An AI-driven method for fast fault localization of cloud rendering services

Abstract

The application discloses an AI-driven cloud rendering service fault rapid positioning method, relates to the technical field of service fault positioning, and is used for solving the problems of low efficiency and large resource waste of rendering retry behavior, acquiring a cloud rendering operation target, performing regional division, collecting the resolution and artifact area of each region, comprehensively calculating the cloud rendering fault indexes of the regions, evaluating fault region positioning features based on the fault indexes and storing the fault region positioning features in a temporary database, setting multiple rendering retries within a retry time, counting retry fault frequencies, detecting the rendering execution time of each time and calculating rendering levels, evaluating fault type positioning features in combination with the retry fault frequencies and the rendering levels, fusing the fault region positioning features in the temporary database, selecting an adaptive fault region or type positioning method for processing, intelligently analyzing multidimensional fault features, realizing rapid and accurate positioning of cloud rendering faults, and improving service stability and operation and maintenance efficiency.

Description

Technical Field

[0001] This invention relates to the field of service fault location technology, and more specifically, to an AI-driven method for rapid fault location of cloud rendering services. Background Technology

[0002] With the rapid development of cloud computing and artificial intelligence technologies, cloud-based graphics rendering services are widely used in scenarios such as games, film and television production, industrial simulation, and architectural visualization. Cloud rendering effectively decouples the computing power of terminal devices by distributing high-intensity graphics rendering tasks to distributed server clusters for execution.

[0003] The existing technology has the following shortcomings:

[0004] Currently, in the execution of actual rendering tasks, due to factors such as network latency, computing resource scheduling, graphics load complexity, and data consistency, there is a lack of dynamic identification capabilities for fault areas and fault types. This fails to meet the rapid response requirements of multi-region, high-complexity rendering tasks, resulting in low efficiency and significant resource waste in rendering retry behavior. Therefore, an AI-driven method for rapid fault location in cloud rendering services is proposed.

[0005] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a method for rapid fault location in AI-driven cloud rendering services. This method utilizes multi-dimensional image feature extraction, an adaptive retry statistics mechanism, and fault feature fusion analysis technology based on AI modeling to address the problems mentioned in the background.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a method for rapid fault location in AI-driven cloud rendering services, comprising the following steps:

[0008] Step S1: Obtain the cloud rendering operation target, divide the cloud rendering operation target into regions, and collect the regional resolution and artifact area of ​​each region.

[0009] Step S2: Calculate the cloud rendering fault index of each region by combining the regional resolution and artifact area of ​​each region, evaluate the fault area location characteristics of the cloud rendering operation target based on the cloud rendering fault index of each region, and input it into the temporary storage database.

[0010] Step S3: Set a retry time, perform multiple cloud rendering retries on the cloud rendering operation target within the retry time, and count the retry failure frequency of the cloud rendering operation target. Detect the rendering execution time of each cloud rendering retry, and calculate the rendering level of the cloud rendering operation target based on the rendering execution time.

[0011] Step S4: Based on the comprehensive assessment of the retry failure frequency and rendering level, evaluate the failure type location characteristics of the cloud rendering operation target, and combine the failure area location characteristics in the temporary storage database to select the failure area location method or failure type location method to process the cloud rendering operation target.

[0012] In a preferred embodiment, in step S1, the cloud rendering operation target is obtained through the cloud rendering platform, and the cloud rendering operation target is divided into a preset number of equal-area division regions.

[0013] The image processing tool library is used to convert the segmented region into a grayscale image and obtain the grayscale value of the pixel in the segmented region.

[0014] The gray values ​​of pixels in the divided region are convolved with the horizontal convolution kernel of the Sobel operator to obtain the gradient matrix of the pixel in the horizontal direction.

[0015] The gray values ​​of pixels in the divided region are convolved with the vertical convolution kernel of the Sobel operator to obtain the gradient matrix of the pixel in the vertical direction.

