Data collaborative operation method and system of edge computing server

Abstract

The application relates to the technical field of edge computing and video stream processing, and discloses a data cooperative operation method and system of an edge computing server, the method comprising inputting a network quality coefficient, a computing load coefficient and a user experience coefficient into a preset decision model to output a decision influence factor; determining a content value coefficient according to image space information index, generating a final resolution control coefficient according to the content value coefficient and the decision influence factor; and generating a control instruction for adjusting the resolution of a video stream based on the resolution control coefficient. The application adopts a nonlinear resolution mapping model and content value self-adaptive adjustment, preferentially guarantees the picture quality of high-value content when resources are limited, avoids loss of key information, guarantees the viewing experience of a client, and realizes the best balance between system efficiency and user experience.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of edge computing and video stream processing, and particularly relates to a data collaborative operation method and system of an edge computing server. BACKGROUND

[0002] With the development of industrial internet and intelligent manufacturing, edge computing undertakes real-time video processing tasks in scenarios such as machine tool processing and visual detection. However, the existing edge video processing system faces some problems. In actual application, wireless network transmission architecture is mostly used. Such architecture has strong volatility, and the traditional video sending process is mostly carried out at a fixed resolution, which is difficult to adapt to changes in bandwidth and delay, resulting in video lag. In addition, different industrial scenarios have different requirements for delay and smoothness, and the existing method lacks fine-grained protection of user experience.

[0003] Therefore, how to provide an edge computing scheme capable of fusing multi-dimensional parameters and making collaborative decisions to realize adaptive and accurate control of video streams is a technical problem to be solved by the technical scheme of the application. SUMMARY

[0004] The purpose of the embodiment of the application is to provide a data collaborative operation method and system of an edge computing server, aiming to solve the above problems.

[0005] The technical scheme of the application provides a data collaborative operation method of an edge computing server, which comprises the following steps:

[0006] querying real-time available bandwidth and network round-trip delay to determine a network quality coefficient;

[0007] obtaining a video encoding queue depth to determine a computing load coefficient;

[0008] calculating an end-to-end delay budget and a client buffer length to determine a user experience coefficient;

[0009] inputting the network quality coefficient, the computing load coefficient and the user experience coefficient into a preset decision model to output a decision influence factor;

[0010] determining a content value coefficient according to image space information index, and generating a final resolution control coefficient according to the content value coefficient and the decision influence factor;

[0011] generating a control instruction for adjusting the resolution of the video stream based on the resolution control coefficient.

[0012] Further, the step of querying real-time available bandwidth and network round-trip delay to determine a network quality coefficient comprises:

[0013] The ratio of real-time available bandwidth to the maximum bandwidth supported by the system is processed, and the upper limit is limited to 1 using the min function to obtain the real-time available bandwidth index.

[0014] The network round-trip time is calculated as a ratio to the maximum allowable round-trip time. After limiting the upper limit to 1 using the min function, the network round-trip time index is obtained.

[0015] Import the real-time available bandwidth index and the network round-trip latency index into the formula. Obtain network quality coefficient ,in, This represents the real-time available bandwidth index. This refers to the network round-trip latency index. and All are weighting coefficients.

[0016] Furthermore, the steps of obtaining the video encoding queue depth and determining the calculation load coefficient include:

[0017] The ratio of the current video encoding queue depth to the maximum allowed queue depth is calculated.

[0018] The difference between 1 and the ratio is calculated, and the difference is then capped at 0 using the max function to obtain the calculated load factor. .

[0019] Furthermore, the step of calculating the end-to-end latency budget and client buffer length to determine the user experience coefficient includes:

[0020] The difference between the total delay budget and the current end-to-end delay budget is compared with the total delay budget, and the delay satisfaction factor is obtained by using the max function to limit the lower limit to 0.

[0021] The ratio of the current client buffer length to the target buffer length is calculated, and the upper limit is set to 1 using the min function to obtain the buffer satisfaction factor.

[0022] Import the delayed satisfaction factor and the buffer satisfaction factor into the formula. Obtain user experience coefficient ,in, For delayed satisfaction factor, For the buffer to satisfy the factor, and All are weighting coefficients.

