A remote sensing monitoring method and system based on edge computing

Abstract

The application discloses a kind of remote sensing monitoring method and system based on edge computing, it is related to remote sensing monitoring technical field, including collection historical remote sensing monitoring data and task division, according to the division result, total task complexity total value and edge node calculation ability value are calculated;With task complexity, clustering is carried out, and the clustering result is output, after the final clustering result is used as individual, population is initialized to carry out overall and local optimization, and the final solution is obtained;The application improves the precision of task scheduling, and the combination of genetic algorithm and ant algorithm enables the application to reasonably schedule according to the actual ability of node, and ensures that high-complexity tasks can be preferentially assigned to nodes with strong capabilities, and more efficient and balanced resource allocation can be achieved, so that the response speed and load balancing capability of the application in the remote sensing monitoring process are significantly improved.

Description

Technical Field

[0001] The present invention relates to the field of remote sensing monitoring technology, and in particular to a remote sensing monitoring method and system based on edge computing. Background Art

[0002] With the continuous development of information technology, remote sensing technology has been widely used in many fields such as earth science, environmental monitoring, and agricultural management. In order to improve the processing efficiency and real-time performance of remote sensing data, edge computing technology has emerged. As a distributed computing architecture, edge computing deploys computing resources at the edge of the network, allowing data storage and processing at or near the data source, significantly reducing dependence on cloud computing resources. Especially in remote areas or when the network connection is unstable, it can effectively realize local processing and intelligent decision-making.

[0003] Traditional remote sensing monitoring methods based on edge computing mostly focus on global optimization, ignoring the characteristics of individual tasks and nodes. As a result, when allocating complex tasks, they are unable to perform fine scheduling based on actual computing power. In addition, they do not fully consider the heterogeneity of edge computing nodes and the complexity of tasks, resulting in unbalanced loads, task delays, and waste of computing resources during task allocation. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a remote sensing monitoring method and system based on edge computing to solve the problems of traditional remote sensing monitoring methods based on edge computing, which mostly focus on global optimization and ignore the characteristics of individual tasks and nodes, resulting in the inability to perform fine scheduling according to actual computing power when allocating complex tasks, and fail to fully consider the heterogeneity of edge computing nodes and the complexity of tasks, resulting in unbalanced load, task delays and waste of computing resources in the task allocation process.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a remote sensing monitoring method based on edge computing, which includes:

[0008] Collect historical remote sensing monitoring data to divide tasks, and calculate the total task complexity and edge node computing capacity based on the division results;

[0009] The historical remote sensing monitoring data includes historical monitoring tasks, execution time of historical tasks and edge node data;

[0010] Based on the complexity of the task, clustering is performed and the clustering results are output. After the final clustering results are used as individuals, the population is initialized for overall and local optimization to obtain the final solution.

[0011] The clustering results include strong and weak cluster clustering results;

[0012] Based on the final solution, the load value is calculated and tested. The maximum and minimum total complexity values ​​are selected based on the test results. These are used as fish individuals in the artificial fish school algorithm. After that, the simulated annealing algorithm is combined with iteration and random perturbation to obtain a load balancing solution.

[0013] The load balancing solutions mentioned above refer to strong and weak load balancing solutions;

[0014] Based on the load balancing solution, the load value is recalculated for secondary detection. After the detection results are displayed through a visual interface, they are stored in the database.

[0015] As a preferred solution of the remote sensing monitoring method based on edge computing described in the present invention, wherein: the clustering is performed based on the complexity of the task, and the clustering results are output. After the final clustering results are used as individuals, the population is initialized for overall and local optimization to obtain the final solution, which includes the following steps:

[0016] Integrate all complexity values ​​and use Excel to draw a box plot based on the integrated total value;

[0017] Based on the boxplot, label low-complexity tasks and high-complexity tasks;

[0018] Integrate the capability values ​​of all edge nodes and arrange them in ascending order;

[0019] When the median of the permutation result is an odd or even number, the average of the middle or middle edge node capability values ​​is taken as the classification threshold;

[0020] Compare the capability value of each edge node with the classification threshold to obtain strong and weak nodes;

[0021] Integrate all tasks and corresponding nodes to obtain the first and second sets, and count the total number of tasks and nodes in the first and second sets respectively;

[0022] The ratio between the total number of tasks and the total number of strong nodes is used as the cluster number of each strong node;

[0023] After initializing each strong node into a strong cluster, the distance from each task to all nodes in the high-complexity task is calculated using the Manhattan distance formula;

[0024] After integrating the distances from each task to all nodes, arrange them in ascending order and assign each task to the corresponding strong cluster according to the minimum distance. Obtain the clustering results of all strong clusters and record the strong cluster labels.

