Distributed intelligent video image scheduling system

Abstract

The invention relates to the technical field of image scheduling, in particular to a distributed intelligent video image scheduling system. The system comprises an acquisition processing unit which is used for collecting video image data by using a camera and preprocessing the video image data; the architecture design unit is used for deploying each node in the network; the scheduling strategy generation unit is used for analyzing video requirements in real time and distributing and managing video resources in combination with a distribution priority algorithm; and the decision execution unit is used for receiving the instruction of the scheduling strategy generation unit and executing the distribution operation of the video image data. In the distributed intelligent video image scheduling system, the network condition and the processing capability of the receiver are considered, and the loss and retransmission of the data packet are avoided, so that the data transmission efficiency and reliability are improved, the data flow can be more accurately controlled, the effective utilization of network resources is ensured, and the user experience is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of image scheduling, and in particular to a distributed intelligent video image scheduling system. Background Art

[0002] With the widespread deployment of video sources such as surveillance cameras, drones, social media, and live broadcast platforms, video data is growing exponentially. These data are not only large in volume, but also diverse in type, containing rich visual information, which requires powerful computing power to process and analyze. Large-scale video data processing requires huge computing and storage resources. How to achieve efficient resource utilization and cost control while ensuring performance is one of the important considerations in system design. Distributed architecture can achieve dynamic allocation and optimized use of resources through resource pooling and load balancing. In distributed systems, data transmission delay is an important factor affecting task execution time. In existing technologies, the selection of data transmission paths is not fully considered, resulting in unnecessary delays. Therefore, a distributed intelligent video image scheduling system is designed. Summary of the invention

[0003] The purpose of the present invention is to provide a distributed intelligent video image scheduling system to solve the problem raised in the above background technology that the selection of data transmission path is not fully considered, resulting in unnecessary delays.

[0004] To achieve the above object, the present invention provides a distributed intelligent video image scheduling system, comprising: Acquisition and processing unit: used to collect video image data using a camera, convert video image data in different formats into a unified format, and pre-process the video image data; Architecture design unit: used to deploy each node in the network and design redundancy and recovery strategies to switch to a backup node when a node fails; Among them, the architecture design unit includes demand analysis module, node deployment module, network construction module, load balancing module and fault tolerance mechanism module; Scheduling strategy generation unit: used to analyze video requirements in real time, adjust the priority and flow of video image data processing, and allocate and manage video resources in combination with the distribution priority algorithm; Wherein, the scheduling strategy generation unit includes a scheduling module, a resource allocation module and a data flow control module; Decision execution unit: used to receive instructions from the scheduling strategy generation unit and execute the allocation operation of video image data, and at the same time, feed back the execution result to the scheduling strategy generation unit.

[0005] As a further improvement of the technical solution, in the architecture design unit: The requirements analysis module is used to clarify system goals, define key performance indicators, create a system framework, and describe the functions of each node and their relationships; The node deployment module is used to deploy video resource tasks according to demand and realize the allocation of various video resource tasks on different nodes; The network construction module is used to determine the communication protocol of video data and plan the network topology; The load balancing module is used to select the load balancing algorithm, adjust the strategy according to the system status, monitor the node health, and redistribute the load when a node failure is detected; The fault-tolerance mechanism module is used for fault detection mechanism, identifying faulty nodes and isolating them from the system, and formulating fault recovery procedures.

[0006] As a further improvement of the present technical solution, the network construction module determines the transmission path and topology according to the specifications and performance of the video data collected by the acquisition and processing unit and the bandwidth, delay, throughput and reliability of the network transmission; and determines the communication protocol for video data transmission according to the format and ability of the acquisition and processing unit to process the video data.

[0007] As a further improvement of the present technical solution, the fault detection mechanism in the fault-tolerant mechanism module includes active detection and passive detection.