[0023] Furthermore, the decision model is as follows:

[0024] in, As a factor influencing decision-making, For network quality coefficients, To calculate the load factor, For user experience coefficient, , and All are weighting coefficients.

[0025] Furthermore, the step of determining the content value coefficient based on the image spatial information index, and generating the final resolution control coefficient based on the content value coefficient and the decision influence factor, includes:

[0026] The difference between the current image spatial information index and the minimum spatial information index is compared with the difference between the maximum spatial information index and the minimum spatial information index.

[0027] First, the lower limit is set to 0 using the max function, and then the upper limit is set to 1 using the min function to obtain the content value coefficient. ;

[0028] Content value coefficient and decision-making influencing factors Import Formula Obtain resolution control coefficients .

[0029] Furthermore, the step of generating control instructions to adjust the resolution of the video stream based on the resolution control coefficient includes:

[0030] Input the resolution control coefficients into the preset control execution model to determine the resolution level;

[0031] The obtained resolution level is used as the target level, and control commands are generated in combination with the current level.

[0032] The control execution model is as follows: ,in, This represents the floor function. For the target resolution level, The maximum number of resolution levels supported by the system. This is the resolution control factor. It is a natural constant.

[0033] The present invention also provides a data collaborative computing system for edge computing servers, the system comprising:

[0034] The network quality assessment module is used to query real-time available bandwidth and network round-trip latency to determine the network quality coefficient.

[0035] The network load determination module is used to obtain the video encoding queue depth and determine the load coefficient.

[0036] The user experience quantification module is used to calculate the end-to-end latency budget and client buffer length to determine the user experience coefficient.

[0037] The decision output module is used to input network quality coefficient, computational load coefficient and user experience coefficient into a preset decision model and output decision influencing factors.

[0038] The parameter integration module is used to determine the content value coefficient based on the image spatial information index, and to generate the final resolution control coefficient based on the content value coefficient and the decision influence factor.

[0039] The control execution module is used to generate control commands to adjust the resolution of the video stream based on the resolution control coefficient.

[0040] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0041] This invention employs a nonlinear resolution mapping model and adaptive content value adjustment. When resources are limited, it prioritizes the image quality of high-value content, avoiding the loss of key information. Through dynamic weight allocation and normalization processing, it achieves adaptive optimization under different industrial scenarios, significantly improving the smoothness and real-time performance of the video stream, ensuring the viewing experience for the client, and achieving the best balance between system efficiency and user experience. Attached Figure Description

[0042] Figure 1 This invention provides a flowchart of a data collaborative computing method for an edge computing server.

[0043] Figure 2 This invention provides a structural block diagram of a data collaborative computing system for an edge computing server. Detailed Implementation

[0044] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0045] Figure 1 This is a flowchart illustrating a data collaborative computing method for an edge computing server provided by the present invention. The technical solution of the present invention provides a data collaborative computing method for an edge computing server, the method comprising:

[0046] Step S100: Query the real-time available bandwidth and network round-trip latency to determine the network quality coefficient;

[0047] Step S200: Obtain the video encoding queue depth and determine the calculation load factor;

[0048] Step S300: Calculate the end-to-end latency budget and client buffer length to determine the user experience coefficient;

[0049] Step S400: Input the network quality coefficient, computational load coefficient, and user experience coefficient into the preset decision model, and output the decision influencing factor;

[0050] Step S500: Determine the content value coefficient based on the image spatial information index, and generate the final resolution control coefficient based on the content value coefficient and the decision influence factor;

[0051] Step S600: Generate control instructions to adjust the resolution of the video stream based on the resolution control coefficient.

[0052] In industrial internet and smart manufacturing scenarios, edge computing needs to handle real-time video tasks such as machine tool processing and visual inspection. The data volume of video tasks is very large, and the edge computing architecture requires a continuous transmission process. Once network fluctuations occur, video will experience stuttering or delay. When multiple video streams are processed concurrently, the computing resources of the edge server are limited, and processing queues are prone to backlog. The above solution introduces real-time available bandwidth, round-trip latency, and encoding queue depth as basic parameters. It combines end-to-end latency budget and client buffer state to quantify user experience, evaluates content importance through image spatial information index, and finally combines network, computing, and user experience parameters into a decision factor, which is jointly adjusted with the content value coefficient to achieve dynamic resolution optimization. Based on dynamically determining the clarity, it ensures the stability of video transmission.

