A method for dynamic planning of detection server resources for industrial video surveillance early warning

Abstract

The present invention discloses a method for dynamic resource planning of a detection server for industrial video monitoring and early warning, which relates to the technical field of industrial video monitoring, and includes: collecting dynamic data and static data of a detection server and preprocessing; based on the preprocessed dynamic data, in combination with a neural network model, dynamically evaluating the resource demand of the detection server and outputting a dynamic resource demand evaluation result; determining a preliminary resource capacity planning based on the preprocessed static data, and revising the preliminary resource capacity planning to obtain a static resource capacity planning result; according to the dynamic resource demand evaluation result and the static resource capacity planning result, in combination with a multi-objective optimization algorithm, obtaining a final resource dynamic planning scheme. The present invention comprehensively captures the characteristics and requirements of an industrial video monitoring system by simultaneously considering dynamic data and static data, and optimizes resource allocation on different time scales.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial video monitoring, and in particular to a method for dynamically planning detection server resources for industrial video monitoring and early warning. Background Art

[0002] Industrial video surveillance systems play an increasingly important role in production safety, quality control, and efficiency improvement. However, with the expansion of monitoring scope and the surge in data volume, traditional static resource allocation methods have been unable to meet the needs of industrial video surveillance systems. In recent years, distributed processing architectures based on cloud computing and edge computing have gradually become mainstream, but how to achieve efficient allocation of detection server resources under dynamic load conditions remains an urgent problem to be solved. Existing resource planning methods mainly focus on static capacity planning or simple dynamic scheduling, lacking comprehensive consideration of the overall system performance and resource utilization. In addition, traditional methods often ignore the special needs of industrial video surveillance systems, such as real-time, reliability, and scalability, leading to problems such as unbalanced resource allocation, system response delays, and low cost-effectiveness.

[0003] In order to overcome the above limitations, researchers have proposed a variety of optimization algorithms, such as heuristic algorithms, machine learning methods, etc., to improve the accuracy and efficiency of resource allocation. However, these methods usually only focus on a single goal, such as minimizing response time or maximizing resource utilization, and it is difficult to balance multiple conflicting optimization goals. At the same time, existing methods often lack sufficient adaptability and robustness when dealing with highly dynamic and uncertain industrial environments. In addition, most studies focus on the construction of theoretical models and lack in-depth consideration and verification of actual industrial scenarios, resulting in poor results in practical applications.

[0004] In summary, the existing industrial video surveillance early warning detection server resource planning technology has the following problems: it cannot effectively balance dynamic resource demand and static resource capacity, lacks comprehensive consideration of multi-objective optimization, and is not adaptable enough in actual industrial scenarios. The method for dynamic planning of detection server resources for industrial video surveillance early warning proposed in the present invention aims to solve the above problems by comprehensively considering dynamic data and static data and combining multi-objective optimization algorithms to achieve efficient dynamic planning of resources. The present invention belongs to the field of industrial video surveillance systems and resource optimization management technology, and provides a new solution for improving the performance, reliability and cost-effectiveness of industrial video surveillance systems. Summary of the invention

[0005] In view of the above-mentioned problems, the present invention is proposed.

[0006] Therefore, the technical problem solved by the present invention is: how to provide a method for dynamic planning of detection server resources for industrial video surveillance early warning.

[0007] To solve the above technical problems, the present invention provides the following technical solutions: a method for dynamic planning of detection server resources for industrial video surveillance and early warning, comprising: collecting dynamic data and static data of the detection server and preprocessing them; based on the preprocessed dynamic data, combined with a neural network model, dynamically evaluating the resource requirements of the detection server and outputting dynamic resource requirement evaluation results; determining a preliminary resource capacity planning based on the preprocessed static data, and revising the preliminary resource capacity planning to obtain a static resource capacity planning result; according to the dynamic resource requirement evaluation result and the static resource capacity planning result, combined with a multi-objective optimization algorithm, obtaining a final resource dynamic planning solution.