[0008] As a further improvement of the technical solution, the node deployment module is combined with the fault tolerance mechanism module, and when a system failure occurs, services can be restored in the correct order. The specific method is as follows: S1. The fault-tolerant mechanism module continuously monitors the health status of all nodes and detects anomalies in a timely manner. When a node detects a fault, it sends a fault report to the fault-tolerant mechanism module and records and analyzes the fault information. Once the fault is confirmed, the fault-tolerant mechanism module immediately isolates the node from the system and formulates and executes a fault recovery process. S2, the node deployment module migrates the services on the faulty node to the healthy node according to the current load and resource allocation of the system, and determines the correct order of task recovery based on the dependencies between video resource tasks to avoid secondary failures caused by improper recovery order; S3. After the service is restored, the fault tolerance mechanism module and the node deployment module jointly check the system status to ensure that all services are running normally.

[0009] As a further improvement of the technical solution, in S2, the specific method for determining the correct order of task recovery according to the dependency relationship between video resource tasks is as follows: S21. Compute node health score: ; in, For Node Final rating of For Node Utilization rate; For Node The weight of utilization rate; For Node failure rate; For Node The weight of the failure rate; For Node availability; For Node The weight of availability; S22. Calculate the dependencies between video resource tasks and generate a task dependency graph: ; in, is a binary variable, Indicates the task Depends on the task ; Indicates the task Not task dependent ; S23, topologically sorting the task dependency graph to determine the execution order list of the tasks; S24, assigning tasks based on the topologically sorted task list, taking into account the health scores of the nodes; S25. After determining the execution order of the tasks and the assigned nodes, implement the task recovery plan.

[0010] As a further improvement of the technical solution, in the scheduling strategy generation unit: The scheduling module is used to adjust the execution order of video resource tasks according to the video image data processing information; The resource allocation module is used to balance the requirements of different video resource tasks and reasonably allocate system resources in the case of multiple video resource tasks by using the task distribution algorithm; The data flow control module is used to adjust the sending rate during the video data transmission process to ensure that the sending rate matches the processing capacity of the node receiver.

[0011] As a further improvement of the technical solution, the scheduling module is based on priority scheduling, specifically as follows: ; in, is the new priority, is the old priority, is the priority adjustment amount.

[0012] As a further improvement of the technical solution, the task distribution algorithm balances the requirements of different video resource tasks and reasonably allocates system resources, as follows: S31. For all tasks to be executed According to its execution time Sort by size; S32. For each task , find the most suitable node to execute; ; in, For the task At the node The fitness score on ; For Node CPU utilization; For Node The weight of CPU utilization; For Node Memory utilization; For Node The weight of memory utilization; For the task From source node to target node Communication delays; For the task From source node to target node The weight of the communication delay; S33, Selection The node corresponding to the maximum value in As If multiple nodes have the same highest score, the node with the lowest load is selected. S34. Once the task Assigned to the node ,renew The state of the task, including increasing its load and reducing available resources, for the next task in the list , repeat steps S32 and S33 until all tasks are assigned.

[0013] As a further improvement of the technical solution, the transmission rate is ensured to match the processing capability of the node receiving party, as follows: ; in, is the new window size, is the maximum window size, is the old window size; is the amount of confirmation data recently received; The receiver buffer capacity, network delay and packet loss rate factors are introduced into the formula of the new window size for optimization. The optimized expression is: ; in, is the new window size after optimization; For network delay; is the round trip time; is the packet loss rate; is the current available capacity of the receiver buffer; The amount of data that has been queued in the receiver's buffer but has not yet been read by the application.