[0053] Furthermore, the above solution utilizes components for real-time measurement of network transmission capacity and latency status, specifically employing bandwidth probing algorithms and round-trip delay calculation models, to reflect the current network transmission stability. It periodically collects available bandwidth and round-trip delay data, and generates a network quality coefficient through normalization and weighted fusion to accurately reflect changes in network transmission capacity. It also monitors the video encoding task queue depth in real time, quantifying server processing pressure by calculating the ratio of the current queue length to the maximum allowed value. Finally, it incorporates an evaluation unit for latency constraints and playback continuity, specifically achieved by calculating latency deviation and buffer fill ratio. For example, it synchronously acquires end-to-end latency deviation and client buffer fill status, combining latency budget and target buffer length to generate a user experience coefficient, ensuring a smooth user interaction experience.

[0054] After the above processing, multi-dimensional parameters can be obtained. A weighted fusion algorithm is used to generate an adjustment strategy. Specifically, the parameters are linearly combined by preset weights to output decision-making factors that reflect the overall state of the system. On this basis, texture complexity can also be calculated by spatial information index to identify high-value image regions. For example, the spatial information index of video frames can be extracted and the content value coefficient can be generated by dynamic range normalization to identify the priority of high texture complexity images.

[0055] After obtaining the above data, it is necessary to apply it comprehensively.

[0056] The adaptive adjustment module refers to the control unit that dynamically balances system decisions and content value. Specifically, it can be implemented through a product coupling algorithm to generate resolution adjustment instructions.

[0057] The resolution control execution module is the component that performs resolution adjustments. Specifically, it can be implemented by mapping discrete resolution levels using a mathematical model to achieve smooth image quality switching.

[0058] Specifically, the decision-making influencing factor is multiplied by the content value coefficient to generate the final resolution control coefficient, ensuring that high-value images maintain their resolution within the limits of system resources. Then, the optimal resolution level is selected through a nonlinear mapping model to achieve smooth adjustment.

[0059] Regarding step S100, the step of querying real-time available bandwidth and network round-trip latency to determine the network quality coefficient includes:

[0060] The ratio of real-time available bandwidth to the maximum bandwidth supported by the system is processed, and the upper limit is limited to 1 using the min function to obtain the real-time available bandwidth index.

[0061] The network round-trip time is calculated as a ratio to the maximum allowable round-trip time. After limiting the upper limit to 1 using the min function, the network round-trip time index is obtained.

[0062] Import the real-time available bandwidth index and the network round-trip latency index into the formula. Obtain network quality coefficient .

[0063] in, The value range is 0-1. This represents the real-time available bandwidth index. This refers to the network round-trip latency index. and All are weighting coefficients, and , and The values ​​range from 0 to 1. Real-time available bandwidth refers to the effective bandwidth available for data transmission in the current network link. This can be obtained in real-time through network probe messages or protocol stack interfaces, reflecting the current network transmission capacity. The maximum bandwidth supported by the system refers to the theoretical maximum transmission rate of the link between the edge computing server and the client. This can be obtained, for example, through network interface configuration parameters, and is used to normalize the real-time bandwidth. Network round-trip latency refers to the time interval between a data packet's arrival at the sender and return point. This can be measured using the ICMP protocol or TCP timestamps, and is used to assess network latency. Maximum allowable round-trip latency refers to the maximum network latency threshold that the system can tolerate. For example, it is set according to the real-time requirements of video streaming, and is used to constrain the reasonable range of latency indicators. The real-time available bandwidth index is a normalized bandwidth utilization indicator, which is used to avoid exceeding the limit through limiting processing. The network round-trip latency index is a normalized inverse latency indicator, obtained through... The transformation maps increased latency to an exponential decrease. Weighting coefficients. and This refers to the dynamic adjustment parameters of bandwidth and latency indicators, which can be set, for example, in industrial testing scenarios. =0.7、 =0.3 prioritizes bandwidth resources and can be set in remote control scenarios. =0.4、 =0.6 to emphasize low latency.