[0016] In a preferred embodiment, in step S1, the gradient magnitude of the pixel is calculated by combining the gradient matrix of the pixel in the horizontal direction and the gradient matrix of the pixel in the vertical direction.

[0017] The average gradient magnitude of all pixels in the divided region is used as the region resolution of the divided region.

[0018] The regions are divided and processed using the Canny edge detection algorithm to generate edge maps of the divided regions.

[0019] In a preferred embodiment, in step S1, the edge map of the divided region is a binary image, where the edge pixel value is 1 and the non-edge pixel value is 0.

[0020] Artifact regions are constructed by connecting edge pixels in the edge map, and the total number of pixels in all closed artifact regions is counted.

[0021] The artifact area is calculated by multiplying the total number of pixels by the preset pixel area.

[0022] In a preferred embodiment, in step S2, the ratio of the artifact area of ​​the segmented region to the region resolution is used as the cloud rendering failure index of the segmented region.

[0023] The average cloud rendering fault index of each region is used as the fault region location feature of the cloud rendering operation target.

[0024] The fault location features of the target cloud rendering operation are transmitted to a temporary storage database for storage.

[0025] In a preferred embodiment, in step S3, the cloud rendering operation target is retried multiple times within a preset retry time.

[0026] After each cloud rendering retry, the resolution of the current cloud rendering target is detected, and the average resolution of the cloud rendering target is calculated.

[0027] The average resolution is compared with the preset resolution baseline. If the average resolution is less than the resolution baseline, the rendering is considered to have failed.

[0028] Count all cloud rendering retries that were judged to have failed to render, and obtain the number of rendering failures.

[0029] The total number of cloud rendering retries is calculated by counting all cloud rendering retries completed within the retry time.

[0030] In a preferred embodiment, in step S3, the ratio between the number of rendering failures and the total number of cloud rendering retries is used as the retry failure frequency of the cloud rendering operation target.

[0031] Obtain the rendering execution time of each cloud rendering retry, and use the ratio of the rendering execution time of each cloud rendering retry to the preset rendering time calibration benchmark as the single rendering execution percentage of each rendering retry;

[0032] The average percentage of all single-render executions is used as the rendering level of the cloud rendering operation target.

[0033] In a preferred embodiment, in step S4, after standardizing the retry failure frequency and rendering level, the fault type localization feature is calculated using a weighted fusion method.

[0034] Extract the stored fault area location features from the temporary storage database and perform standardization processing. Calculate the difference between the fault type location features and the fault area location features to obtain the difference value.

[0035] In a preferred embodiment, in step S4, if the difference value is greater than a preset fault judgment threshold, then a fault type location method is selected for processing.

[0036] If the difference value is less than or equal to the preset fault judgment threshold, then the fault area location method is selected for processing.

[0037] The technical effects and advantages of this invention are as follows:

[0038] This invention acquires the target cloud rendering operation, divides the target into regions, collects the regional resolution and artifact area of ​​each region, calculates the cloud rendering fault index for each region based on the combined regional resolution and artifact area, evaluates the fault region location characteristics of the target cloud rendering operation based on the cloud rendering fault index of each region, and transmits the data to a temporary storage database. A retry time is set, and multiple cloud rendering retries are performed on the target cloud rendering operation within the retry time, and the retry failure frequency of the target cloud rendering operation is statistically analyzed. The rendering execution time of each cloud rendering retry is detected, and the rendering level of the target cloud rendering operation is calculated based on the rendering execution time. The fault type location characteristics of the target cloud rendering operation are evaluated by combining the retry failure frequency and rendering level. The fault region location characteristics in the temporary storage database are combined to select either a fault region location method or a fault type location method for processing the target cloud rendering operation. Through intelligent analysis of multi-dimensional fault characteristics, rapid and accurate fault location in cloud rendering services is achieved, improving service stability and operational efficiency, and effectively reducing resource waste caused by rendering anomalies. Attached Figure Description

[0039] Figure 1 This is a flowchart illustrating the implementation of an AI-driven method for rapid fault location in cloud rendering services according to the present invention.