[0008] As a preferred solution of the method for dynamic planning of detection server resources for industrial video surveillance early warning described in the present invention, the process of obtaining the dynamic resource demand evaluation result includes the following steps: preliminary evaluation of the detection server resource demand based on the improved RNN network:

[0009] D(t) ;

[0010] Where D(t) is the preliminary resource demand assessment value; is the hidden state output of the RNN network at time t; is the context vector of the RNN network at time t; , is the weight matrix; , is the bias vector; and is the activation function.

[0011] Combining the preliminary resource demand assessment value D(t) with the impact of historical data, we can get the revised resource demand assessment value D(t)':

[0012] ;

[0013] in, is a historical data point, is the feature vector at the current moment; is the number of historical data points used to calculate similarity; and To adjust the parameters.

[0014] Combine the corrected resource demand assessment value D(t)' with the forecast error correction to obtain the dynamic resource demand assessment result :

[0015] ;

[0016] in, is the error correction coefficient; is the exponential moving average of the forecast errors.

[0017] = - ;

[0018] in, is the actual observed resource demand.

[0019] As a preferred solution of the method for dynamic resource planning of the detection server of the industrial video surveillance early warning described in the present invention, wherein: determining the preliminary resource capacity planning based on the pre-processed static data, and revising the preliminary resource capacity planning, and obtaining the static resource capacity planning result includes the following steps: obtaining the preliminary resource capacity planning according to the server hardware configuration and the monitoring scene type :

[0020] ;

[0021] in, Configure metrics for server hardware; To monitor the complexity of the scene; is the reference scene complexity; is the factor affecting scene complexity.

[0022] Combined with the change rate of detection algorithm complexity, preliminary resource capacity planning Make corrections to obtain the revised resource capacity plan :

[0023] ;

[0024] in, is the algorithm complexity change rate; is the factor affecting algorithm complexity.

[0025] The revised resource capacity plan Combined with historical load data, static resource capacity planning results are obtained :

[0026] ;

[0027] in, Represents CPU, memory and network resources respectively; is the weight of CPU, memory and network resources, which sums up to 1; is the efficiency factor of CPU, memory and network resources; Historical load data for CPU, memory, and network resources; is the reference load value; Non-linear tuning parameters for CPU, memory, and network resources.

[0028] As a preferred solution of the method for dynamic resource planning of the detection server for industrial video surveillance early warning described in the present invention, obtaining the final dynamic resource planning solution includes using a multi-objective optimization algorithm to combine the dynamic resource demand evaluation results and the static resource capacity planning results to construct multiple multi-objective optimization problems:

[0029] Minimize the deviation between resource allocation and dynamic demand. The objective function is expressed as:

[0030]

[0031] Minimize the deviation of resource allocation from static capacity. The objective function is expressed as:

[0032]

[0033] Minimize the resource allocation cost, the objective function is expressed as:

[0034]

[0035] Maximize system performance, the objective function is expressed as:

[0036]

[0037] Use the non-dominated sorting genetic algorithm II to solve multiple multi-objective optimization problems and obtain a set of Pareto optimal solutions. Select the one with the lowest comprehensive score S(x) from the Pareto optimal solutions as the final resource dynamic planning solution. .

[0038] As a preferred solution of the method for dynamic planning of detection server resources for industrial video surveillance and early warning described in the present invention, the dynamic data refers to data that changes in real time during the operation of the detection server, and the dynamic data includes real-time video traffic, CPU utilization, memory usage, network bandwidth occupancy, current number of detection tasks, detection queue length, average response time, abnormal event frequency, time, date type, special event mark, and system operation status; the static data refers to the statistical results of historical operation data in the industrial video monitoring system and the collection of configuration and environmental parameters that affect the system operation; the static data includes server hardware configuration, historical peak load, average load, detection algorithm complexity, and monitoring scene type.