[0064] The solution described above maps bandwidth and latency to the 0-1 range using double normalization, eliminating dimensional differences. For bandwidth, a positive ratio is used to directly reflect resource utilization; for example, when real-time bandwidth reaches 80% of maximum bandwidth... Take 0.8. For the delay index, use reverse processing and add 1-operation to make... As the actual latency increases, The corresponding decrease occurs, for example, when the actual latency reaches 90% of the maximum allowable value. =0.9, at this time =0.1. A comprehensive network quality assessment value is formed by linearly weighting and fusing the two indices, where... and The configuration can be dynamically adjusted according to the application scenario, such as improving bandwidth in bandwidth-sensitive video surveillance scenarios. Weighting, improving performance in latency-sensitive remote operation scenarios. Weighting. This evaluation mechanism can simultaneously capture bandwidth resource sufficiency and transmission real-time performance, providing a multi-dimensional decision-making basis for resolution adjustment; specifically, by introducing a two-way index fusion mechanism, it overcomes the one-sidedness of single-parameter evaluation.

[0065] It is worth mentioning that existing technologies mostly use fixed weight allocation, while this solution supports dynamic adjustment based on business needs. and For example, in machine tool processing scenarios, priority is given to ensuring low latency, and in visual inspection scenarios, priority is given to ensuring high bandwidth, thereby achieving scenario-adaptive optimization of network quality assessment.

[0066] In practical applications, the resolution level is automatically reduced to compress the amount of data when bandwidth is sufficient but latency is too high, or the encoding parameters are dynamically adjusted when latency is controllable but bandwidth is limited. This evaluation method provides a reliable network status input for video stream resolution control, avoids video stuttering or resource waste caused by misjudgment of network quality, and ensures the real-time performance and stability of video processing tasks in industrial scenarios.

[0067] Step S200, the step of obtaining the video encoding queue depth and determining the calculation load coefficient includes:

[0068] The ratio of the current video encoding queue depth to the maximum allowed queue depth is calculated.

[0069] The difference between 1 and the ratio is calculated, and the difference is then capped at 0 using the max function to obtain the calculated load factor. ; The value range is 0-1.

[0070] The video encoding queue depth refers to the number of video frames currently awaiting processing backlogged in the encoding queue. This can be achieved by real-time statistics of the number of frames with incomplete encoding tasks in the queue, representing the real-time computational load pressure on the edge server. The maximum allowed queue depth refers to the system-set upper limit threshold for queue capacity, which can be implemented using a preset fixed value or dynamically adjusted based on server hardware performance, to prevent task backlog caused by queue overflow. (Calculate the load factor.) This refers to the normalized load pressure index, specifically achieved by inversely transforming the ratio of the current queue depth to the maximum allowable depth and then applying a limit. When the queue depth approaches the system's capacity limit, Approaching 0 triggers the resolution adjustment mechanism.

[0071] Furthermore, when the edge server processes multiple video streams, the number of frames to be processed in the video encoding queue is monitored in real time. The current queue depth is divided by the preset maximum allowed queue depth to obtain the original ratio reflecting the current load pressure. By subtracting this ratio from 1 and applying a lower limit, the load factor is calculated when the queue depth reaches 90% of the maximum allowed value. The value will drop below 0.1; at this point, the system automatically triggers a resolution adjustment mechanism, reducing the amount of encoding computation by lowering the video stream resolution, thereby avoiding processing delays caused by queue overflow. When the queue depth returns to a normal level, Once the value rises above 0.5, the system gradually restores the original resolution settings. This dynamic adjustment process ensures real-time video processing while maintaining smooth playback on the client side.

[0072] In practical applications, the above technical solution can sense the computing load status of the edge server in real time in scenarios of concurrent processing of multiple video streams. When the video encoding queue approaches its capacity limit, it can automatically trigger resolution adjustment to prevent processing delays caused by task backlog. At the same time, by dynamically balancing computing resource consumption and video quality requirements, it can maintain the continuity and smoothness of video playback on the client side while ensuring the stable operation of the system.

[0073] Step S300, the step of calculating the end-to-end latency budget and client buffer length to determine the user experience coefficient includes:

[0074] The difference between the total delay budget and the current end-to-end delay budget is compared with the total delay budget, and the delay satisfaction factor is obtained by using the max function to limit the lower limit to 0.

