Remote sensing intelligent interpretation proxy method and device based on multi-agent cooperation and medium

Abstract

The invention discloses a remote sensing intelligent interpretation proxy method and device based on multi-agent cooperation and a medium, and the method comprises the steps: extracting user query task core information based on a GisRs Brain VLLM model; a plurality of agents cooperate to carry out query task planning, remote sensing data retrieval query, calculation code generation and query model configuration; a plurality of agents cooperatively execute a query calculation task, save calculation data, perform feature analysis according to a calculation result, and generate a preliminary analysis conclusion; and performing result inspection and analysis summarization on the preliminary analysis conclusion to generate a final analysis report. According to the invention, by adapting to the multi-modal sensing feature extraction module special for remote sensing, various modal data can be sensed and understood, the multi-modal information can be effectively integrated and processed, complex information in the remote sensing image can be accurately understood and analyzed, key information in the image can be accurately extracted, and the sensing and understanding capability of the remote sensing image can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of remote sensing data intelligent query and analysis, and particularly to a remote sensing intelligent interpretation agent method, device and medium based on multi-agent cooperation. Background Art

[0002] With the rapid development of artificial intelligence technology, large cross-modal models (Vision-Language Model, abbreviated as VLLM) and intelligent agent (Agent) technology have become essential technical routes for large models to land in business scenarios. These technologies provide intelligent solutions for various complex tasks by simulating the cognitive and decision-making processes of humans. Although LLM and Agent technology have made significant progress in multiple fields, however, in the field of remote sensing, the application of these technologies faces a series of challenges.

[0003] In the application in the field of remote sensing, it is required that the VLLM can combine the information of remote sensing images, fully understand the user's needs and usage scenarios, and require the intelligent agent to be able to process and analyze a large amount of remote sensing data, be able to perform complex planning and decision-making, and further analyze to provide decision support. However, the existing LLM technology often lacks the extraction of remote sensing image features, and the existing multi-agent system has limitations in planning ability, making it difficult to meet the complexity and variability of remote sensing tasks. Therefore, it is necessary to develop an intelligent agent technology specifically for the remote sensing field to overcome the limitations of the existing technology and meet the professional needs of the remote sensing field.

[0004] The main deficiencies of the existing VLLM and Agent technologies mainly include: First, the perception of remote sensing images is insufficient, and the ability to process multi-modal information is lacking: The data involved in the remote sensing field includes not only images, but also multiple modalities such as text and geographical information. The existing technologies often cannot effectively integrate and process these multi-modal information, restricting the comprehensiveness and accuracy of decision-making information. The traditional VLLM and Agent technologies have deficiencies in the perception of remote sensing images, making it difficult to accurately understand and analyze the complex information in remote sensing images and unable to accurately extract the key information in the images.

[0005] Second, the multi-agent planning ability is poor, and there is a lack of domain-specific adaptation: The tasks in the remote sensing field are highly professional, and it is required that the intelligent agent can be adapted according to the domain characteristics. The existing VLLM and Agent technologies often lack in-depth understanding and adaptation to the specific needs of the remote sensing field, resulting in difficulty in achieving the best results in practical applications.

[0006] For example: The invention application with the application number 202310501982.1 discloses an intelligent interpretation method and system for remote sensing image data. Using the application solution, key pixels in remote sensing images can be optimized specifically without manual intervention, greatly improving the interpretation efficiency and accuracy, and providing a more reliable data basis for the work of image interpretation. Although its solution mentions intelligent interpretation, its technical solution cannot be applied to remote sensing image perception, is difficult to accurately understand and analyze complex information in remote sensing images, and cannot accurately extract key information in the images.

[0007] Another example: The invention application with the application number 202111638510.8 discloses an intelligent retrieval method for satellite remote sensing data, which relates to the technical field of remote sensing data processing; through its method, the query threshold can be reduced, the data query process can be simplified, and the user experience can be improved; by establishing a temporary database, the query speed can be increased and the query results can be displayed. However, its solution also has the problems that it cannot perform remote sensing image perception according to the user's query requirements, is difficult to accurately understand and analyze complex information in remote sensing images, and cannot accurately extract key information in the images.