[0014] Compared with the prior art, the present invention has the following beneficial effects: 1. In this distributed intelligent video image scheduling system, the receiver buffer capacity, network delay and packet loss rate factors are introduced into the window size for window size adjustment, which not only takes into account the network conditions and the processing capacity of the receiver, but also avoids the loss and retransmission of data packets, thereby improving the efficiency and reliability of data transmission, and can more accurately control the data flow, ensure the effective use of network resources and improve user experience; 2. In this distributed intelligent video image scheduling system, the scheduling strategy can more accurately allocate tasks to the most suitable nodes with lower current loads by dynamically evaluating the resource usage of each node and combining the task time, thereby improving resource utilization efficiency and reducing waste. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is the overall flow chart of the present invention; Figure 2 It is a system flow chart of the architecture design unit in the present invention; Figure 3 It is a system flow chart of the scheduling strategy generation unit in the present invention; The meaning of each number in the figure is: 1. Acquisition and processing unit; 2. Architecture design unit; 21. Demand analysis module; 22. Node deployment module; 23. Network construction module; 24. Load balancing module; 25. Fault tolerance mechanism module; 3. Scheduling strategy generation unit; 31. Scheduling module; 32. Resource allocation module; 33. Data flow control module; 4. Decision execution unit. DETAILED DESCRIPTION

[0016] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. Example

[0017] See also Figure 1-Figure 3 As shown, a distributed intelligent video image scheduling system is provided, including an acquisition processing unit 1, an architecture design unit 2, a scheduling strategy generation unit 3 and a decision execution unit 4; The acquisition and processing unit 1 is used to collect video image data using a camera, convert video image data of different formats into a unified format, and pre-process the video image data; such as scaling, decoding, noise filtering, etc., to improve the efficiency of subsequent processing; Architecture design unit 2 is used to deploy each node in the network, determine the location, type, and number of nodes, and design redundancy and recovery strategies to enhance the reliability and fault tolerance of the system, so that it can switch to a backup node when a node fails; achieve load balancing, ensure the reasonable distribution of workloads of each node, and formulate fault tolerance and recovery strategies so that the system can still operate when some nodes fail; The architecture design unit 2 includes a demand analysis module 21, a node deployment module 22, a network construction module 23, a load balancing module 24 and a fault tolerance mechanism module 25; The requirements analysis module 21 is used to clarify the system goals, including the expected functions, performance requirements and non-functional requirements, define key performance indicators such as latency, throughput, availability, etc. to measure system performance; create a system framework to determine the basic structure of the system, including various components and the relationship between them; describe the functions of each node and the relationship between them, that is, the responsibilities and operations of each part of the system, and how they work together; The node deployment module 22 is used to deploy video resource tasks according to demand and realize the allocation of various video resource tasks on different nodes; according to the results of the demand analysis module 21, it decides how to deploy video resource tasks on physical or virtual servers, ensures that video resource tasks are reasonably allocated to different nodes according to their types, performance requirements and dependencies, and implements node deployment strategies to optimize resource utilization and system performance.

[0018] The network construction module 23 is used to determine the communication protocol of the video data and plan the network topology; the network construction module 23 determines the transmission path and topology according to the specifications and performance of the video data collected by the acquisition and processing unit 1 and the bandwidth, delay, throughput and reliability of the network transmission; and determines the communication protocol for video data transmission according to the format and ability of the acquisition and processing unit 1 to process the video data. By selecting appropriate communication protocols, such as TCP / IP, HTTP / 2, WebSockets, etc., to support the efficient transmission of video data, the network topology is designed, including determining the network hierarchy, node connection method and data flow path, to support the scalability and performance requirements of the system; The load balancing module 24 is used to select a load balancing algorithm, adjust the strategy according to the system status, monitor the node health, and redistribute the load when a node failure is detected; a load balancing algorithm such as polling, least number of connections or IP hash is used to disperse the load and ensure that the workload of each node is balanced. The node health is monitored, and when a node failure is detected, the load can be automatically redistributed to maintain the service level of the system, and the strategy can be dynamically adjusted to cope with changes in the system status, such as traffic peaks or fluctuations in node resources.