[0039] As a preferred solution of the method for dynamic planning of detection server resources for industrial video surveillance and early warning described in the present invention, the preprocessing of the dynamic data includes data cleaning, data standardization, time series processing, feature engineering, data aggregation, data smoothing, and anomaly detection; the preprocessing of the static data includes data encoding, feature construction, historical data statistics, correlation analysis, and establishment of a data update mechanism in addition to data cleaning and data standardization.

[0040] As a preferred solution of the method for dynamic planning of detection server resources for industrial video surveillance and early warning described in the present invention, the historical data impact includes resource usage patterns, load trends, anomaly detection, and seasonal adjustments.

[0041] To further solve the above technical problems, the present invention provides the following technical solutions: a detection server resource dynamic planning system in an industrial video surveillance analysis and early warning scenario, comprising: a data acquisition and preprocessing module, used to acquire dynamic data and static data of the detection server and perform preprocessing; a dynamic resource demand assessment module, used to dynamically assess the resource demand of the detection server based on the preprocessed dynamic data in combination with a neural network model and output the dynamic resource demand assessment result; a static resource capacity planning module, used to determine a preliminary resource capacity planning based on the preprocessed static data, and to correct the preliminary resource capacity planning to obtain a static resource capacity planning result; a multi-objective optimization decision module, used to obtain a final resource dynamic planning solution based on the dynamic resource demand assessment result and the static resource capacity planning result in combination with a multi-objective optimization algorithm.

[0042] A computer device includes a memory and a processor, wherein the memory stores a computer program, and is characterized in that when the processor executes the computer program, the steps of the method for dynamically planning detection server resources for industrial video surveillance and early warning as described above are implemented.

[0043] A computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps of the method for dynamically planning detection server resources for industrial video surveillance and early warning as described above are implemented.

[0044] Beneficial effects of the present invention: First, the present invention takes into account both dynamic data and static data, and can more comprehensively capture the characteristics and requirements of industrial video surveillance systems. Dynamic data may include factors that change over time, such as real-time video traffic, complexity of processing tasks, and system load. Static data may involve relatively stable system parameters such as hardware configuration, network topology, and storage capacity. By constructing a dynamic resource demand assessment model and a static resource capacity planning model, the present invention can optimize resource allocation on different time scales. Secondly, the present invention introduces a multi-objective optimization algorithm, which is a significant highlight. In industrial video surveillance systems, it is often necessary to consider multiple conflicting goals at the same time, such as minimizing delay, maximizing throughput, optimizing energy consumption, and improving reliability. The multi-objective optimization algorithm can find the best balance between these goals, thereby obtaining a solution with the best overall performance. In addition, the dynamic programming characteristics of the present invention enable it to adapt to uncertainties and changes in industrial environments. For example, during peak production or when special events occur, the system can quickly adjust resource allocation to cope with sudden high load conditions. At the same time, during low load periods, effective resource recovery and energy saving can also be achieved through dynamic programming. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0046] Figure 1 An overall flow chart of a method for dynamically planning detection server resources for industrial video surveillance early warning provided by an embodiment of the present invention.

[0047] Figure 2 A computer equipment diagram of the method for dynamically planning detection server resources for industrial video surveillance and early warning according to the present invention. DETAILED DESCRIPTION

[0048] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0049] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0050] Example 1, reference Figure 1 , as an embodiment of the present invention, provides a method for dynamically planning detection server resources for industrial video surveillance early warning, comprising:

[0051] S1: Collect dynamic data and static data of the detection server and perform preprocessing.

[0052] In the present invention, dynamic data refers to data that changes in real time during the operation of the detection server. Dynamic data consists of real-time video traffic, CPU utilization, memory usage, network bandwidth occupancy, current number of detection tasks, detection queue length, average response time, abnormal event frequency, time, date type, special event mark, and system operation status.

[0053] Static data refers to the statistical results of historical operation data in industrial video surveillance systems and the collection of configuration and environmental parameters that affect system operation. Static data includes server hardware configuration, historical peak load, average load, detection algorithm complexity, and monitoring scenario type.