[0008] Therefore, in practical applications, a multi-modal perception intelligent agent method adapted to the remote sensing field is needed. Based on the concurrent collaboration mode of multiple intelligent agents adapted to remote sensing, the image perception ability and planning ability of intelligent agents in the remote sensing field are improved, and the multi-modal information processing ability is enhanced to better meet the professional needs of the remote sensing field. Summary of the Invention

[0009] Aiming at the above existing problems, the purpose of the present invention is to provide a remote sensing intelligent interpretation agent method, device and medium based on multi-agent collaboration, which can accurately understand the user's query information, use multiple agents to cooperate in remote sensing image perception, accurately understand and analyze complex information in remote sensing images, and accurately extract key information in remote sensing images.

[0010] The embodiments of the present invention provide a remote sensing intelligent interpretation agent method, device and medium based on multi-agent collaboration.

[0011] First aspect: A remote sensing intelligent interpretation agent method based on multi-agent collaboration, including the steps of:

[0012] S101. Analyze the user's query task based on the GisRs Brain VLLM model and extract the core information of the user's query task;

[0013] S102. According to the core information of the query task, multiple agents cooperate to perform query task planning, remote sensing data retrieval query, generate calculation codes and query model configuration;

[0014] S103. Based on the execution result of S102, multiple agents collaborate to execute the query calculation task, save the calculation data, perform feature analysis based on the calculation result, and generate a preliminary analysis conclusion.

[0015] S104. After multiple agents collaborate to check the result and analyze and summarize the preliminary analysis conclusion, a final analysis report is generated.

[0016] Among them, multiple agents are driven by the GisRs Brain VLLM model to collaborate.

[0017] Further, the extraction of the core information of the user query task includes:

[0018] Based on the user query task, the GisRs Brain VLLM model uses an image feature extractor to extract the image features of the user input image, uses a text feature extractor to extract the text features of the user input text, and then takes the image features and text features as multi-modal hybrid features for analysis by the GisRs Brain VLLM model to extract the core information of the query task, where:

[0019] The core information of the query task includes the user intention, the remote sensing data range, the remote sensing data time, and the source of the remote sensing image.

[0020] Further, the multiple agents include:

[0021] The main remote sensing interpretation agent Master decomposes and plans the query task according to the core information of the query task, and guides other agents to execute subtasks.

[0022] The remote sensing data agent Agent1 conducts remote sensing data retrieval and query.

[0023] The remote sensing calculation agent Agent2 performs code calculation on the remote sensing data retrieval and query results.

[0024] The remote sensing analysis agent Agent3 conducts feature analysis on the code calculation results.

[0025] Among them, instant communication is carried out between each Agent, and between Master and each Agent.

[0026] Further, when the main remote sensing interpretation agent Master receives the error message during the execution process of other agents, it re-plans the query task.

[0027] Further, the remote sensing data retrieval and query includes the steps of:

[0028] S201. Agent1 generates a query statement according to the query task decomposition and planning, in combination with the core information of the user query task and the remote sensing data warehouse structure.

[0029] S202. Invoke the data retrieval engine to execute the query statement and obtain the query result of the remote sensing image data;

[0030] S203. Return the storage path of the queried remote sensing image data and notify other agents of the abnormal query result.

[0031] Further, when the agent executes the query task calculation task:

[0032] Agent2 combines the core information of the user query task and the data structure of the queried remote sensing image data to generate executable calculation code, or selects the built-in remote sensing calculation interface for code calculation.

[0033] Further, in S4, after multiple agents collaborate to check, analyze, and summarize the preliminary analysis conclusion, a final analysis report is generated, including:

[0034] Master starts the verification of the result report based on the execution results of each Agent, where:

[0035] Agent1 confirms that the data used for calculation is consistent with the query result;

[0036] Agent2 confirms that the calculation result is correct and draws an analysis chart;

[0037] Agent3 confirms that the production of the analysis chart is completed and generates a final analysis report.