[0025] Similarly, we use the ratio between the total number of tasks and the total number of weak nodes as the cluster number for each weak node, perform the same operation, obtain the clustering results of all weak clusters, and record the weak cluster labels;

[0026] Use hot encoding technique to convert strong cluster labels into binary values;

[0027] After taking each strong cluster in the strong cluster clustering result as a column and the task as a row, a strong assignment matrix is ​​constructed;

[0028] After taking the strong assignment matrix as an individual, randomly generate a population, initialize the binary value and the population, and define the objective function combined with the binary value to minimize the total execution time of the overall task;

[0029] Calculate the objective function value of each individual, use roulette to select the individual with the smallest objective function value as the optimal individual, select individuals for task exchange through crossover and mutation operations, and generate new individuals. During the iterative process, when the number of iterations reaches the maximum, stop the iteration and output the strong basic matrix;

[0030] Similarly, the same operation is performed according to each weak cluster label to obtain the weak basic matrix;

[0031] According to the strong basic matrix, each strong cluster is initialized as each ant individual in the ant algorithm, and the ratio between the node capability value and the total complexity value is used as the path matching value;

[0032] After randomly setting the adjustment values ​​of all paths using a random number generator, the probability of assigning ants to tasks is calculated based on the path matching values.

[0033] Define the objective function to minimize the total task execution time of individual ants;

[0034] The ant algorithm is used to select paths based on the distribution probability. Based on the path selection results, the objective function value of the ant individual is calculated, and the minimum objective function value is selected to update the path. During the iterative process, when the number of iterations reaches the maximum number, the iteration is stopped and the final solution of the strong cluster is output;

[0035] Similarly, each weak cluster in the weak basic matrix is ​​treated as an ant individual in the ant algorithm and the same operation is performed to obtain the final solution of the weak cluster;

[0036] The strong and weak clusters are ultimately resolved to the minimum total task execution time of each cluster.

[0037] As a preferred solution of the remote sensing monitoring method based on edge computing described in the present invention, wherein: the load value is calculated according to the final solution and then detected, and the maximum and minimum total complexity values ​​are selected according to the detection results, and are used as fish individuals in the artificial fish school algorithm, and then combined with the simulated annealing algorithm for iteration and random perturbation to obtain a load balancing solution, which includes the following steps:

[0038] According to the strong and weak final solutions, after extracting all clusters in the strong and weak final solutions, the total execution time mean of all tasks in the cluster is calculated, and the standard deviation of the cluster is taken as the first load value to obtain the first load values ​​of the strong and weak clusters respectively;

[0039] Set the judgment threshold to and , compare the first load values ​​of the strong and weak clusters with the corresponding thresholds and By comparing, we can obtain clusters with strong and weak load imbalance respectively;

[0040] Use the maximum operation to select the maximum total complexity value from the unbalanced clusters, and obtain the maximum total complexity value of the strong and weak clusters respectively;

[0041] Further, according to the strong and weak final solutions, the minimum total complexity value is selected from the clusters of the strong and weak final solutions respectively using the minimization operation, and the maximum and minimum total complexity values ​​are used as fish individuals in the artificial fish school algorithm, and the fish population is randomly generated and initialized;

[0042] The fish individuals and populations include new strong and weak fish individuals and populations respectively;

[0043] The standard deviation of the total execution time of all clusters is used as the objective function value of the individual fish;

[0044] Calculate the objective function value of each individual fish, and use the artificial fish school algorithm to simulate the behavior of the fish school to update the individual fish state. After finding the minimum objective function value through the foraging, grouping, and tail-chasing behaviors of the individual fish, set the initial temperature and cooling parameters of the simulated annealing algorithm;

[0045] The exponential decay method is combined with the cooling parameter to gradually reduce the initial temperature. After the temperature is updated, the fish school is disturbed by random perturbation operation to generate a new fish school. When the number of iterations reaches the maximum number, the iteration and random perturbation operation are stopped, and the load balancing solution is output.

[0046] As a preferred solution of the remote sensing monitoring method based on edge computing described in the present invention, wherein: the recalculation of the load value for secondary detection based on the load balancing solution refers to recalculating the mean of all total execution times of all strong and weak clusters respectively according to the load balancing solution, and taking the standard deviation as the second load value, setting the number of comparisons, and comparing the second load value with the corresponding threshold value. and Compare again, when the second load value of the strong and weak clusters is less than or equal to the corresponding threshold and If the number of comparisons reaches the maximum, the final result is taken as the load balancing result.