[0019] The fault-tolerant mechanism module 25 is used for fault detection mechanism, identifying faulty nodes and isolating them from the system, and formulating fault recovery processes; the fault-tolerant mechanism module 25 is a critical component in distributed systems, high-availability architectures, and cloud computing platforms, and is designed to ensure that the system can operate normally even when some components fail. This module monitors and responds to the health status within the system through active detection and passive detection to identify and locate faulty nodes. When a fault occurs, the faulty node is immediately isolated to prevent the fault from spreading and reduce the impact on the system. Formulate and execute fault recovery processes, including automatic restart of services, data recovery mechanisms, and system restarts when necessary, to restore normal operation as soon as possible. The fault detection mechanism in the fault-tolerant mechanism module 25 includes active detection and passive detection.

[0020] Among them, active detection refers to the inspection process performed by the system regularly or at a predetermined frequency, which does not rely on the feedback of the node. It is mainly achieved through the following means: Means 1 is heartbeat detection. Each node in the system will regularly send a heartbeat signal to the fault tolerance mechanism module to prove that it is still active and healthy. If the heartbeat of a node is not received for a period of time, it is considered that the node may have failed.

[0021] The second method is performance indicator monitoring. In addition to heartbeats, a series of performance indicators are also monitored, such as CPU usage, memory usage, disk I / O, network latency, etc., to determine whether the node is in normal working condition. If the performance indicators are found to be out of the normal range, it may indicate that the node is facing some kind of problem.

[0022] The main purpose of active detection is to detect potential problems in a timely manner, even before a node fails completely, allowing the system to take preventive measures to reduce the impact of the failure.

[0023] By combining the node deployment module 22 with the fault tolerance mechanism module 25, when a system failure occurs, services can be restored in the correct order. The specific method is as follows: S1. The fault-tolerant mechanism module 25 continuously monitors the health status of all nodes and detects anomalies in a timely manner; this is active detection, usually using heartbeat detection and performance indicator monitoring methods. When a node detects its own fault, it sends a fault report to the fault-tolerant mechanism module 25, and records and analyzes the fault information, which is passive detection; once the fault is confirmed, the fault-tolerant mechanism module 25 immediately isolates the node from the system to avoid the spread of the fault, and formulates and executes a fault recovery process; for transient faults, the fault-tolerant mechanism module 25 attempts to automatically restart the service to see if the service function can be restored; if the fault causes data damage or loss, the fault-tolerant mechanism module triggers the data recovery process, restores data from the backup, and ensures data consistency; in the case of severe faults, the fault-tolerant mechanism module may initiate a system restart to clear the fault state and reinitialize the system.

[0024] S2, the node deployment module 22 migrates the services on the faulty node to the healthy node according to the current load and resource allocation of the system to ensure task continuity; and determines the correct order of task recovery according to the dependency relationship between video resource tasks to avoid secondary failures caused by improper recovery order; Among them, according to the dependency relationship between video resource tasks, the specific method of determining the correct order of task recovery is as follows: S21. Compute node health score: ; in, For Node Final rating of For Node Utilization rate; For Node The weight of utilization rate; For Node failure rate; For Node The weight of the failure rate; For Node availability; For Node The weight of availability; S22. Calculate the dependencies between video resource tasks and generate a task dependency graph: ; in, is a binary variable, Indicates the task Depends on the task ; Indicates the task Not task dependent ; This function takes two parameters and As input, returns a value representing the task Depends on the task , if the task Need to be on task After completion, you can start or continue, then the function returns This means that there is a To the task Dependencies; if the task Not task dependent , then the function returns This indicates that the task and tasks Is independent, or task May depend on other tasks but is not a task .

[0025] S23, topologically sorting the task dependency graph to determine the execution order list of the tasks; S24. Based on the topologically sorted task list, consider the node health score to assign tasks; give priority to nodes with higher health scores to execute tasks, especially the earlier tasks in the task dependency relationship. This can maximize the overall efficiency and stability of the system; S25. After determining the execution order of tasks and the assigned nodes, implement the task recovery plan; if a problem occurs in a video resource task or node, reschedule the task to a healthier node based on the health score and task dependency graph to maintain service continuity and data integrity; S3. After the service is restored, the fault tolerance mechanism module 25 and the node deployment module 22 jointly check the system status to ensure that all services are running normally.