[0040] Figure 2 This is a schematic diagram illustrating the steps of a method for rapid fault location in an AI-driven cloud rendering service according to the present invention. Detailed Implementation

[0041] 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.

[0042] This invention acquires the target cloud rendering operation, divides the target into regions, collects the regional resolution and artifact area of ​​each region, calculates the cloud rendering fault index for each region based on the combined regional resolution and artifact area, evaluates the fault region location characteristics of the target cloud rendering operation based on the cloud rendering fault index of each region, and transmits the data to a temporary storage database. A retry time is set, and multiple cloud rendering retries are performed on the target cloud rendering operation within the retry time, and the retry failure frequency of the target cloud rendering operation is statistically analyzed. The rendering execution time of each cloud rendering retry is detected, and the rendering level of the target cloud rendering operation is calculated based on the rendering execution time. The fault type location characteristics of the target cloud rendering operation are evaluated by combining the retry failure frequency and rendering level. The fault region location characteristics in the temporary storage database are combined to select either a fault region location method or a fault type location method for processing the target cloud rendering operation. By combining intelligent analysis of multi-dimensional fault characteristics, rapid and accurate fault location in cloud rendering services is achieved, improving service stability and operational efficiency.

[0043] Example 1: A method for rapid fault location in AI-driven cloud rendering services, such as... Figures 1 to 2 As shown, it includes the following steps:

[0044] Step S1: Obtain the cloud rendering operation target, divide the cloud rendering operation target into regions, and collect the regional resolution and artifact area of ​​each region.

[0045] Step S2: Calculate the cloud rendering fault index of each region by combining the regional resolution and artifact area of ​​each region, evaluate the fault area location characteristics of the cloud rendering operation target based on the cloud rendering fault index of each region, and input it into the temporary storage database.

[0046] Step S3: Set a retry time, perform multiple cloud rendering retries on the cloud rendering operation target within the retry time, and count the retry failure frequency of the cloud rendering operation target. Detect the rendering execution time of each cloud rendering retry, and calculate the rendering level of the cloud rendering operation target based on the rendering execution time.

[0047] Step S4: Based on the comprehensive assessment of the retry failure frequency and rendering level, evaluate the failure type location characteristics of the cloud rendering operation target, and combine the failure area location characteristics in the temporary storage database to select the failure area location method or failure type location method to process the cloud rendering operation target.

[0048] The specific implementation is as follows:

[0049] In step S1, the cloud rendering operation target is obtained through the cloud rendering platform. The cloud rendering operation target refers to the target input image data for the cloud rendering service to perform image rendering processing.

[0050] In the rapid fault localization process of cloud rendering services, a matching localization method is selected for different types of anomalies. Cloud rendering fault types include fault area localization and fault type localization. Fault area localization is used to identify local areas with quality anomalies in the cloud rendering operation target. If the area resolution is lower and the artifact area is larger, it indicates that the rendering quality of the area is worse, and the cloud rendering fault index of the corresponding area is larger, and the fault area localization feature is stronger. Fault type localization is used to identify overall rendering anomalies of the cloud rendering operation target caused by systemic problems. The higher the retry failure frequency and the longer the rendering execution time, the higher the rendering level, and the stronger the corresponding fault type localization feature.

[0051] The cloud rendering operation target is divided into a preset number of equal-area regions. The preset number can be configured based on the resolution information, analysis accuracy requirements, or computing resource configuration of the cloud rendering operation target. The specific settings for dividing the cloud rendering operation target into multiple equal-area regions are to be performed by professionals.

[0052] For example, if the pixel size of the target of the cloud rendering operation is Then, dividing the horizontal and vertical directions into m rows and n columns respectively, we get the total number of... The regions are divided into areas, and the area of ​​each region is: ,in, The area to be divided for cloud rendering operations.

[0053] The image processing tool library is used to convert the segmented region into a grayscale image and obtain the grayscale value of the pixel in the segmented region. Then, the horizontal and vertical convolution kernels in the Sobel operator are used to calculate the gradient magnitude of the pixel in the segmented region.