[0054] Furthermore, the preprocessing of dynamic data includes data cleaning, data standardization, time series processing, feature engineering, data aggregation, data smoothing, and anomaly detection; the preprocessing of static data includes data encoding, feature construction, historical data statistics, correlation analysis, and establishment of a data update mechanism in addition to data cleaning and data standardization. The preprocessing here adopts conventional techniques in the field, so the present invention will not repeat them.

[0055] S2: Based on the preprocessed dynamic data and combined with the neural network model, the resource requirements of the detection server are dynamically evaluated and the dynamic resource requirement evaluation results are output.

[0056] S2.1: Preliminary evaluation of detection server resource requirements based on the improved RNN network:

[0057] D(t) ;

[0058] Where D(t) is the preliminary resource demand assessment value; is the hidden state output of the RNN network at time t; is the context vector of the RNN network at time t; , is the weight matrix; , is the bias vector; and is the activation function.

[0059] S2.2: Combine the preliminary resource demand assessment value D(t) with the impact of historical data to obtain the revised resource demand assessment value D(t)':

[0060] ;

[0061] in, is a historical data point, is the feature vector at the current moment; is the number of historical data points used to calculate similarity; and To adjust the parameters.

[0062] It should be noted that the impact of historical data includes: resource usage patterns: capturing periodic or repetitive resource usage patterns by comparing the similarity between current data and historical data; load trends: identifying long-term growth or decline trends in resource demand; anomaly detection: identifying potential abnormal resource demand by comparing with historical data; seasonal adjustment: taking into account possible seasonal changes in industrial production.

[0063] S2.3: Combine the corrected resource demand assessment value D(t)' with the forecast error correction to obtain the dynamic resource demand assessment result :

[0064] ;

[0065] in, is the error correction coefficient; is the exponential moving average of the forecast error;

[0066] = - ;

[0067] in, is the actual observed resource demand.

[0068] S3: Determine a preliminary resource capacity planning based on the preprocessed static data, and revise the preliminary resource capacity planning to obtain a static resource capacity planning result.

[0069] S3.1: Get preliminary resource capacity planning based on server hardware configuration and monitoring scenario type :

[0070] ;

[0071] in, Configure metrics for server hardware; To monitor the complexity of the scene; is the reference scene complexity; is the factor affecting scene complexity.

[0072] S3.2: Considering the complexity of the detection algorithm, the revised resource capacity planning is obtained :

[0073] ;

[0074] in, is the algorithm complexity change rate; is the factor affecting algorithm complexity.

[0075] S3.3: Consider historical load data to obtain static resource capacity planning results :

[0076] ;

[0077] in, Represents CPU, memory and network resource types respectively; is the weight of CPU, memory and network resources, which sums up to 1; is the efficiency factor of CPU, memory and network resources; Historical load data for CPU, memory, and network resources; is the reference load value; Non-linear tuning parameters for CPU, memory, and network resources.

[0078] It should be noted that CPU resources are used to process video streams and execute analysis algorithms. Memory resources are used to store temporary data and cache video frames. Network resources are used to transmit video data and communicate with other system components. i∈{c, m, n} means that these three resource types are calculated separately, and then the results are multiplied to obtain a comprehensive planning result that takes into account all resource types. Each resource type has its own specific weight , efficiency factor and nonlinear adjustment parameters ,This allows the model to be flexibly adjusted according to the characteristics and importance of different resource types.

[0079] S4: Based on the dynamic resource demand assessment results and the static resource capacity planning results, combined with the multi-objective optimization algorithm, the final resource dynamic planning solution is obtained.