[0026] Furthermore, the scheduling strategy generation unit 3 is used to analyze video demand in real time, adjust the priority and flow direction of video image data processing, and allocate and manage video resources in combination with the distribution priority algorithm; the scheduling strategy generation unit 3 works in coordination with the architecture design unit 2 to adjust the scheduling strategy according to the system architecture and current load, and send scheduling instructions to the decision execution unit to guide resource allocation and data flow control; The scheduling strategy generation unit 3 includes a scheduling module 31, a resource allocation module 32 and a data flow control module 33; The scheduling module 31 is used to adjust the execution order of the video resource tasks according to the video image data processing information; the scheduling module 31 is based on priority scheduling, which is a common scheduling algorithm that determines the execution order based on the priority of the task or request. In the system, different tasks may have different urgency, importance or urgent needs for resources. Priority scheduling allows the system to dynamically adjust the execution order of tasks based on these factors, thereby ensuring that high-priority tasks can be processed in a timely manner.

[0027] High-priority tasks can be processed faster, reducing waiting time; when resources are limited, critical tasks (such as safety-related tasks or time-sensitive tasks) can be executed first; by prioritizing high-value tasks, the benefits of resources can be maximized; as follows: ; in, is the new priority, is the old priority, The priority adjustment amount is calculated based on certain conditions (such as the waiting time of the task, changes in resource requirements, etc.).

[0028] The resource allocation module 32 is used to balance the requirements of different video resource tasks and reasonably allocate system resources by using a task distribution algorithm in the case of multiple video resource tasks; resource allocation refers to the process of reasonably allocating system resources (such as CPU time, memory, disk space, network bandwidth, etc.) in a multi-task or multi-user environment. An effective resource allocation strategy can balance the requirements of different tasks, avoid resource contention and deadlock problems, and ensure the overall performance and efficiency of the system.

[0029] Ensure that all tasks have access to sufficient resources to prevent some tasks from being left unserved for a long time. Avoid wasting resources through reasonable allocation, so that the system can complete as much work as possible in the shortest time. Prevent low-priority or long-running tasks from being unable to complete due to insufficient resources.

[0030] Among them, the task distribution algorithm balances the needs of different video resource tasks and reasonably allocates system resources, as follows: S31. For all tasks to be executed According to its execution time Sort by size; S32. For each task , find the most suitable node to execute; ; in, For the task At the node The fitness score on ; For Node CPU utilization; For Node The weight of CPU utilization; For Node Memory utilization; For Node The weight of memory utilization; For the task From source node to target node Communication delays; For the task From source node to target node The weight of the communication delay; S33, Selection The node corresponding to the maximum value in As If multiple nodes have the same highest score, the node with the lowest load is selected. S34. Once the task Assigned to the node ,renew The state of the task, including increasing its load and reducing available resources, for the next task in the list , repeat steps S32 and S33 until all tasks are assigned; The data flow control module 33 is used to adjust the sending rate during the video data transmission process to ensure that the sending rate matches the processing capacity of the node receiving party; data flow control is a mechanism to avoid data overflow or packet loss due to processing capacity or buffer capacity limitations of the receiving party during the data transmission process. It ensures that the sending rate matches the processing capacity of the receiving party, thereby maintaining the stability and reliability of communication.

[0031] By adjusting the sending rate, the network or receiving end buffer overflow is avoided and the packet loss is reduced. For real-time applications (such as voice and video streaming), data flow control helps maintain the required delay and packet loss rate; by dynamically adjusting the transmission rate, the speed of data transmission can be increased without affecting data integrity.