[0047] As a preferred solution of the remote sensing monitoring method based on edge computing described in the present invention, the display of detection results through a visual interface refers to using the Matplotlib tool to draw a bar chart and a stacked chart, displaying the total execution time mean and standard deviation of strong and weak nodes through a bar chart, and using a stacked chart to display the changes before and after load optimization.

[0048] As a preferred solution of the remote sensing monitoring method based on edge computing described in the present invention, the storage using a database refers to storing the bar chart and the stacked chart through the MySQL library, and converting the comparison result of the first load value and the second load value into XML format and encrypting it using a symmetric encryption algorithm, and storing the encrypted file in the MySQL library.

[0049] As a preferred solution of the remote sensing monitoring method based on edge computing described in the present invention, wherein: the collecting of historical remote sensing monitoring data to perform task division, and calculating the total task complexity value and the edge node computing capacity value according to the division result, include the following steps:

[0050] Obtain historical monitoring tasks, historical task execution times, and edge node data from remote sensing probes using IoT technology;

[0051] The historical monitoring task data includes image data and image detection data collected in different time periods;

[0052] Preprocessing of collected data, including denoising, cropping and smoothing of image data in history, and image detection data, including feature extraction and classification of preprocessed data in history;

[0053] The execution time of the historical tasks refers to the execution time of each task in the historical monitoring tasks;

[0054] The edge node data refers to the performance data of the node, including node type, CPU performance and memory performance;

[0055] Taking preprocessing as the total task and the execution time of historical tasks as the operation complexity value of each task, we sum the operation complexity values ​​of the tasks to obtain the total complexity values ​​of the total tasks in different historical time periods.

[0056] Using edge node data, use the HiBench tool to obtain the computing power value of each edge node;

[0057] Normalize all the total complexity values ​​and the computing power values ​​of the edge nodes.

[0058] In a second aspect, the present invention provides a remote sensing monitoring system based on edge computing, comprising:

[0059] The acquisition and calculation module is used to collect remote sensing monitoring data for task division and calculate the total task complexity and edge node computing power value;

[0060] Clustering and allocation module, which is used to combine task complexity, perform clustering and global and local optimization to obtain the final solution;

[0061] The detection and load balancing module is used to calculate the load value according to the final solution, perform detection, and perform iteration and random perturbation to obtain the load balancing solution and secondary detection;

[0062] The display and storage module is used to display the test results through a visual interface and then store them in a database.

[0063] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the remote sensing monitoring method based on edge computing as described in the first aspect of the present invention is implemented.

[0064] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the edge computing-based remote sensing monitoring method as described in the first aspect of the present invention.

[0065] The beneficial effects of the present invention are as follows: the present invention greatly improves the accuracy of task scheduling by dividing task complexity into low-complexity tasks and high-complexity tasks, and dividing nodes into strong nodes and weak nodes. The combination of genetic algorithm and ant algorithm enables the present invention to reasonably schedule according to the actual capabilities of the nodes, and ensures that high-complexity tasks can be preferentially allocated to nodes with stronger capabilities, avoiding the waste of computing resources and task delays caused by node heterogeneity and task complexity in traditional methods. Moreover, the combination of artificial fish school algorithm and simulated annealing algorithm enables the present invention to achieve more efficient and balanced resource allocation when processing large-scale tasks, so that the response speed and load balancing capability of the present invention in the remote sensing monitoring process are significantly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0067] Figure 1 This is a flow chart of the remote sensing monitoring method based on edge computing in Example 1.

[0068] Figure 2 This is a structural diagram of the remote sensing monitoring system based on edge computing in Example 1.

[0069] Figure 3 This is a flowchart of task clustering in Example 1.

[0070] Figure 4 This is a flow chart of the ant colony algorithm path selection and optimization in Example 1. DETAILED DESCRIPTION

[0071] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0072] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. 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.

[0073] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0074] Example 1, with reference to Figures 1 to 4 , which is the first embodiment of the present invention, provides a remote sensing monitoring method based on edge computing, comprising the following steps:

[0075] S1. Collect historical remote sensing monitoring data to divide tasks, and calculate the total task complexity and edge node computing capacity based on the division results;

[0076] Historical remote sensing monitoring data includes historical monitoring tasks, execution time of historical tasks, and edge node data;

[0077] Specifically, collecting historical remote sensing monitoring data to divide tasks, and calculating the total task complexity value and edge node computing capacity value based on the division results include the following steps:

[0078] Obtain historical monitoring tasks, historical task execution times, and edge node data from remote sensing probes using IoT technology;