[0054] The horizontal and vertical convolution kernels in the Sobel operator are as follows:

[0055] The horizontal convolution kernel is: ,in, The kernel is a horizontal convolution kernel;

[0056] The vertical convolution kernel is: ,in, The kernel is a vertical convolution kernel;

[0057] The pixels in the divided region are convolved with the horizontal convolution kernel to obtain the gradient matrix of the pixels in the horizontal direction: ,in, To divide the region, the coordinates are The grayscale value of the pixel at that location. This represents a two-dimensional convolution operation. To divide the area The gradient matrix in the horizontal direction of the pixel;

[0058] Each pixel in the divided region is convolved with a vertical convolution kernel to obtain the gradient matrix of the pixel in the vertical direction: ,in, To divide the area The gradient matrix in the vertical direction of the pixel;

[0059] Calculate the gradient magnitude of each pixel in the divided region: ,in, To divide the area The gradient magnitude of the pixel at that location;

[0060] The average gradient magnitude of all pixels in the divided region is taken as the region resolution of the divided region. The region resolution is used to reflect the clarity of details in the divided region.

[0061] The larger the gradient magnitude, the higher the visual clarity of the divided region and the greater the region resolution.

[0062] After the regions are segmented and processed by the Canny edge detection algorithm, edge maps of the segmented regions are generated. The edge maps of the segmented regions are binary images, in which edge pixels have a value of 1 and non-edge pixels have a value of 0.

[0063] Based on the edge map of the segmented region, the artifact region is constructed by connecting the edge pixels in the edge map. The total number of pixels in all closed artifact regions is counted, and the artifact area is calculated by multiplying it by the preset pixel area. This area is used to reflect the size of the false pixel range caused by rendering anomalies in the segmented region.

[0064] The preset pixel area is used to represent the standard area value corresponding to a single pixel in physical space. It can be converted according to the image resolution set by the cloud rendering platform and the actual rendering size, or it can be predefined as a fixed constant under a unified rendering task, which is set by professionals.

[0065] It should be noted that a cloud rendering platform refers to a cloud computing system that provides remote image rendering services. It receives rendering requests from clients or edge devices and extracts image frame data from the received rendering tasks, serving as the target of the cloud rendering operation to be analyzed in this invention. OpenCV is an open-source computer vision and image processing library that can be used to convert color images to grayscale images. The Sobel operator is an image edge detection tool that applies two... The convolution kernel is used to calculate the gradient magnitude of the pixel, and the region resolution is calculated based on the gradient magnitude of the pixel. The Canny edge detection algorithm is a multi-stage edge detection method. By calling the Canny edge detection algorithm in the OpenCV image processing tool library, the edge map of the divided region is extracted. When constructing the artifact region, the contour retrieval mode of the OpenCV image processing tool library is used to extract the closed artifact contour in the edge map.

[0066] In step S2, the ratio of the artifact area of ​​the divided region to the region resolution is used as the cloud rendering fault index of the divided region.

[0067] The average cloud rendering fault index of each region is used as the fault region location feature of the cloud rendering operation target.

[0068] The larger the artifact area, the lower the regional resolution, indicating that the image content of the divided region is less clear, and the greater the cloud rendering failure index of the divided region.

[0069] The fault location features of the target cloud rendering operation are transmitted to a temporary storage database for storage.

[0070] By calculating the ratio of the artifact area to the region resolution as the cloud rendering fault index of the region, and taking the average of the cloud rendering fault indices of all regions, the fault area location features of the cloud rendering operation target are obtained and cached in a temporary storage database. This achieves efficient management and fast retrieval of fault feature data, improves the accuracy and response speed of cloud rendering service fault location, and helps to detect and handle rendering anomalies in a timely manner.

[0071] It should be noted that the temporary storage database refers to a system that stores the analytical data generated by the cloud rendering operation target during the rapid fault location process of cloud rendering.

