Large language model assisted remote sensing index calculation method and device and medium

Abstract

The invention discloses a large language model assisted remote sensing index calculation method and device and a medium, and the method comprises the steps: inputting a calculation request of a user and a predefined system cue word into a large language model to analyze the intention of the user, and generating a tool call instruction; according to a tool name in the tool calling instruction, calling a corresponding MCP remote sensing tool set service to execute auxiliary remote sensing index calculation; recombining the tool execution result and the remote sensing tool set description information into a new cue word, and inputting the new cue word into the large language model; tool calling and result feedback operation is executed circularly until the large language model does not generate a new tool calling instruction any more, and final response data generated by the large language model is output to the client side.

Description

Technical Field

[0001] The present application relates to the field of remote sensing technology, and in particular to a method, device and medium for calculating remote sensing indicators assisted by a large language model. Background Art

[0002] Currently, remote sensing data processing relies heavily on specialized software tools and manual workflows. Users must master remote sensing terminology, data format specifications, and spatial analysis algorithm principles to complete basic operations such as vegetation index calculation and surface temperature inversion. This technical barrier to entry excludes non-professional users from remote sensing applications, severely restricting the large-scale adoption of remote sensing technology across diverse fields, such as agricultural monitoring and environmental assessment.

[0003] While existing automation solutions can execute single remote sensing metric calculations through pre-set scripts, they suffer from the inherent flaw of rigid processes. Their toolchains only support the linear execution of fixed algorithm combinations, preventing the flexible deployment of tool sets based on dynamic user needs. Furthermore, traditional architectures suffer from significant deficiencies in tool collaboration and state management. Each computing module is typically an independent service, lacking a unified scheduling hub. This fragmented operation significantly increases cognitive load, and errors can easily lead to the failure of the entire analysis process. Summary of the Invention

[0004] The embodiments of the present application provide a large language model-assisted remote sensing indicator calculation method, device and medium to solve the above-mentioned technical problems.

[0005] On the one hand, the embodiments of the present application provide a large language model-assisted remote sensing indicator calculation method, including: Input the user's computing request and predefined system prompt words into the large language model to analyze the user's intention and generate tool call instructions; According to the tool name in the tool calling instruction, the corresponding MCP remote sensing tool set service is called to perform auxiliary remote sensing indicator calculation; Recombining the tool execution results and the remote sensing toolset description information into new prompt words, and inputting the new prompt words into the large language model; The tool calling and result feedback operations are executed cyclically until the large language model no longer generates new tool calling instructions, so as to output the final response data generated by the large language model to the client.

[0006] In one implementation of the present application, the user's computing request and predefined system prompt words are input into the large language model, specifically including: Receive natural language computing requests input by the user through the client, and determine the remote sensing indicators to be calculated in the computing requests; Determining remote sensing tool set description information corresponding to the remote sensing indicator to be calculated, and using the remote sensing tool set description information as a system prompt word; The calculation request and the system prompt word are input into a large language model.

[0007] In one implementation of the present application, parsing the user's intent and generating a tool call instruction specifically includes: Identifying a geographic region identifier in the computation request; The geographic area identifier is matched with the preset remote sensing data spatial range, the MCP tool name to be called and the tool calling area parameters are determined according to the matching result, and a tool calling instruction corresponding to the MCP tool name is generated.

[0008] In one implementation of the present application, calling the corresponding MCP remote sensing tool set service according to the tool name in the tool calling instruction specifically includes: Verify that the tool call area parameters comply with the parameter specifications of the corresponding MCP remote sensing tool set service; If parameter validation fails, an error message is returned to the large language model. If parameter validation succeeds, the tool execution process is triggered.

[0009] In one implementation of the present application, performing auxiliary remote sensing index calculation specifically includes: According to the tool name, determine whether it contains the corresponding target remote sensing indicator; If not, then retrieving the corresponding remote sensing image data according to the tool calling parameters, and performing target remote sensing index calculation on the remote sensing image data; Perform spatial clipping on the calculation results according to the requested geographic area range corresponding to the tool call area parameter, and generate the calculation result storage path and download address.

[0010] In one implementation of the present application, the tool call and result feedback operations are executed cyclically, specifically including: Return the calculation result of each tool call in the data format of structured data; the structured data includes a status code, result file address and error description field; According to the status code in the structured data, the large language model is used to determine whether to initiate a call to a new tool.

[0011] In one implementation of the present application, outputting the final response data generated by the large language model to the client specifically includes: Automatically integrate intermediate results generated by multi-cycle step tool calls to generate a natural language description; the natural language description includes a link to a downloadable result file; Obtaining an execution status log of a multi-loop step tool call, and determining corresponding error handling suggestions based on the execution status log; The execution status log and the error handling suggestion are added to the natural language description to generate final response data corresponding to the large language model.