[0075] The ratio of the current client buffer length to the target buffer length is calculated, and the upper limit is set to 1 using the min function to obtain the buffer satisfaction factor.

[0076] Import the delayed satisfaction factor and the buffer satisfaction factor into the formula. Obtain user experience coefficient ,in, For delayed satisfaction factor, For the buffer to satisfy the factor, and All are weighting coefficients.

[0077] in, The value range is 0-1. The delay satisfaction factor is the proportion of the current remaining delay budget to the total delay budget. Specifically, it can be calculated by dividing the total delay budget by (total delay budget - current delay budget) and applying the max function to limit the lower bound to 0. It is used to quantify the degree of satisfaction of delay constraints. When the actual delay is close to the budget threshold, the factor approaches 0, triggering the system to prioritize reducing the resolution to control the delay. The buffer satisfaction factor is the ratio of the client buffer fullness to the target value. Specifically, it can be achieved by dividing the current buffer length by the target buffer length and applying the min function to limit the upper limit to 1. It is used to reflect the risk state of playback smoothness. When the buffer is lower than the target value, the factor decreases linearly, prompting the system to increase the resolution to maintain smooth playback. and All are weighting coefficients, and , and The values ​​range from 0 to 1, and the weighting coefficients are... and This refers to the contribution ratio of the delay factor and the buffer factor in the comprehensive evaluation. Specifically, it can be achieved by using a preset fixed value or a dynamic adjustment strategy to adjust the priority of the two according to the needs of different industrial scenarios.

[0078] The application process of the above scheme is as follows: First, calculate the ratio of the remaining delay budget to the total delay budget to generate a delay satisfaction factor. When the actual delay approaches the budget threshold, this factor tends to be close to 0, prompting the system to prioritize delay-sensitive scenarios. Then, calculate the ratio of the client buffer length to the target value to generate a buffer satisfaction factor. When the buffer length is lower than the target value, this factor decreases linearly, triggering a resolution adjustment mechanism. Finally, the two factors are weighted and summed according to preset weights to generate a comprehensive user experience coefficient. The weight coefficients can be configured differently according to the needs of different industrial scenarios. For example, in a machine tool processing scenario requiring real-time control, the following applies: Set to 0.7 Set it to 0.3 to make the delay satisfaction factor dominant; in visual inspection scenarios that require continuous and smooth playback, set it to... Set to 0.3. Setting it to 0.7 ensures the buffer factor dominates. The delay factor and buffer factor are limited using the max and min functions respectively, preventing numerical anomalies caused by sudden network fluctuations.

[0079] In practical applications, the above solution can achieve refined protection of user experience based on the dynamically changing end-to-end latency and client buffer status in industrial scenarios. When the latency is close to the budget threshold, the resolution control coefficient is automatically reduced to prioritize avoiding video stuttering caused by latency exceeding the limit; when the buffer is below the target value, the resolution control coefficient is increased to prioritize maintaining playback smoothness.

[0080] Step S400, the decision model is:

[0081] in, As a factor influencing decision-making, For network quality coefficients, To calculate the load factor, For user experience coefficient, , and All are weighting coefficients.

[0082] The network quality index is a quantitative indicator that reflects the current network transmission capacity. It can be calculated by combining real-time bandwidth and round-trip latency, and is used to characterize the quality of the network transmission environment. To calculate the load factor, which is a quantitative indicator reflecting the processing pressure of edge servers, it can be calculated by comparing the video encoding queue depth with the system's carrying capacity, and is used to characterize the degree of strain on computing resources. The user experience coefficient is a quantitative indicator that reflects the perceived quality of service on the user end. Specifically, it can be calculated by combining the delay budget satisfaction and the buffer state, and is used to characterize the terminal performance of service quality. , and All are weighting coefficients, and , , and The values ​​range from 0 to 1. The weight coefficient is a parameter used to adjust the importance of different factors in decision-making. Specifically, it can be assigned values ​​through preset strategies or dynamic algorithms to achieve priority adjustment in different scenarios.