[0072] In step S3, a retry time is set for the target cloud rendering operation. After detecting that the initial rendering does not meet the expected rendering quality standard, multiple cloud rendering retries are initiated for the target cloud rendering operation within the retry time. The retry time is determined by multiplication based on the preset maximum number of retries and the standard time quota corresponding to a single rendering task. During this period, a sufficient number of independent rendering operations are performed to obtain stable statistical samples, while avoiding resource waste and redundant delays.

[0073] Cloud rendering retry refers to the process in which, if a target cloud rendering operation fails to meet the preset quality benchmark after the first rendering is completed, multiple independent and repeated rendering operations are performed on it in a scheduled manner within a set retry time. Each cloud rendering retry is executed on an independent computing node and has independent input, execution process and output results, ensuring that each rendering attempt has statistical independence for fault collection.

[0074] After each cloud rendering retry, the quality of the rendering output is evaluated. First, the original pixel matrix of the cloud rendering target is obtained. Then, the distribution data of all pixels of the cloud rendering target is extracted using the image analysis unit. Based on preset valid pixel determination rules, each pixel is judged according to the rules, and pixels that meet the valid pixel determination conditions are recorded as valid pixels. The number of valid pixels of the cloud rendering target is counted, and then the average resolution of the cloud rendering target is calculated, defined as:

[0075] ;

[0076] in, This represents the average resolution after the i-th cloud rendering retry; This indicates the number of regions to be divided in the cloud rendering operation target; This represents the actual resolution of the j-th region in the i-th rendering.

[0077] It should be noted that the image parsing unit is the module responsible for quality detection and parameter extraction of the rendered output image during the cloud rendering retry process. The effective pixel judgment rule is the judgment criterion used to determine whether a certain pixel in the cloud rendering output image has image quality validity. It includes multiple evaluation dimensions and is used to exclude invalid pixels in areas with rendering anomalies, image distortion, or quality degradation.

[0078] Compare the average resolution with the preset resolution benchmark. Comparison: If If the result is negative, the rendering is considered a failure; otherwise, it is considered a valid rendering.

[0079] The resolution baseline value is set based on a historical set of valid rendering samples. The mean and standard deviation of the number of valid pixels for each type of sample are statistically analyzed. An adjustment coefficient is set in combination with the minimum acceptable quality requirements in practical applications. The resolution baseline value is obtained by subtracting the product of the standard deviation and the adjustment coefficient from the mean number of valid pixels, in order to determine whether the rendering result meets the standard.

[0080] Record and statistically analyze all cloud rendering retry results executed within the retry time frame, and sum them up to satisfy the requirements. The number of retries and the total number of rendering attempts completed within the retry time are used to obtain the number of rendering failures and the total number of cloud rendering retries, respectively.

[0081] The ratio of the number of rendering failures to the total number of cloud rendering retries is used as the retry failure frequency of the cloud rendering operation target. The retry failure frequency of the cloud rendering operation target reflects the relative proportion of rendering failures during repeated rendering, and serves as an important parameter for determining whether fault type localization is necessary.

[0082] After each cloud rendering retry is completed, the actual execution time of the rendering task on the cloud computing platform is recorded, that is, the total time from task submission to output result completion. This time is used as the rendering execution time for each cloud rendering retry. The preset rendering time calibration benchmark is retrieved. The rendering time calibration benchmark is the standard rendering time estimated based on task complexity and resource configuration, which is used to measure rendering efficiency and performance deviation. This will not be elaborated here.

[0083] The ratio of the rendering execution time of each cloud rendering retry to the preset rendering time calibration benchmark is calculated to obtain the single rendering execution percentage of each rendering retry. If the single rendering execution percentage is greater than 1, it means that the rendering efficiency is lower than the standard level. If the single rendering execution percentage is less than 1, it means that the rendering efficiency is higher than the standard level.

[0084] The average percentage of single-render executions across all retries is used to obtain the rendering level of the current cloud rendering operation target. The rendering level reflects the degree of deviation of the overall rendering performance of the cloud rendering operation target within the retry cycle. The higher the rendering level, the longer the average rendering time, and the more necessary it is to locate the fault type.