[0032] Among them, ensure that the sending rate matches the processing capacity of the node receiver, as follows: Data flow control is based on the sliding window protocol, which controls the data flow by setting the number of packets that a sender can send continuously (i.e., the window size): ; in, is the new window size, is the maximum window size, Is the old window size; reflects the amount of data that the sender can currently send, which changes with the amount of confirmed data grows by the accumulation of It is the amount of confirmed data recently received, usually indicated by the receiver through the sequence number and confirmation number in the ACK (acknowledgement) packet; The receiver buffer capacity, network delay and packet loss rate factors are introduced into the formula of the new window size for optimization. The optimized expression is: ; in, is the new window size after optimization; is the network delay, which is the average time it takes for a data packet to be transmitted in the network; is the round trip time; is the packet loss rate; The current available capacity of the receiver buffer. This value reflects the processing capability of the receiver. If it is very small, it means that the receiver may be processing a large amount of data and cannot receive more data temporarily; This is the amount of data that has been queued in the receiver's buffer but has not yet been read by the application. This value also needs to be fed back by the receiver to the sender to ensure that the sender does not send too much data and cause the receiver's buffer to overflow.

[0033] In the formula Expressions are used to calculate potential data loss due to network latency and packet loss. Network latency and round trip time The ratio represents the proportion of time that a data packet stays in the network per unit time, multiplied by the current window size Then multiply it by the packet loss rate , the potential amount of packet loss due to network delay can be estimated; However, in this expression and The ratio is not a direct loss rate, but is used to quantify the impact of network delay on the potential loss of data packets. can be obtained by the sender through measurement, and and It needs to be estimated through network monitoring or statistical methods; By incorporating these parameters into the formula, we are able to obtain a more accurate and practical window size adjustment algorithm, which not only takes into account the network conditions and the processing power of the receiver, but also avoids the loss and retransmission of data packets, thereby improving the efficiency and reliability of data transmission.

[0034] Finally, the decision execution unit 4 is used to receive the instruction of the scheduling strategy generation unit 3 and execute the allocation operation of the video image data, and at the same time, feed back the execution result to the scheduling strategy generation unit 3.

[0035] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and descriptions are only preferred examples of the present invention and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A distributed intelligent video image scheduling system, It is characterized in that include: The acquisition and processing unit (1) is used to collect video image data using a camera, convert video image data in different formats into a unified format, and pre-process the video image data; Architecture Design Unit (2): used to deploy each node in the network and design redundancy and recovery strategies to switch to a backup node when a node fails; The architecture design unit (2) includes a demand analysis module (21), a node deployment module (22), a network construction module (23), a load balancing module (24) and a fault tolerance mechanism module (25); Scheduling strategy generation unit (3): used to analyze video demand in real time, adjust the priority and flow of video image data processing, and allocate and manage video resources in combination with a distribution priority algorithm; The scheduling strategy generation unit (3) includes a scheduling module (31), a resource allocation module (32) and a data flow control module (33); A decision execution unit (4) is used to receive instructions from the scheduling strategy generation unit (3), execute the allocation operation of the video image data, and feed back the execution result to the scheduling strategy generation unit (3).

2. The distributed intelligent video image scheduling system according to claim 1, It is characterized in that In the architecture design unit (2): The requirement analysis module (21) is used to clarify the system goals, define key performance indicators, create a system framework, and describe the functions of each node and their relationships; The node deployment module (22) is used to deploy video resource tasks according to demand, and realize the allocation of each video resource task on different nodes; The network construction module (23) is used to determine the communication protocol of the video data and plan the network topology; The load balancing module (24) is used to select a load balancing algorithm, adjust the strategy according to the system status, monitor the node health, and redistribute the load when a node failure is detected; The fault-tolerance mechanism module (25) is used for fault detection mechanism, identifying faulty nodes and isolating them from the system, and formulating fault recovery procedures.

3. The distributed intelligent video image scheduling system according to claim 2, Features: The network construction module (23) determines the transmission path and topology structure according to the specifications and performance of the video data collected by the acquisition processing unit (1) and the bandwidth, delay, throughput and reliability of the network transmission; and determines the communication protocol for video data transmission according to the format and capability of the acquisition processing unit (1) to process the video data.