[0083] Furthermore, the aforementioned decision-making model is an integrated parameter processing function. Through normalization, it converts the three heterogeneous parameters—network, computation, and user—into coefficients with uniform dimensions. A weighted summation method is used to generate a comprehensive decision factor. When network quality deteriorates, the first weight coefficient can be increased to favor network optimization; when computational resources are strained, the second weight coefficient can be increased to strengthen load balancing; and if user experience declines, the third weight coefficient is increased to prioritize service quality. Each weight coefficient can be dynamically adjusted during system operation. For example, in industrial vision inspection scenarios, when the detection algorithm requires high-resolution input, a higher third weight coefficient can be pre-set to ensure user experience. When applied to real-world scenarios, this allows the resolution adjustment strategy to be dynamically optimized based on changes in network conditions, server load, and user needs, avoiding video stuttering or resource waste caused by single-factor decisions and improving the adaptability and stability of video streaming transmission in complex industrial environments.

[0084] Step S500, the step of determining the content value coefficient based on the image spatial information index, and generating the final resolution control coefficient based on the content value coefficient and the decision influence factor, includes:

[0085] The difference between the current image spatial information index and the minimum spatial information index is compared with the difference between the maximum spatial information index and the minimum spatial information index.

[0086] First, the lower limit is set to 0 using the max function, and then the upper limit is set to 1 using the min function to obtain the content value coefficient. ;

[0087] Content value coefficient and decision-making influencing factors Import Formula Obtain resolution control coefficients .

[0088] in, The image spatial information index, ranging from 0 to 1, is a quantitative indicator reflecting the complexity of image details. It can be implemented using the contrast parameter or edge density parameter of the gray-level co-occurrence matrix, and is used to characterize the effective information contained in a video frame. The maximum and minimum spatial information indices are statistical extreme values ​​dynamically recorded during system operation. These can be maintained using a sliding window mechanism to construct a dynamic normalization benchmark. The `max` function's lower limit of 0 forces negative values ​​in the calculation result to zero, which can be implemented using conditional statements to eliminate evaluation errors caused by low-complexity images. The `min` function's upper limit of 1 constrains results exceeding a threshold within a unit interval, which can be implemented using numerical truncation operations to maintain consistency in the evaluation scale across different resolution levels.

[0089] Furthermore, regarding the calculation principle of the above process, firstly, the difference between the current image spatial information index and the historical minimum value is calculated, and then the ratio of this difference to the difference range of the historical maximum and minimum values ​​is processed to map the absolute complexity to a relative proportion space. If the current value is lower than the historical minimum value, the max function forces the difference ratio to zero to avoid negative evaluation results from low-complexity content. Then, the min function constrains the normalization result to the 0-1 range, forming a standardized coefficient that is independent of the resolution level. This double limiting process ensures that high-complexity images obtain coefficient values ​​close to 1, and low-complexity images obtain coefficient values ​​close to 0, while eliminating the evaluation bias caused by the difference in the absolute dimensions of the spatial information index under different scenarios.

[0090] In addition, regarding the generation process of the resolution control coefficients, The value range is 0-1. This formula indicates that resolution control is influenced by both system decisions and content value. When content value is high, the system tends to maintain resolution, but it is constrained by decision factors. The content value coefficient is a quantitative indicator that dynamically reflects the value of video content based on the image spatial information index. Specifically, it can be implemented using difference ratio processing combined with a limiting function, prioritizing the resolution quality of high-value content. The decision influence factor is a system state constraint indicator generated by integrating network quality, computational load, and user experience parameters. Specifically, it can be implemented using a weighted summation model, representing the global constraint strength of the current system on resolution adjustment. The resolution control coefficient is a dynamic adjustment parameter generated through multiplication operations, specifically implemented using normalized numerical multiplication operations, used to balance the dual needs of content value and system resource status.

[0091] Furthermore, when video content has high spatial complexity, the content value coefficient approaches 1. If the system's decision influence factor is high at this point, the resolution control coefficient will approach 1, triggering operations to maintain or increase the resolution. When network quality deteriorates or computational load exceeds limits, the decision influence factor approaches 0. Even if the content value coefficient is high, the resolution control coefficient will still decrease, forcibly implementing a resolution downgrade strategy. This mechanism achieves dual constraints through a mathematical product relationship: the content value coefficient provides a positive gain for resolution control, while the decision influence factor, as a global constraint coefficient, dynamically limits the range of resolution adjustments.