[0038] Further, the analysis of the calculation result features includes:

[0039] Agent3 generates a final analysis report according to the summary report template; among them, the final analysis report includes: requirement analysis, data selection, calculation method, statistical results and analysis, and planning suggestions.

[0040] In a second aspect: An electronic device includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the steps of the method provided in the first aspect are implemented.

[0041] In a third aspect: A non-transitory computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method provided in the first aspect are implemented.

[0042] Advantages of the present invention:

[0043] 1. The present invention can perceive and understand various modal data such as images, texts, and geographical information through an adapted multi-modal feature extraction module, effectively integrate and process this multi-modal information, accurately understand and analyze complex information in remote sensing images, accurately extract key information in the images, improve the perception and understanding ability of remote sensing images, and thus more accurately extract and analyze remote sensing data.

[0044] 2. By introducing a multi-agent role assignment and multi-modal information interaction mechanism in the field of remote sensing, the present invention strengthens the in-depth understanding and adaptation to specific requirements in the field of remote sensing, meets the complexity and variability of remote sensing tasks, designs a concurrent collaboration mode for remote sensing multi-agent interaction for the data extraction, calculation, and analysis processes of remote sensing tasks, improves the efficiency and accuracy of remote sensing computing task planning and decision-making, and enhances the application effect in practical applications.

[0045] 3. The present invention solves the deficiencies existing in the existing LLM and intelligent agent application technologies through a multi-modal perception module and a concurrent collaboration mode of multi-agents based on remote sensing adaptation, improves the performance and applicability of intelligent agent technology in the field of remote sensing, and can better meet the professional needs in the field of remote sensing. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a schematic structural diagram of a remote sensing intelligent interpretation agent method based on multi-agent collaboration of the present invention;

[0047] Figure 2 It is a schematic flow diagram of a remote sensing intelligent interpretation agent device based on multi-agent collaboration of the present invention;

[0048] Figure 3 It is a schematic structural diagram of an electronic device of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0049] The embodiments of the present invention will be described in detail below. The examples of the embodiments are shown in the drawings, where the same or similar symbols represent the same or similar elements or elements with the same or similar functions from beginning to end. The embodiments described below with reference to the drawings are exemplary and are only used to explain the present invention and should not be construed as a limitation of the present invention.

[0050] When dealing with remote sensing images, the existing technology often lacks the ability to deeply understand and analyze the image content, resulting in the inability to accurately extract key information in the images. It lacks in-depth understanding and adaptation to specific requirements in the field of remote sensing and is difficult to meet the complexity and variability of remote sensing tasks, resulting in difficulty in achieving the best effect in practical applications.

[0051] To implement the method of the present invention, the present invention discloses an architecture for multi-agent collaboration in the field of remote sensing based on the GisRs Brain VLLM model, as Figure 1As shown, it includes the GisRs Brain VLLM model and multiple agents.

[0052] Among them, the user query task analysis is mainly responsible for parsing by the GisRs Brain VLLM model, and then each step is completed based on the collaboration of multiple remote sensing agents.

[0053] The GisRs Brain VLLM model is a cross-modal large model in the field of remote sensing, with the ability of remote sensing computing knowledge and multi-modal information understanding. The GisRs Brain VLLM model includes a text feature extractor and an image feature extractor, which can complete the text-image cross-modal feature extraction task and better understand the intention of the user query task.

[0054] After obtaining the task information, the GisRs Brain VLLM drives the collaboration of multiple agents to complete the remaining steps; the present invention uses 4 agents, including: the main remote sensing interpretation agent Master, the remote sensing data agent Agent1, the remote sensing computing agent Agent2, and the remote sensing analysis agent Agent3. Instant communication exists between Master and each Agent, and between each Agent.

[0055] Among them, Master decomposes and plans the query task according to the core information of the query task, and guides other agents to execute sub-tasks; Agent1 conducts remote sensing data retrieval and query; Agent2 generates code and performs calculations based on the remote sensing data retrieval and query results; Agent3 conducts feature analysis on the code calculation results.