4. The distributed intelligent video image scheduling system according to claim 3, Features: The fault detection mechanism in the fault-tolerant mechanism module (25) includes active detection and passive detection.

5. The distributed intelligent video image scheduling system according to claim 4, Features: The node deployment module (22) is combined with the fault tolerance mechanism module (25) to restore services in the correct order when a system failure occurs. The specific method is as follows: S1, the fault tolerance mechanism module (25) continuously monitors the health status of all nodes and detects abnormalities in a timely manner; when a node detects a fault, it sends a fault report to the fault tolerance mechanism module (25) and records and analyzes the fault information; once the fault is confirmed, the fault tolerance mechanism module (25) immediately isolates the node from the system and formulates and executes a fault recovery process; S2, the node deployment module (22) migrates the services on the failed node to the healthy node according to the current load and resource allocation of the system, and determines the correct order of task recovery according to the dependency relationship between the video resource tasks; S3. After the service is restored, the fault tolerance mechanism module (25) and the node deployment module (22) jointly check the system status to ensure that all services are running normally.

6. The distributed intelligent video image scheduling system according to claim 5, Features: In S2, according to the dependency relationship between the video resource tasks, the specific method for determining the correct order of task recovery is as follows: S21. Compute node health score: ； in, For Node Final rating of For Node Utilization rate; For Node The weight of utilization rate; For Node failure rate; For Node The weight of the failure rate; For Node availability; For Node The weight of availability; S22. Calculate the dependencies between video resource tasks and generate a task dependency graph: ； in, is a binary variable, Representation Task Depends on the task ; Representation Task Not task dependent ; S23, topologically sorting the task dependency graph to determine the execution order list of the tasks; S24, assigning tasks based on the topologically sorted task list, taking into account the health scores of the nodes; S25. After determining the execution order of the tasks and the assigned nodes, implement the task recovery plan.

7. The distributed intelligent video image scheduling system according to claim 6, It is characterized in that In the scheduling strategy generation unit (3): The scheduling module (31) is used to adjust the execution order of the video resource tasks according to the video image data processing information; The resource allocation module (32) is used to balance the requirements of different video resource tasks by using a task distribution algorithm in the case of multiple video resource tasks, and reasonably allocate system resources; The data flow control module (33) is used to adjust the sending rate during the video data transmission process to ensure that the sending rate matches the processing capability of the node receiving party.

8. The distributed intelligent video image scheduling system according to claim 7, Features: The scheduling module (31) is based on priority scheduling, specifically as follows: ； in, is the new priority, is the old priority, is the priority adjustment amount.

9. The distributed intelligent video image scheduling system according to claim 8, Features: The task distribution algorithm balances the requirements of different video resource tasks and allocates system resources reasonably, as follows: S31. For all tasks to be executed According to its execution time Sort by size; S32. For each task , find the most suitable node to execute; ； in, For the task At the node The fitness score on ; For Node CPU utilization; For Node The weight of CPU utilization; For Node Memory utilization; For Node The weight of memory utilization; For the task From source node to target node Communication delays; For the task From source node to target node The weight of the communication delay; S33, Selection The node corresponding to the maximum value in As If multiple nodes have the same highest score, the node with the lowest load is selected. S34. Once the task Assigned to the node ,renew The state of the task, including increasing its load and reducing available resources, for the next task in the list , repeat steps S32 and S33 until all tasks are assigned.

10. The distributed intelligent video image scheduling system according to claim 9, Features: The transmission rate is ensured to match the processing capability of the node receiver, as follows: ； in, is the new window size, is the maximum window size, is the old window size; is the amount of confirmation data recently received; The receiver buffer capacity, network delay and packet loss rate factors are introduced into the formula of the new window size for optimization. The optimized expression is: ； in, is the new window size after optimization; For network delay; is the round trip time; is the packet loss rate; is the current available capacity of the receiver buffer; The amount of data that has been queued in the receiver's buffer but has not yet been read by the application.

Images