[0079] The historical monitoring task data includes image data and image detection data collected in different time periods;

[0080] Preprocessing of collected data, including denoising, cropping and smoothing of image data in history, and image detection data, including feature extraction and classification of preprocessed data in history;

[0081] The execution time of the historical tasks refers to the execution time of each task in the historical monitoring tasks;

[0082] The edge node data refers to the performance data of the node, including node type, CPU performance and memory performance;

[0083] Take preprocessing as the total task and the execution time of historical tasks as the operation complexity value of each task;

[0084] Sum up the operational complexity values ​​of the tasks to obtain the total complexity values ​​of the total tasks in different historical time periods;

[0085] Using edge node data, use the HiBench tool to obtain the computing power value of each edge node;

[0086] Normalize all the total complexity values ​​and the computing power values ​​of the edge nodes.

[0087] The data quality can be improved by adopting image preprocessing technology, and the computing power of edge nodes can be obtained using the HiBench tool. The task complexity and node capability data are made comparable through normalization operations. This enables the present invention to make reasonable allocations based on the computing power of the nodes, thereby avoiding resource waste and task delays.

[0088] S2. Clustering is performed based on the complexity of the task, and the clustering results are output. The final clustering results are used as individuals, and the population is initialized for overall and local optimization to obtain the final solution.

[0089] Clustering results include strong and weak clustering results;

[0090] Specifically, clustering is performed based on the complexity of the task, and the clustering results are output. After the final clustering results are used as individuals, the population is initialized for overall and local optimization. The final solution is obtained by the following steps:

[0091] After integrating all the total complexity values, the integrated total complexity values ​​were plotted as a box plot using Excel tools;

[0092] According to the box plot, all the complexity values ​​between the lower quartile and the median in the box plot are integrated and marked as low-complexity tasks, and all the complexity values ​​between the median and the upper quartile and the complexity values ​​above the upper quartile are integrated and marked as high-complexity tasks;

[0093] Integrate the capability values ​​of all edge nodes and arrange them in ascending order;

[0094] When the median in the permutation result is an odd number, the middle edge node capability value is taken as the classification threshold;

[0095] When the median in the permutation result is an even number, the average value of the ability values ​​of the middle edge nodes is taken as the classification threshold;

[0096] Compare the capability value of each edge node with the classification threshold. If the capability value of the edge node is greater than or equal to the classification threshold, the edge node is classified as a strong node, otherwise it is classified as a weak node.

[0097] All high-complexity tasks and strong nodes are grouped into the first set, and all low-complexity tasks and weak nodes are grouped into the second set;

[0098] Count the total number of high-complexity tasks and the total number of strong nodes in the first set;

[0099] The ratio between the total number of tasks and the total number of strong nodes is used as the cluster number of each strong node;

[0100] If there is no remainder in the ratio, the current number of clusters is used as the basis, otherwise the number of clusters of all strong nodes is increased until there is no remainder;

[0101] Initialize each strong node as a strong cluster, and get A strong cluster;

[0102] according to A strong cluster is formed, and the distance from each task to all nodes in the highly complex task is calculated using the Manhattan distance formula;

[0103] After integrating the distances from each task to all nodes, arrange them in ascending order and assign each task to the corresponding strong cluster according to the minimum distance. Obtain the clustering results of all strong clusters and record the strong cluster labels.

[0104] Similarly, count the total number of low-complexity tasks and the total number of weak nodes in the second set, and perform the same operation to obtain the clustering results of all weak clusters and record the weak cluster labels;

[0105] Based on each strong cluster label, the strong cluster label is converted into a binary value using the hot encoding technique;

[0106] Further, based on the strong cluster clustering results, each strong cluster in the clustering results is used as a column and the task as a row to construct a strong assignment matrix;

[0107] After taking the strong assignment matrix as an individual, randomly generate the population and initialize the binary value and population;

[0108] Define the objective function combined with binary values ​​to minimize the total execution time of the overall task:

[0109]

[0110] Where, represents the individual objective function value, Indicates the total number of tasks, represents the total number of clusters, Indicates a task Assign to cluster The binary value in , Indicates a task The total complexity of Represents a cluster Node capability value;

[0111] Calculate the objective function value of each individual, use roulette to select the individual with the smallest objective function value as the optimal individual, select individuals for task exchange through crossover and mutation operations, and generate new individuals. During the iterative process, when the number of iterations reaches the maximum, stop the iteration and output the strong basic matrix;

[0112] Similarly, according to each weak cluster label, the hot encoding technique is used to convert the weak cluster label into a binary value;

[0113] Further, based on the weak cluster clustering results, each weak cluster in the clustering results is used as a column and the task as a row to construct a weak assignment matrix. Then, the same operation is performed to obtain a weak basic matrix.