[0080] Specifically, the following multi-objective optimization problem is constructed by combining the dynamic resource demand evaluation results and the static resource capacity planning results using a multi-objective optimization algorithm:

[0081] Minimize the deviation between resource allocation and dynamic demand. The objective function is expressed as:

[0082]

[0083] Minimize the deviation of resource allocation from static capacity. The objective function is expressed as:

[0084]

[0085] Minimize the resource allocation cost, the objective function is expressed as:

[0086]

[0087] Maximize system performance, the objective function is expressed as:

[0088]

[0089] The constraints are:

[0090] Allocate resources not less than dynamic requirements:

[0091] ;

[0092] Allocate resources not exceeding the static capacity:

[0093] ;

[0094] Allocate resources not less than the minimum requirements:

[0095] ;

[0096] Allocate resources without exceeding the maximum limit:

[0097] ;

[0098] in, is the resource allocation scheme to be optimized, is the decision variable.

[0099] The non-dominated sorting genetic algorithm II is used to solve the above multi-objective optimization problem and obtain a set of Pareto optimal solutions. The one with the lowest comprehensive score S(x) is selected from the Pareto optimal solutions as the final resource dynamic planning solution. ,

[0100] ;

[0101] Embodiment 2 is an embodiment of the present invention, which provides a detection server resource dynamic planning system in an industrial video monitoring analysis and early warning scenario, including:

[0102] The data acquisition and preprocessing module is used to collect and preprocess the dynamic and static data of the detection server; this module is responsible for acquiring various dynamic data (such as real-time video traffic, CPU utilization, memory usage, etc.) and static data (such as server hardware configuration, historical peak load, etc.) from the industrial video surveillance system in real time, and performing preprocessing operations such as cleaning, normalization and feature extraction on these data to provide high-quality input data for subsequent analysis.

[0103] The dynamic resource demand assessment module is used to dynamically evaluate the resource requirements of the detection server based on the preprocessed dynamic data and combined with the neural network model, and output the dynamic resource demand assessment results; this module uses an improved recursive neural network (RNN) or other suitable deep learning models to analyze the preprocessed dynamic data, consider the impact of historical data and prediction error correction, and finally output accurate dynamic resource demand assessment results.

[0104] The static resource capacity planning module is used to determine the preliminary resource capacity planning based on the preprocessed static data, and to revise the preliminary resource capacity planning to obtain the static resource capacity planning results. The module first determines the preliminary resource capacity planning according to the server hardware configuration and the monitoring scenario type, and then makes revisions considering the complexity of the detection algorithm and the historical load data, and finally outputs the static resource capacity planning results that comprehensively consider various factors.

[0105] The multi-objective optimization decision-making module is used to obtain the final resource dynamic planning solution based on the dynamic resource demand evaluation results and the static resource capacity planning results, combined with the multi-objective optimization algorithm; this module uses multi-objective optimization algorithms such as the improved NSGA-II (non-dominated sorting genetic algorithm II), comprehensively considering multiple objectives such as the deviation between resource allocation and dynamic demand, the deviation between resource allocation and static capacity, resource allocation cost and system performance, and finally outputs a resource dynamic planning solution that meets multiple constraints by defining a comprehensive scoring function and selecting the best solution in the Pareto optimal solution set.

[0106] Example 3

[0107] It is an embodiment of the present invention, which is different from the previous embodiment in that:

[0108] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0109] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0110] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0111] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0112] Example 4

[0113] As an embodiment of the present invention, a method for dynamic planning of detection server resources for industrial video surveillance early warning is provided. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0114] Specifically, a medium-sized industrial park is selected, which includes 100 high-definition network cameras and 10 edge servers. The performance of the present invention, the static allocation method and the simple dynamic scheduling method during 72 hours of continuous operation is compared.

[0115] During the experiment, various scenarios of normal working days were simulated, including peak periods, trough periods, and emergencies, and the performance of the three methods on four key indicators: resource utilization, response time, energy efficiency, and system reliability were comprehensively tested. Experimental data was collected every 5 seconds, including dynamic data such as CPU usage, memory usage, and network traffic, as well as static data such as server configuration and historical load. The improved RNN model was used to predict resource demand, and the NSGA-II algorithm was applied for multi-objective optimization to generate a new resource allocation plan every 30 minutes. In addition, emergencies such as factory fire alarms, network attacks, and equipment failures were simulated to evaluate the system's response capabilities. After the experiment, a large amount of collected data was analyzed in detail, the statistical values ​​of various performance indicators were calculated, and the economic benefits were evaluated.