[0056] Specifically, Master is responsible for decomposing and planning the remote sensing query task, guiding other agents to execute sub-tasks, tracking the overall progress, analyzing the execution information collected from other agents or creating new artifacts. If Master receives error messages during the execution process of other agents, Master can re-plan the overall task.

[0057] For example, if Agent1 gets an empty result when retrieving data according to the core information of the query task, Master will initiate an alternative data selection plan, or initiate multi-round inquiries from the user to change other data retrieval conditions to ensure the continuation of the user query task.

[0058] Agent1 is good at remote sensing data retrieval and query. It can generate query statements based on the core information in the user's query task (such as data range, time node, remote sensing image source, etc.) combined with the remote sensing data warehouse structure, can call the data retrieval engine to execute the query statements and obtain the remote sensing data query results, can correctly return the storage path of the queried remote sensing image data, and can communicate with the Master and other agents in a timely manner for abnormal retrieval results, with corresponding abnormal compatibility processing mechanisms.

[0059] Agent2 is an agent good at code calculation and encoding. Combining the core information of the user's query task and the structure of the queried remote sensing image data, Agent2 can generate executable calculation codes, or select built-in remote sensing calculation interfaces, and configure the correct data input and output paths for code calculation.

[0060] Agent3 is used for generating data analysis conclusions and writing analysis reports. Based on the document writing function in the remote sensing field of VLLM, it can analyze and summarize the execution results of Agent1 and / or Agent2, and can write the analysis report required by the user according to the summary report template and give planning suggestions in combination with the reference knowledge in the remote sensing field.

[0061] Table 1 Multi-agent Collaboration Process Table

[0062]

[0063]

[0064] As shown in Table 1, the following is an example process of the architecture of the present invention for processing a user request. Taking the user request: "Analyze the change in vegetation coverage in XX region in the past 5 years" as an example.

[0065] The GisRs Brain VLLM model drives the cooperation of each agent to complete the query task. Each Agent maintains instant communication with the Master, and each Agent also communicates with each other. The Master fixedly checks the working conditions of each Agent before and after the query task starts and ends. The execution results and abnormal information are synchronized between each Agent and the Master to ensure the synchronization of the execution parameters of the entire query task and the normal flow of the execution status.

[0066] Based on the above architecture of multi-agent collaboration in the remote sensing field, taking the user query task "Analyze the change in vegetation coverage in XX region in the past 5 years" as an example, the present invention discloses a remote sensing intelligent interpretation agent method based on multi-agent collaboration, which uses the architecture of multi-agent collaboration in the remote sensing field for user query tasks, as Figure 2 shown, and the steps included are:

[0067] S101. Analyze the user's query task based on the GisRs Brain VLLM model and extract the core information of the user's query task.

[0068] Based on the user's query task, the GisRs Brain VLLM model uses an image feature extractor to extract the image features of the user's input image and a text feature extractor to extract the text features of the user's input text. Then, the image features and text features are used as multi-modal hybrid features to analyze and extract the core information of the query task through the GisRs Brain VLLM model. The core information of the query task includes: user intent, remote sensing data range, remote sensing data time, and the source of remote sensing images, etc.

[0069] During the query process, the GisRs Brain VLLM model drives the coordinated actions of the Master, Agent1, Agent2, and Agent3 to jointly complete data acquisition, data processing, result review, and analysis report generation. In each step, the corresponding Agent will complete the specific execution tasks corresponding to that step according to its own role.

[0070] Specifically, Agent1 is responsible for data work. In the data acquisition stage, it will automatically generate a SQL statement for remote sensing data retrieval based on the key information such as data time and range obtained from the core information of the query task, and check and confirm the data source. In the data processing stage, Agent1 is responsible for generating the storage path for the remote sensing image data obtained by the query on an annual basis according to the calculation requirements. In the result review stage, Agent1 has no specific execution tasks. At this time, it will receive the confirmation information returned by other agents and check for data anomalies. Agent2 and Agent3 also have different specific execution tasks in different stages, and the entire process is driven by the core tool of the GisRS Brain VLLM model.