[0114] According to the strong basic matrix, each strong cluster is initialized after being used as each ant individual in the ant algorithm;

[0115] According to the individual ant, the ratio between the node capability value and the total complexity value is used as the path matching value;

[0116] Use a random number generator to randomly set the adjustment values ​​of all paths;

[0117] Combine the path matching value and the adjustment value to calculate the allocation probability between ant individuals and tasks:

[0118]

[0119]

[0120] Where, Indicates a task With individual ants The probability of distribution between Indicates a task With individual ants The adjustment value of the task With individual ants The matching value of Indicates the total number of tasks;

[0121] Define the objective function to minimize the total task execution time of individual ants:

[0122]

[0123] Where, represents the objective function value of the ant individual, represents the total number of ant individuals, Indicates a task The total complexity of Represents an ant individual Node capability value;

[0124] The ant algorithm is used to select paths based on the distribution probability. Based on the path selection results, the objective function value of the ant individual is calculated, and the minimum objective function value is selected to update the path. During the iterative process, when the number of iterations reaches the maximum number, the iteration is stopped and the final solution of the strong cluster is output;

[0125] Similarly, each weak cluster in the weak basic matrix is ​​initialized as each ant individual in the ant algorithm, and the same operation is performed to obtain the final solution of the weak cluster;

[0126] The strong and weak clusters are ultimately resolved to the minimum total task execution time of each cluster.

[0127] By analyzing the complexity of tasks through box plots, tasks are divided into low-complexity and high-complexity categories, and classified according to the computing power values ​​of edge nodes, strong nodes are reasonably assigned to high-complexity tasks, and weak nodes are reasonably assigned to low-complexity tasks. The use of Manhattan distance further refines the matching degree between tasks and nodes by quantifying the distance between tasks and nodes. The strong and weak cluster labels are converted into binary values ​​through unique encoding technology, which effectively simplifies the subsequent optimization algorithm operations. Subsequently, the genetic algorithm generates an initial solution through global optimization search, providing a set of potential path or task allocation schemes. On this basis, the ant algorithm can further optimize path selection and task allocation. By simulating the process of ant foraging, the quality of the solution in the search space is refined, and the efficiency of task execution is improved. Therefore, the genetic algorithm provides a global search framework, and the ant algorithm improves the path selection accuracy through local optimization. Through the linkage between these algorithms, the adaptability and optimization effect of the present invention in complex environments are significantly improved.

[0128] S3. Calculate the load value based on the final solution and then perform a test. Select the maximum and minimum total complexity values ​​based on the test results as the fish individuals in the artificial fish school algorithm. Then, combine the simulated annealing algorithm for iteration and random perturbation to obtain a load balancing solution.

[0129] Load balancing solutions refer to strong and weak load balancing solutions;

[0130] Specifically, according to the final solution, the load value is calculated and then tested. The maximum and minimum total complexity values ​​are selected according to the test results. They are used as fish individuals in the artificial fish school algorithm and then combined with the simulated annealing algorithm for iteration and random perturbation to obtain a load balancing solution. The following steps are included:

[0131] According to the strong and weak final solutions, all clusters in the strong and weak final solutions are extracted;

[0132] After calculating the mean total execution time of all tasks in the cluster using the mean formula, the standard deviation of the cluster is taken as the first load value to obtain the first load values ​​of the strong and weak clusters respectively;

[0133] The judgment threshold is set according to the accuracy requirements and personal experience. and , compare the first load values ​​of the strong and weak clusters with the corresponding thresholds and For comparison, when the first load value of the strong cluster is greater than the threshold When , it indicates that the load of the strong cluster is unbalanced. Similarly, when the first load value of the weak cluster is greater than the threshold When , it indicates that the load of the weak cluster is unbalanced;

[0134] According to the unbalanced clusters, the maximum complexity total value is selected from the clusters using the maximum operation to obtain the maximum complexity total value of the strong and weak clusters respectively;

[0135] Further, according to the strong and weak final solutions, the minimum total complexity value is selected from the clusters of the strong and weak final solutions respectively using the minimization operation;

[0136] The maximum complexity total value and the minimum complexity total value are used as fish individuals in the artificial fish swarm algorithm, and the fish population is randomly generated and initialized;

[0137] The fish individuals and populations include new strong and weak fish individuals and populations respectively;

[0138] The standard deviation of the total execution time of all clusters is used as the objective function value of the individual fish;

[0139] Calculate the objective function value for each individual fish, and use the artificial fish school algorithm to simulate the behavior of the fish school to update the individual fish state. Find the minimum objective function value through the foraging, grouping, and chasing behaviors of the individual fish.