[0092] Step S600, the step of generating control instructions to adjust the resolution of the video stream based on the resolution control coefficient includes:

[0093] Input the resolution control coefficients into the preset control execution model to determine the resolution level;

[0094] The obtained resolution level is used as the target level, and control commands are generated in combination with the current level.

[0095] The control execution model is as follows: ,in, This represents the floor function. For the target resolution level, The maximum number of resolution levels supported by the system. This is the resolution control factor. It is a natural constant.

[0096] The above content is essentially an instruction output process, the core of which lies in the control execution model, in which... This represents the floor function, which removes the decimal part of the calculated result and keeps only the integer part. Specifically, it can be implemented using the floor function in the mathematical operation library, and is used to map continuous values ​​to discrete resolution levels. For the target resolution level, an integer. The target resolution level L refers to the final output video resolution level of the system. For example, when N=5, L=3 corresponds to 720p resolution. The maximum number of resolution levels supported by the system. positive integers (e.g.) (Corresponding to 240p, 480p, 720p, 1080p and 4K). The resolution control coefficient (RC) is an adjustment parameter calculated by comprehensively considering network quality, computing load, user experience, and content value. Its value ranges from 0 to 1 and is used to reflect the intensity of the system's demand for resolution adjustment. ε is the natural constant, which refers to the base of the exponential function in mathematics. It has a value of 2.71828 and is used to construct nonlinear mapping functions. ,parameter This refers to the adjustment factor that controls the shape of the curve, which can be specifically calibrated experimentally. The mapping relationship approaches linear as it approaches 0. Increase height The rate of change of resolution in the value region is accelerated.

[0097] Furthermore, the model constructs a normalized nonlinear mapping using an exponential function to convert the resolution control coefficients into the target resolution level. When When =0, The corresponding lowest resolution level; when When =1, This corresponds to the highest resolution level. The parameter k controls the curve shape: when... When = 0.1, the mapping approximates a linear distribution; when When =5, The value rapidly approaches 1 in the 0.6-1 range, making it easier to maintain high resolution for high-value content. Through N-fold expansion and +0.5 rounding, the continuous calculation results are converted into integer resolution levels. For example, when N=5 and the calculated value is 3.4, L=3 is obtained after processing. This model supports flexible configuration of the number of resolution levels. For example, in industrial inspection scenarios, N=3 can be set to correspond to low, medium, and high resolutions. In practical applications, this model, integrating multi-dimensional parameters, can dynamically adapt to network fluctuations and computational load, achieving intelligent adjustment of resolution levels. For instance, when network bandwidth suddenly drops, the model automatically reduces the resolution level of low-value video streams, prioritizing the transmission quality of critical inspection images. When the load on edge servers increases, the mapping curve shape is adjusted through parameter k to quickly reduce the processing pressure on non-critical streams. The discrete resolution level settings match the standardized video specification requirements in industrial scenarios, avoiding compatibility issues caused by traditional continuous adjustment. The configurable nature of parameter k allows the system to optimize adjustment strategies according to different industrial scenarios. For example, a larger k value is used in precision machining inspection to enhance sensitivity in high RC regions, while a smaller k value is used in ordinary monitoring scenarios to maintain a smooth adjustment.

[0098] Figure 2 This invention provides a structural block diagram of a data collaborative computing system for an edge computing server. In a preferred embodiment of the invention, the system 10 further includes:

[0099] Network quality assessment module 11 is used to query real-time available bandwidth and network round-trip latency to determine the network quality coefficient;

[0100] The network load determination module 12 is used to obtain the video encoding queue depth and determine the calculated load coefficient.

[0101] The user experience quantification module 13 is used to calculate the end-to-end latency budget and the client buffer length to determine the user experience coefficient.

[0102] The decision output module 14 is used to input the network quality coefficient, computational load coefficient and user experience coefficient into the preset decision model and output the decision influencing factor.

[0103] The parameter integration module 15 is used to determine the content value coefficient based on the image spatial information index, and to generate the final resolution control coefficient based on the content value coefficient and the decision influence factor.

[0104] The control execution module 16 is used to generate control instructions to adjust the resolution of the video stream based on the resolution control coefficient.