[0071] S102. Based on the core information of the query task, multiple agents collaborate to perform query task planning, remote sensing data retrieval query, generate calculation code, and query model configuration.

[0072] The Master is mainly used for task planning and decomposition; first, the Master is responsible for task splitting. According to the core information of the query task obtained in the S101 stage, it assigns the specific execution tasks of Agent1 to Agent3 and generates the specific task execution parameters for these 3 agents.

[0073] Agent1 will automatically generate a SQL statement for remote sensing data retrieval based on the data query instructions and key information parameters such as data time and range in the query task core information assigned by the Master, and check and confirm the data source; Agent2 will generate code for calculating the vegetation index of remote sensing images based on the data calculation instructions and vegetation index tasks assigned by the Master; Agent3 will plan a method for dividing data by year according to the data analysis instructions and five-year calculation tasks assigned by the Master. At the end of step S102, the execution results of Agent1 to Agent3 are summarized and returned to the Master.

[0074] S103. Based on the execution results of S102, multiple agents collaborate to execute query and calculation tasks, save the calculation data, perform feature analysis based on the calculation results, and generate preliminary analysis conclusions.

[0075] Based on the execution results of S102, the Master has received the data query results returned by Agent1, the vegetation index calculation code generated by Agent2, and the data annual statistical caliber generated by Agent3. At this time, it will assign a new round of specific tasks to Agent1 to Agent3.

[0076] Agent1 generates the annual storage path for the data required for the remote sensing calculation queried to provide Agent2 to perform the vegetation index calculation task on the remote sensing data for each year and save the data obtained from the annual vegetation index (NDVI) calculation; Agent3 performs summary statistics based on the 5-year NDVI calculation results to obtain preliminary change analysis conclusions.

[0077] S104. After multiple agents collaborate to check and analyze and summarize the preliminary analysis conclusions, a final analysis report is generated.

[0078] The Master receives the execution results of Agent1 to Agent3 in S103, determines that the calculation task has been completed, and at this time starts result verification. All agents will participate in the review and summary of the execution results to ensure the accuracy and reliability of the analysis results.

[0079] Agent1 confirms that the data on which the calculation is based is consistent with the query results, Agent2 confirms that the results are correct, and draws a change rate chart based on the analysis conclusions of Agent3. Agent3 confirms that the analysis chart is completed and generates a final analysis report according to the summary report template, including parts such as requirements analysis, data selection, calculation methods, statistical results and analysis, and planning suggestions for the entire calculation process.

[0080] Finally, the final output of the entire user query task process is an "Analysis Report on the Changes in Vegetation Coverage in XX Region in the Past 5 Years" accompanied by specific application data, calculation processes, and results.

[0081] The present invention also provides an electronic device, Figure 3 which is a schematic structural diagram of the electronic device provided by an embodiment of the present invention. As Figure 3 shown, the electronic device may include: a processor, a communications interface, a memory, and a communication bus. Among them, the processor, the communication interface, and the memory complete mutual communication through the communication bus. The processor can call the logical instructions in the memory to execute the following methods, for example:

[0082] S101. Analyze the user query task based on the GisRs Brain VLLM model, and extract the core information of the user query task;

[0083] S102. According to the core information of the query task, multiple agents cooperate to perform query task planning, remote sensing data retrieval query, generate calculation codes, and query model configuration;

[0084] S103. Based on the execution result of S102, multiple agents cooperate to execute the query calculation task, save the calculation data, perform feature analysis based on the calculation result, and generate a preliminary analysis conclusion;

[0085] S104. After multiple agents cooperate to check and analyze and summarize the preliminary analysis conclusion, generate a final analysis report.

[0086] Among them, multiple agents are driven by the GisRs Brain VLLM model to cooperate.

[0087] In addition, when the logical instructions in the above-mentioned memory are implemented in the form of software functional units and sold or used as an independent product, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs that can store program codes.