[0140] The initial temperature and cooling parameters of the simulated annealing algorithm are set through experience and prior knowledge;

[0141] The exponential decay method is combined with the cooling parameter to gradually reduce the initial temperature. After the temperature is updated, the fish school is disturbed by random perturbation operation to generate a new fish school. When the number of iterations reaches the maximum number, the iteration and random perturbation operation are stopped, and the load balancing solution is output.

[0142] By calculating the standard deviation of the task execution time in strong and weak clusters respectively, and combining the complexity of the tasks and the capabilities of the nodes, clusters with uneven loads can be accurately identified. This enables the present invention to achieve efficient resource utilization by dynamically adjusting task allocation. Moreover, when the load is uneven, the present invention can adjust according to the maximum and minimum values ​​of the complexity, and flexibly respond to different load conditions, thereby improving the stability and response speed of the present invention. In addition, the present invention uses the artificial fish school algorithm to perform local search in the solution space, simulating the foraging behavior of the fish school to find a better solution, while the simulated annealing algorithm performs a global search for the solution through temperature control to prevent the trap of the local optimal solution and help the algorithm jump out of the local optimal solution. This makes the local exploration of the artificial fish school algorithm provide a more accurate search direction for the simulated annealing, and the simulated annealing algorithm optimizes the solution space of the artificial fish school algorithm through global exploration, so that the two achieve a higher balance in the quality of the solution, which not only increases the diversity of the search and prevents the dilemma of the local optimal solution, but also accelerates the optimization process and avoids the optimization bottleneck that may be caused by a single algorithm.

[0143] S4. Based on the load balancing solution, recalculate the load value for secondary detection, display the detection results through a visual interface, and store them in a database;

[0144] Specifically, based on the load balancing solution, recalculating the load value for secondary detection means recalculating the mean of all total execution time of all strong and weak clusters respectively according to the load balancing solution, taking the standard deviation as the second load value, setting the number of comparisons, and comparing the second load value with the corresponding threshold. and Compare again, when the second load value of the strong and weak clusters is less than or equal to the corresponding threshold and If the number of comparisons reaches the maximum, the final result is taken as the load balancing result.

[0145] Through secondary detection, the load balancing situation can be accurately evaluated, and when the load exceeds the threshold, the optimization process is started to ensure load balancing. The setting of the maximum number of comparisons effectively avoids over-optimization and ensures the efficiency and stability of the process.

[0146] Furthermore, displaying the detection results through a visual interface means using the Matplotlib tool to draw bar charts and stacked charts, using bar charts to display the mean and standard deviation of the total execution time of strong and weak nodes, and using stacked charts to display the changes before and after load optimization.

[0147] By using the Matplotlib tool to draw bar charts and stacked charts, we can intuitively display the execution time of strong and weak nodes and the load optimization effects. The bar chart shows the mean and standard deviation of the node execution time, helping to analyze the execution efficiency and stability between nodes. The stacked chart intuitively shows the changes before and after load optimization, which helps to evaluate the effectiveness of the optimization strategy.

[0148] Furthermore, using a database for storage means storing the bar chart and the stacked chart through a MySQL library, converting the comparison result of the first load value and the second load value into an XML format, encrypting the result using a symmetric encryption algorithm, and storing the encrypted file in the MySQL library.

[0149] By using the MySQL database to store histograms, stacked charts, and load value data, combined with symmetric encryption algorithms and XML format storage, the security and scalability of data management are effectively improved.

[0150] This embodiment also provides a remote sensing monitoring system based on edge computing, including:

[0151] The acquisition and calculation module is used to collect remote sensing monitoring data for task division and calculate the total task complexity and edge node computing power value;

[0152] Clustering and allocation module, which is used to combine task complexity, perform clustering and global and local optimization to obtain the final solution;

[0153] The detection and load balancing module is used to calculate the load value according to the final solution, perform detection, and perform iteration and random perturbation to obtain the load balancing solution and secondary detection;

[0154] The display and storage module is used to display the test results through a visual interface and then store them in a database.

[0155] This embodiment also provides a computer device suitable for the remote sensing monitoring method based on edge computing, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the remote sensing monitoring method based on edge computing proposed in the above embodiment.