[0116] Table 1: Performance comparison of different methods on various indicators

[0117] ,

[0118] As shown in Table 1, the average resource utilization rate of the present invention reaches 85.7%, which is much higher than 62.3% of the static allocation method and 73.9% of the simple dynamic scheduling method. This shows that the present invention can allocate and utilize server resources more effectively and reduce resource waste.

[0119] The average response time of the present invention is 78ms, which is half as fast as the static allocation method (156ms) and significantly better than the simple dynamic scheduling method (112ms). This significant performance improvement is due to the fact that the present invention can more accurately predict resource requirements and optimize allocation.

[0120] The present invention can process 1250 tasks per kilowatt-hour of energy consumption, which is 47% higher than the static allocation method (850 tasks / kWh) and 19% higher than the simple dynamic scheduling method (1050 tasks / kWh). This reflects the accuracy of the present invention in resource allocation and avoids unnecessary energy waste.

[0121] The system reliability of the present invention reaches 99.95%, which is higher than the other two methods. This is mainly due to the fact that the present invention can better cope with load fluctuations and emergencies, and reduce the probability of system failures and service interruptions.

[0122] The present invention shows obvious advantages in all four key indicators. The present invention not only improves resource utilization, but also significantly reduces response time, while improving energy efficiency and system reliability. These improvements are mainly attributed to the present invention combining dynamic data and static data analysis, using advanced neural network models for resource demand prediction, and adopting multi-objective optimization algorithms to balance various performance indicators.

[0123] It is particularly noteworthy that the present invention performs well in handling load fluctuations. In the case of simulated sudden high load, the present invention can quickly adjust resource allocation and control the increase in response time within 15%, while other methods increase the response time by 40%-60% under the same circumstances.