[0088] An embodiment of the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it is implemented to execute the methods provided in the above various embodiments, for example, including:

[0089] S101. Analyze the user's query task based on the GisRs Brain VLLM model and extract the core information of the user's query task;

[0090] S102. According to the core information of the query task, multiple agents collaborate to perform query task planning, remote sensing data retrieval query, generate calculation code, and query model configuration;

[0091] S103. Based on the execution results of S102, multiple agents collaborate to execute the query calculation task, save the calculation data, perform feature analysis based on the calculation results, and generate a preliminary analysis conclusion;

[0092] S104. After multiple agents collaborate to check the results and analyze and summarize the preliminary analysis conclusion, generate a final analysis report.

[0093] Among them, multiple agents are driven by the GisRs Brain VLLM model to collaborate.

[0094] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative labor.

[0095] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence, or the part that contributes to the prior art can be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0096] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A remote sensing intelligent interpretation agent method based on multi-agent collaboration, characterized in that: Includes steps: S101. Based on the GisRs Brain VLLM model, analyze the user query tasks and extract the core information of the user query tasks; S102, based on the query task core information, multiple agents collaborate to perform query task planning, remote sensing data retrieval query, computational code generation and query model configuration; S103, based on the execution result of S102, multiple agents collaborate to execute the query calculation task, save the calculation data, perform feature analysis based on the calculation results, and generate preliminary analysis conclusions; S104, after multiple agents collaborate to check and analyze the preliminary analysis conclusions, a final analysis report is generated; Among them, multiple agents collaborate driven by the GisRs Brain VLLM model.

2. The method according to claim 1, characterized in that: The extracting of core information of the user query task includes: The GisRs Brain VLLM model uses an image feature extractor to extract image features of user input images based on user query tasks, and uses a text feature extractor to extract text features of user input texts. Then, the image features and text features are analyzed as multimodal mixed features by the GisRs Brain VLLM model to extract the core information of the query task, including: The core information of the query task includes user intention, remote sensing data range, remote sensing data time and remote sensing image source.

3. The method according to claim 1, characterized in that The plurality of agents include: The remote sensing interpretation master agent Master decomposes and plans the query task based on the core information of the query task and guides other agents to perform subtasks; Remote sensing data agent Agent1, performs remote sensing data retrieval query; Remote sensing computing agent Agent2 performs code calculation on remote sensing data retrieval query results; Remote sensing analysis agent Agent3 performs feature analysis on the code calculation results; Among them, there is instant communication between each Agent and between the Master and each Agent.

4. The method according to claim 3, characterized in that When the remote sensing interpretation master agent Master receives error information from the execution process of other agents, it re-plans the query task.

5. The method according to claim 3, characterized in that: Conduct remote sensing data retrieval query, including the following steps: S201, Agent 1 generates query statements based on the query task decomposition and planning, combined with the core information of the user's query task and the remote sensing data warehouse structure; S202, calling a data retrieval engine to execute a query statement and obtain remote sensing image data query results; S203. Return the storage path of the queried remote sensing image data and notify other agents of the abnormal query results.

6. The method according to claim 3, characterized in that When the agent performs the query task calculation task: Agent2 combines the core information of the user's query task and the data structure of the queried remote sensing image data to generate executable computing code, or selects the built-in remote sensing computing interface to perform code calculation.

7. The method according to claim 3, characterized in that In S4, multiple agents collaborate to check and analyze the preliminary analysis conclusions and generate a final analysis report, including: The Master starts the result report verification based on the execution results of each Agent, including: Agent 1 confirms that the data used for calculation is consistent with the query result; Agent 2 confirms that the calculation results are correct and draws analysis charts; Agent3 confirms that the analysis chart is complete and generates a final analysis report.

8. The method according to claim 7, characterized in that The calculation result characteristic analysis includes: Agent3 generates a final analysis report based on the summary report template; the final analysis report includes: demand analysis, data selection, calculation method, statistical results and analysis, and planning suggestions.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the method according to any one of claims 1 to 8 are implemented.

10. A non-transitory 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 according to any one of claims 1 to 8 are implemented.

Images