[0156] The computer device may be a terminal, comprising a processor, memory, a communication interface, a display, and an input device connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores an operating system and computer programs. The internal memory provides an environment for the operating system and computer programs stored in the non-volatile storage media. The communication interface of the computer device is used to communicate with external terminals via wired or wireless communication. Wireless communication may be achieved via Wi-Fi, a carrier network, NFC (near-field communication), or other technologies. The display of the computer device may be a liquid crystal display or an electronic ink display. The input device may be a touchscreen overlay on the display, buttons, a trackball, or a touchpad on the computer device housing, or an external keyboard, touchpad, or mouse.

[0157] This embodiment also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the remote sensing monitoring method based on edge computing proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, disk or optical disk.

[0158] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. 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 remote sensing monitoring method based on edge computing, characterized by: include, Collect historical remote sensing monitoring data to divide tasks, and calculate the total task complexity and edge node computing capacity based on the division results; The historical remote sensing monitoring data includes historical monitoring tasks, execution time of historical monitoring tasks and edge node data; Based on the complexity of the task, clustering is performed and the clustering results are output. After the final clustering results are used as individuals, the population is initialized for overall and local optimization. The final solution is obtained by the following steps: Integrate all complexity values ​​and use Excel to draw a box plot based on the integrated total value; Based on the boxplot, label low-complexity tasks and high-complexity tasks; Integrate the capability values ​​of all edge nodes and arrange them in ascending order; When the median of the permutation result is an odd or even number, the average of the middle or middle edge node capability values ​​is taken as the classification threshold; Compare the capability value of each edge node with the classification threshold to obtain strong and weak nodes; Integrate all tasks and corresponding nodes to obtain the first and second sets, and count the total number of tasks and nodes in the first and second sets respectively; The ratio between the total number of tasks and the total number of strong nodes is used as the cluster number of each strong node; After initializing each strong node into a strong cluster, the distance from each task to all nodes in the high-complexity task is calculated using the Manhattan distance formula; After integrating the distances from each task to all nodes, arrange them in ascending order and assign each task to the corresponding strong cluster according to the minimum distance. Obtain the clustering results of all strong clusters and record the strong cluster labels. Similarly, we use the ratio between the total number of tasks and the total number of weak nodes as the cluster number for each weak node, perform the same operation, obtain the clustering results of all weak clusters, and record the weak cluster labels; Use hot encoding technique to convert strong cluster labels into binary values; After taking each strong cluster in the strong cluster clustering result as a column and the task as a row, a strong assignment matrix is ​​constructed; After taking the strong assignment matrix as an individual, randomly generate a population, initialize the binary value and the population, and define the objective function combined with the binary value to minimize the total execution time of the overall task; Calculate the objective function value of each individual, use roulette to select the individual with the smallest objective function value as the optimal individual, select individuals for task exchange through crossover and mutation operations, and generate new individuals. During the iterative process, when the number of iterations reaches the maximum, stop the iteration and output the strong basic matrix; Similarly, the same operation is performed according to each weak cluster label to obtain the weak basic matrix; According to the strong basic matrix, each strong cluster is initialized as each ant individual in the ant algorithm, and the ratio between the node capability value and the total complexity value is used as the path matching value; After randomly setting the adjustment values ​​of all paths using a random number generator, the probability of assigning ants to tasks is calculated based on the path matching values. Define the objective function to minimize the total task execution time of individual ants; The ant algorithm is used to select paths based on the distribution probability. Based on the path selection results, the objective function value of the ant individual is calculated, and the minimum objective function value is selected to update the path. During the iterative process, when the number of iterations reaches the maximum number, the iteration is stopped and the final solution of the strong cluster is output; Similarly, each weak cluster in the weak basic matrix is ​​treated as an ant individual in the ant algorithm and the same operation is performed to obtain the final solution of the weak cluster; The final solution of the strong and weak clusters refers to the minimum total task execution time of each cluster; The clustering results include strong and weak cluster clustering results; Based on the final solution, the load value is calculated and tested. The maximum and minimum total complexity values ​​are selected based on the test results. These are used as fish individuals in the artificial fish school algorithm. After that, the simulated annealing algorithm is combined with iteration and random perturbation to obtain a load balancing solution. The load balancing solutions mentioned above refer to strong and weak load balancing solutions; Based on the load balancing solution, the load value is recalculated for secondary detection. After the detection results are displayed through a visual interface, they are stored in the database.