[0124] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for dynamic planning of detection server resources for industrial video surveillance early warning, characterized in that: include: Collect dynamic data and static data of the detection server and perform preprocessing; Based on the pre-processed dynamic data, combined with the neural network model, dynamically evaluate the resource demand of the detection server and output the dynamic resource demand evaluation result; Determine a preliminary resource capacity planning based on the preprocessed static data, and modify the preliminary resource capacity planning to obtain a static resource capacity planning result; According to the dynamic resource demand evaluation result and the static resource capacity planning result, combined with a multi-objective optimization algorithm, a final resource dynamic planning solution is obtained; The process of obtaining the dynamic resource demand assessment result includes the following steps: A preliminary assessment of the detection server resource requirements is conducted based on the improved RNN network: D(t)=ReLU(W1·h(t)+b1)*Sigmoid(W2·c(t)+b2); Where D(t) is the preliminary resource demand assessment value; h(t) is the hidden state output of the RNN network at time t; c(t) is the context vector of the RNN network at time t; W1, W2 are weight matrices; b1, b2 are bias vectors; ReLU and sigmoid are activation functions; Combining the preliminary resource demand assessment value D(t) with the impact of historical data, we can get the revised resource demand assessment value D(t)': Among them, x i is a historical data point, x t is the feature vector at the current moment; n is the number of historical data points used to calculate similarity; α and β are adjustment parameters; Combine the corrected resource demand assessment value D(t)' with the forecast error correction to obtain the dynamic resource demand assessment result D final : D final =D(t)'+λ·EMA(e t ); Where λ is the error correction coefficient; EMA(e t ) is the exponential moving average of the forecast error; e t =D(t)’-D actual ; Among them, D actual is the actual observed resource demand; Determining a preliminary resource capacity planning based on the preprocessed static data, and revising the preliminary resource capacity planning to obtain a static resource capacity planning result comprises the following steps: Based on the server hardware configuration and monitoring scenario type, we get the initial resource capacity planning C1: C1=H·log(1+exp(η·(S-S ref ))); Among them, H is the server hardware configuration index; S is the monitoring scene complexity index; S ref is the reference scene complexity; η is the scene complexity influencing factor; Combined with the change rate of the detection algorithm complexity, the preliminary resource capacity planning C1 is revised to obtain the revised resource capacity planning C2: in, is the algorithm complexity change rate; δ is the algorithm complexity influencing factor; Combine the revised resource capacity planning C2 with the historical load data to obtain the static resource capacity planning result C final : Among them, c, m, and n represent CPU, memory, and network resources respectively; P i is the weight of CPU, memory and network resources, which sums up to 1; E i is the efficiency factor of CPU, memory and network resources; L i,hist Historical load data for CPU, memory, and network resources; ref is the reference load value; μ i Non-linear adjustment parameters for CPU, memory and network resources; Obtaining the final resource dynamic planning solution includes using a multi-objective optimization algorithm to combine dynamic resource demand assessment results and static resource capacity planning results to construct multiple multi-objective optimization problems: Minimize the deviation between resource allocation and dynamic demand. The objective function is expressed as: f1(x)=|R(x)-D final |; Minimize the deviation of resource allocation from static capacity. The objective function is expressed as: f2(x)=|R(x)-C final |; Minimize the resource allocation cost, the objective function is expressed as: f3(x)=Cost(R(x)); Maximize system performance, the objective function is expressed as: f4(x)=-Performance(R(x)); Use the non-dominated sorting genetic algorithm II to solve multiple multi-objective optimization problems and obtain a set of Pareto optimal solutions. Select the one with the lowest comprehensive score S(x) from the Pareto optimal solutions as the final resource dynamic planning solution R final ; Among them, R(x) is the resource allocation scheme to be optimized, and x is the decision variable; The dynamic data refers to the data that changes in real time during the operation of the detection server, and the dynamic data includes real-time video traffic, CPU utilization, memory usage, network bandwidth occupancy, number of current detection tasks, detection queue length, average response time, abnormal event frequency, time, date type, special event mark, and system operation status; The static data refers to the statistical results of historical operating data in the industrial video surveillance system and a collection of configuration and environmental parameters that affect the system operation; the static data includes server hardware configuration, historical peak load, average load, detection algorithm complexity, and monitoring scene type.

2. The method for dynamic planning of detection server resources for industrial video surveillance early warning according to claim 1, characterized in that: The preprocessing of the dynamic data includes data cleaning, data standardization, time series processing, feature engineering, data aggregation, data smoothing, and anomaly detection; the preprocessing of the static data includes data encoding, feature construction, historical data statistics, correlation analysis, and establishment of a data update mechanism in addition to data cleaning and data standardization.

3. The method for dynamic planning of detection server resources for industrial video surveillance early warning according to claim 2, characterized in that: The historical data impacts include resource usage patterns, load trends, anomaly detection, and seasonal adjustments.

4. A system using the method for dynamic planning of detection server resources for industrial video surveillance early warning according to any one of claims 1 to 3, characterized in that: include: Data collection and preprocessing module, used to collect dynamic data and static data of the detection server and perform preprocessing; A dynamic resource demand assessment module is used to dynamically assess the resource demand of the detection server based on the pre-processed dynamic data and in combination with a neural network model and output the dynamic resource demand assessment result; The static resource capacity planning module is used to determine the preliminary resource capacity planning based on the pre-processed static data, and to modify the preliminary resource capacity planning to obtain the static resource capacity planning result; The multi-objective optimization decision module is used to obtain the final resource dynamic planning solution based on the dynamic resource demand evaluation results and the static resource capacity planning results, combined with the multi-objective optimization algorithm.

5. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for dynamically planning detection server resources for industrial video surveillance and early warning according to any one of claims 1 to 3 are implemented.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for dynamically planning detection server resources for industrial video surveillance and early warning according to any one of claims 1 to 3 are implemented.

Images