[0085] In step S4, after standardizing the retry failure frequency and rendering level, the fault type localization features are calculated using a weighted fusion method. The specific calculation formula is as follows:

[0086] ;

[0087] in, Locate the characteristics of the fault type. To determine the frequency of retrying faults, For rendering levels, and These are the weighting coefficients.

[0088] It should be noted that the standardization methods include, but are not limited to, standard linear transformation based on interval scaling, Z-Score standardization based on statistics, or normalization based on nonlinear mapping functions. The application methods of standardization will not be elaborated here. The weighting coefficients are set based on historical cloud rendering fault sample data. The correlation between the retry fault frequency and the rendering level in each sample and the final fault type determination result is statistically analyzed. The Pearson correlation coefficient is used to calculate the linear correlation between the two and the labeled fault type, and the weighting coefficients are allocated proportionally according to the magnitude of the correlation coefficient.

[0089] The stored fault area location features are extracted from the temporary storage database and standardized. The difference between the fault type location features and the fault area location features is calculated to obtain the difference value between the two. Difference value Reflecting the relative severity between the two fault location characteristics, the difference value is... Compared with the preset fault judgment threshold The comparison was made based on the following criteria:

[0090] like If the severity of the fault type location feature exceeds that of the fault area location feature, it indicates that the cloud rendering operation target may fail due to faults such as network problems or rendering engine problems, and the fault type location method should be used first.

[0091] like If the severity of the fault area location feature is greater than or equal to the fault type location feature, it indicates that the rendering failure of the cloud rendering operation target is mainly concentrated in certain specific areas, such as areas with low resolution or areas with many artifacts. In this case, the fault area location method is preferred.

[0092] It should be noted that the fault judgment threshold is based on a large number of historical cloud rendering fault cases. The standardized difference distribution between the corresponding fault type location features and the fault area location features is calculated, and the statistical quantile of this difference distribution is selected as the fault judgment threshold to distinguish the severity difference between the two fault location features.

[0093] This step, based on the fault characteristics of the cloud rendering operation target, prioritizes fault type location methods or fault area location methods to accurately locate and handle faults, thereby improving the stability of cloud rendering services and the efficiency of fault diagnosis.

[0094] Finally, it should be noted that in this paper, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations.

[0095] 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 limitation, 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.

[0096] In this document, the singular forms “a,” “an,” and “the” may also include the plural forms unless the context clearly indicates otherwise. It should also be understood that terms such as “comprising / including” or “having” specify the presence of the stated features, integrals, steps, operations, components, parts, or combinations thereof, but do not preclude the possibility of the presence or addition of one or more other features, integrals, steps, operations, components, parts, or combinations thereof. Meanwhile, the term “and / or” as used in this specification includes any and all combinations of the associated listed items.

[0097] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The various embodiments can be combined as needed, and the same or similar parts can be referred to each other.

[0098] The above description of the disclosed embodiments will enable those skilled in the art to make or use various modifications to these embodiments. It will be readily apparent to those skilled in the art that the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for rapid fault location in AI-driven cloud rendering services, characterized in that: Includes the following steps: Step S1: Obtain the cloud rendering operation target, divide the cloud rendering operation target into regions, and collect the regional resolution and artifact area of ​​each region. Step S2: Calculate the cloud rendering fault index of each region by combining the regional resolution and artifact area of ​​each region, evaluate the fault area location characteristics of the cloud rendering operation target based on the cloud rendering fault index of each region, and input it into the temporary storage database. Step S3: Set a retry time, perform multiple cloud rendering retries on the cloud rendering operation target within the retry time, and count the retry failure frequency of the cloud rendering operation target. Detect the rendering execution time of each cloud rendering retry, and calculate the rendering level of the cloud rendering operation target based on the rendering execution time. Step S4: Based on the comprehensive assessment of the retry failure frequency and rendering level, evaluate the failure type location characteristics of the cloud rendering operation target, and combine the failure area location characteristics in the temporary storage database to select the failure area location method or failure type location method to process the cloud rendering operation target.