[0105] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A data collaborative computing method for an edge computing server, characterized in that, The method includes: Query real-time available bandwidth and network round-trip latency to determine the network quality coefficient; Obtain the video encoding queue depth to determine the calculation load factor; Calculate the end-to-end latency budget and client buffer length to determine the user experience coefficients; Input the network quality coefficient, computational load coefficient, and user experience coefficient into the preset decision model, and output the decision influencing factors; The content value coefficient is determined based on the image spatial information index, and the final resolution control coefficient is generated based on the content value coefficient and the decision influence factor. Control commands to adjust the resolution of the video stream are generated based on the resolution control coefficient; The steps for querying real-time available bandwidth and network round-trip latency to determine the network quality coefficient include: The ratio of real-time available bandwidth to the maximum bandwidth supported by the system is processed, and the upper limit is limited to 1 using the min function to obtain the real-time available bandwidth index. The network round-trip time is calculated as a ratio to the maximum allowable round-trip time. After limiting the upper limit to 1 using the min function, the network round-trip time index is obtained. Import the real-time available bandwidth index and the network round-trip latency index into the formula. Obtain network quality coefficient ,in, This represents the real-time available bandwidth index. This refers to the network round-trip latency index. and All are weighting coefficients; The decision model is as follows: in, As a factor influencing decision-making, For network quality coefficients, To calculate the load factor, For user experience coefficient, , and All are weighting coefficients; Obtain user experience coefficient ,in, For delayed satisfaction factor, For the buffer to satisfy the factor, and All are weighting coefficients; The steps of determining the content value coefficient based on the image spatial information index, and generating the final resolution control coefficient based on the content value coefficient and the decision influence factor, include: The difference between the current image spatial information index and the minimum spatial information index is compared with the difference between the maximum spatial information index and the minimum spatial information index. First, the lower limit is set to 0 using the max function, and then the upper limit is set to 1 using the min function to obtain the content value coefficient. ; Content value coefficient and decision-making influencing factors Import Formula Obtain resolution control coefficients ; The step of generating control instructions to adjust the resolution of the video stream based on the resolution control coefficient includes: Input the resolution control coefficients into the preset control execution model to determine the resolution level; The obtained resolution level is used as the target level, and control commands are generated in combination with the current level. The control execution model is as follows: ,in, This represents the floor function. For the target resolution level, The maximum number of resolution levels supported by the system. This is the resolution control factor. It is a natural constant; ,parameter It refers to the adjustment factor that controls the shape of the curve.

2. The data collaborative computing method for edge computing servers according to claim 1, characterized in that, The steps of obtaining the video encoding queue depth and determining the load coefficient include: The ratio of the current video encoding queue depth to the maximum allowed queue depth is calculated. The difference between 1 and the ratio is calculated, and the difference is then capped at 0 using the max function to obtain the calculated load factor. .

3. The data collaborative computing method for edge computing servers according to claim 2, characterized in that, The steps for calculating the end-to-end latency budget and client buffer length to determine the user experience coefficients include: The difference between the total delay budget and the current end-to-end delay budget is compared with the total delay budget, and the delay satisfaction factor is obtained by using the max function to limit the lower limit to 0. The ratio of the current client buffer length to the target buffer length is calculated, and the upper limit is set to 1 using the min function to obtain the buffer satisfaction factor. Import the delayed satisfaction factor and the buffer satisfaction factor into the formula. Obtain user experience coefficient ,in, For delayed satisfaction factor, For the buffer to satisfy the factor, and All are weighting coefficients.

4. A data collaborative computing system for an edge computing server, characterized in that, The system is used to implement the data collaborative computing method for edge computing servers as described in any one of claims 1-3, and the system includes: The network quality assessment module is used to query real-time available bandwidth and network round-trip latency to determine the network quality coefficient. The network load determination module is used to obtain the video encoding queue depth and determine the load coefficient. The user experience quantification module is used to calculate the end-to-end latency budget and client buffer length to determine the user experience coefficient. The decision output module is used to input network quality coefficient, computational load coefficient and user experience coefficient into a preset decision model and output decision influencing factors. The parameter integration module is used to determine the content value coefficient based on the image spatial information index, and to generate the final resolution control coefficient based on the content value coefficient and the decision influence factor. The control execution module is used to generate control commands to adjust the resolution of the video stream based on the resolution control coefficient.

Images