothing factor. is the g-th resource load metric corresponds to a preset non-zero proportionality coefficient, and all resource load metrics correspond to a preset non-zero proportionality coefficient, and the sum of all such coefficients is one. Z represents the total number of resource load metrics ; is a preset non-zero influence coefficient, is the g-th resource load metric corresponds to a non-linear conversion function, ; is the g-th resource load metric corresponds to a sensitivity adjustment value, is a non-zero constant; It is an interaction term, which is obtained by calculating the product one of any two resource load indicators, then multiplying the product one by the correlation coefficient of the two resource load indicators to get the product two, and finally summing up all the product twos; It represents the resource load index of the current game image scene.

8. The real-time game image optimization method based on intelligent hierarchical rendering according to claim 7, characterized in that The usage logic of fuzzy logic is: Define the input variables as the rendering complexity index, resource load index, and target frame rate, define the output variable as the type of strategy for real-time game image optimization, perform fuzzy processing on the input variables, convert the values of the input variables into fuzzy sets, perform fuzzy processing on the output variable, convert the values of the output variable into fuzzy sets, formulate fuzzy rules to describe the optimization requirements under different combinations of data types, and infer the type of strategy for real-time game image optimization that needs to be performed by passing the fuzzified input variables through the fuzzy rules.

Images