2. The method for rapid fault location of AI-driven cloud rendering service according to claim 1, characterized in that: In step S1, the cloud rendering operation target is obtained through the cloud rendering platform, and the cloud rendering operation target is divided into a preset number of equal-area division regions; The image processing tool library is used to convert the segmented region into a grayscale image and obtain the grayscale value of the pixel in the segmented region. The gray values ​​of pixels in the divided region are convolved with the horizontal convolution kernel of the Sobel operator to obtain the gradient matrix of the pixel in the horizontal direction. The gray values ​​of pixels in the divided region are convolved with the vertical convolution kernel of the Sobel operator to obtain the gradient matrix of the pixel in the vertical direction.

3. The method for rapid fault location in AI-driven cloud rendering services according to claim 2, characterized in that: In step S1, the gradient magnitude of the pixel is calculated by combining the gradient matrix in the horizontal direction and the gradient matrix in the vertical direction of the pixel. The average gradient magnitude of all pixels in the divided region is used as the region resolution of the divided region. The regions are divided and processed using the Canny edge detection algorithm to generate edge maps of the divided regions.

4. The method for rapid fault location of AI-driven cloud rendering service according to claim 3, characterized in that: In step S1, the edge map of the divided region is a binarized image, where edge pixels have a value of 1 and non-edge pixels have a value of 0. Artifact regions are constructed by connecting edge pixels in the edge map, and the total number of pixels in all closed artifact regions is counted. The artifact area is calculated by multiplying the total number of pixels by the preset pixel area.

5. The method for rapid fault location in AI-driven cloud rendering services according to claim 4, characterized in that: In step S2, the ratio of the artifact area of ​​the divided region to the region resolution is used as the cloud rendering fault index of the divided region. The average cloud rendering fault index of each region is used as the fault region location feature of the cloud rendering operation target. The fault location features of the target cloud rendering operation are transmitted to a temporary storage database for storage.

6. The method for rapid fault location in AI-driven cloud rendering services according to claim 1, characterized in that: In step S3, the cloud rendering operation target is retried multiple times within a preset retry time. After each cloud rendering retry, the resolution of the current cloud rendering target is detected, and the average resolution of the cloud rendering target is calculated, defined as: ； in, This represents the average resolution after the i-th cloud rendering retry; This indicates the number of regions to be divided in the cloud rendering operation target; This represents the actual resolution of the j-th region in the i-th rendering; The average resolution is compared with the preset resolution baseline. If the average resolution is less than the resolution baseline, the rendering is considered to have failed. Count all cloud rendering retries that were judged to have failed to render, and obtain the number of rendering failures. The total number of cloud rendering retries is calculated by counting all cloud rendering retries completed within the retry time.

7. The method for rapid fault location of AI-driven cloud rendering service according to claim 6, characterized in that: In step S3, the ratio between the number of rendering failures and the total number of cloud rendering retries is used as the retry failure frequency of the cloud rendering operation target. Obtain the rendering execution time of each cloud rendering retry, and use the ratio of the rendering execution time of each cloud rendering retry to the preset rendering time calibration benchmark as the single rendering execution percentage of each rendering retry; The average percentage of all single-render executions is used as the rendering level of the cloud rendering operation target.

8. The method for rapid fault location of AI-driven cloud rendering service according to claim 7, characterized in that: In step S4, after standardizing the retry failure frequency and rendering level, the fault type localization features are calculated using a weighted fusion method. Extract the stored fault area location features from the temporary storage database and perform standardization processing. Calculate the difference between the fault type location features and the fault area location features to obtain the difference value.

9. The method for rapid fault location of AI-driven cloud rendering service according to claim 8, characterized in that: In step S4, if the difference value is greater than the preset fault judgment threshold, then the fault type location method is selected for processing. If the difference value is less than or equal to the preset fault judgment threshold, then the fault area location method is selected for processing.

Images