2. The remote sensing monitoring method based on edge computing according to claim 1, characterized in that: The method of calculating the load value according to the final solution and then performing a test, and selecting the maximum and minimum complexity total values ​​according to the test results as the fish individuals in the artificial fish school algorithm, and then performing iteration and random perturbation in combination with the simulated annealing algorithm to obtain a load balancing solution includes the following steps: According to the strong and weak final solutions, after extracting all clusters in the strong and weak final solutions, the total execution time mean of all tasks in the cluster is calculated, and the standard deviation of the cluster is taken as the first load value to obtain the first load values ​​of the strong and weak clusters respectively; Set the judgment threshold to and The first load values ​​of the strong and weak clusters are compared with the corresponding thresholds and By comparing, we can obtain clusters with strong and weak load imbalance respectively; Use the maximum operation to select the maximum total complexity value from the unbalanced clusters, and obtain the maximum total complexity value of the strong and weak clusters respectively; Further, according to the strong and weak final solutions, the minimum total complexity value is selected from the clusters of the strong and weak final solutions respectively using the minimization operation, and the maximum and minimum total complexity values ​​are used as fish individuals in the artificial fish school algorithm, and the fish population is randomly generated and initialized; The fish individuals and populations include new strong and weak fish individuals and populations respectively; The standard deviation of the total execution time of all clusters is used as the objective function value of the individual fish; Calculate the objective function value of each individual fish, and use the artificial fish school algorithm to simulate the behavior of the fish school to update the individual fish state. After finding the minimum objective function value through the foraging, grouping, and tail-chasing behaviors of the individual fish, set the initial temperature and cooling parameters of the simulated annealing algorithm; The exponential decay method is combined with the cooling parameter to gradually reduce the initial temperature. After the temperature is updated, the fish school is disturbed by random perturbation operation to generate a new fish school. When the number of iterations reaches the maximum number, the iteration and random perturbation operation are stopped, and the load balancing solution is output.

3. The remote sensing monitoring method based on edge computing according to claim 2, characterized in that: The recalculation of the load value for secondary detection based on the load balancing solution refers to recalculating the mean of all total execution times of all strong and weak clusters according to the load balancing solution, taking the standard deviation as the second load value, setting the number of comparisons, and comparing the second load value with the corresponding threshold value. and Compare again, when the second load value of the strong and weak clusters is less than or equal to the corresponding threshold and If the number of comparisons reaches the maximum, the final result is taken as the load balancing result.

4. The remote sensing monitoring method based on edge computing according to claim 3, characterized in that: The display of the detection results through a visual interface refers to using the Matplotlib tool to draw a bar chart and a stacked chart, using the bar chart to display the total execution time mean and standard deviation of the strong and weak nodes, and using the stacked chart to display the changes before and after the load optimization.

5. The remote sensing monitoring method based on edge computing according to claim 4, characterized in that: The storage using a database refers to storing the bar chart and the stacked chart through the MySQL library, converting the comparison result of the first load value and the second load value into XML format, encrypting it using a symmetric encryption algorithm, and storing the encrypted file in the MySQL library.

6. The remote sensing monitoring method based on edge computing according to claim 5, characterized in that: The collecting of historical remote sensing monitoring data to divide tasks, and calculating the total task complexity value and edge node computing capacity value according to the division results, includes the following steps: Obtain historical monitoring tasks, their execution times, and edge node data from remote sensing probes using IoT technology; The historical monitoring task includes collecting image data and image detection data in different time periods; Preprocessing of collected data, including denoising, cropping and smoothing of image data in history, and image detection data, including feature extraction and classification of preprocessed data in history; The execution time of the historical monitoring task refers to the execution time of each task in the historical monitoring task; The edge node data refers to the performance data of the node, including node type, CPU performance and memory performance; Taking the preprocessing as the total task and the execution time of the historical monitoring task as the operation complexity value of each task, the operation complexity values ​​of the tasks are summed up to obtain the total complexity values ​​of the total tasks in different historical time periods; Using edge node data, use the HiBench tool to obtain the computing power value of each edge node; Normalize all the total complexity values ​​and the computing power values ​​of the edge nodes.

7. A remote sensing monitoring system based on edge computing, based on the remote sensing monitoring method based on edge computing according to any one of claims 1 to 6, characterized in that: include, The acquisition and calculation module is used to collect remote sensing monitoring data for task division and calculate the total task complexity and edge node computing power value; Clustering and allocation module, which is used to combine task complexity, perform clustering and global and local optimization to obtain the final solution; The detection and load balancing module is used to calculate the load value according to the final solution, perform detection, and perform iteration and random perturbation to obtain the load balancing solution and secondary detection; The display and storage module is used to display the test results through a visual interface and then store them in a database.

8. 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 edge computing-based remote sensing monitoring method according to any one of claims 1 to 6 are implemented.

9. 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 remote sensing monitoring method based on edge computing according to any one of claims 1 to 6 are implemented.

Images