[0012] In one implementation of the present application, the present invention further includes: Before calling the MCP Remote Sensing Toolset service for the first time, load the predefined prompt word template for remote sensing applications and obtain the real-time description information of the MCP Remote Sensing Toolset; The prompt word template is bound to the real-time description information, and an initialization context environment of a large language model is dynamically constructed.

[0013] On the other hand, an embodiment of the present application further provides a large language model-assisted remote sensing index calculation device, the device comprising: at least one processor; and, a memory communicatively coupled to the at least one processor; In which, the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the large language model-assisted remote sensing indicator calculation method as described above.

[0014] On the other hand, an embodiment of the present application also provides a non-volatile computer storage medium storing computer executable instructions. When the computer executable instructions are executed, a large language model-assisted remote sensing indicator calculation method as described above is implemented.

[0015] The present application provides a method, device, and medium for calculating remote sensing indicators assisted by a large language model, which have at least the following beneficial effects: By inputting the user's natural language computing request and predefined system prompt words into the large language model, the user's intention is directly analyzed and tool call instructions are generated. Non-professionals do not need to master remote sensing professional terminology or algorithm principles. Users can directly complete remote sensing indicator calculations through natural language instructions, significantly lowering the threshold for technology use; the tool call instructions generated based on the large language model accurately trigger the MCP remote sensing toolset service, automatically matching computing requirements with tool functions, avoiding the tedious manual operation of searching for tools and configuring parameters in traditional solutions, and solving the problem of rigid tool call process; by reorganizing the tool execution results and toolset description information into new prompt words and inputting them into the large language model, and looping until no new tool call instructions are generated, autonomous and collaborative execution of multi-step remote sensing tasks is achieved; finally, the response data generated by the large language model is output to the client, intelligently integrating the execution information of the entire process, avoiding omissions or misoperations caused by manual collection of fragmented outputs in the traditional model, and improving the availability and operability of remote sensing analysis results. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings: Figure 1 A flowchart of a method for calculating remote sensing indicators assisted by a large language model provided in an embodiment of the present application; Figure 2 A schematic diagram of the internal structure of a large language model-assisted remote sensing index calculation device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0017] To make the purpose, technical solutions, and advantages of this application more clear, the technical solutions of this application will be clearly and completely described below in conjunction with the specific embodiments of this application and the corresponding drawings. Obviously, the embodiments described are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0018] The following describes in detail the technical solutions provided by various embodiments of the present application in conjunction with the accompanying drawings.

[0019] Figure 1 A flowchart of a large language model-assisted remote sensing index calculation method provided in an embodiment of the present application.

[0020] The analysis method involved in the embodiments of the present application can be implemented by a terminal device or a server, and the present application does not impose any special restrictions on this. For ease of understanding and description, the following embodiments are described in detail using a server as an example.

[0021] It should be noted that the server can be a single device or a system composed of multiple devices, that is, a distributed server, and this application does not make any specific restrictions on this.

[0022] like Figure 1 As shown, the embodiment of the present application provides a large language model assisted remote sensing index calculation method, including: Step 101: Input the user's computing request and predefined system prompt words into the large language model to analyze the user's intention and generate a tool call instruction.

[0023] It should be noted that prompt words are divided into system prompt words and user prompt words. System prompt words are fixed and represent the common remote sensing processing process. User prompt words are used to input user business computing requirements and are entered by users when asking questions.

[0024] First, prepare a Large Language Model (LLM) that supports Model Context Protocol (MCP) tool calls and deploy the LLM. Also prepare the remote sensing metrics to be calculated, such as NDVI, FVC, TVDI, and PDI, listing them based on the actual application. Then, write MCP services based on the remote sensing metrics, creating a tool for each metric.

[0025] It should be noted that the MCP service is deployed as a remote service or a local resource service as needed. The MCP client represents a one-to-one connection with the server, which is equivalent to a client that needs to be implemented to implement database interaction in the application.

[0026] In this example, a natural language computing request entered by the user through the client, along with a predefined system prompt, is input into the LLM. For example, a natural language computing request such as "Get vegetation coverage in the southern mountainous area" is used. It should be noted that the system prompt includes the remote sensing processing rule base and the MCP toolset description specification. The remote sensing processing rule base, for example, describes the process logic of "first search for images, then calculate indicators," and the MCP toolset description specification, for example, uses the unified prefix "mcp" for tool names.

[0027] Specifically, the large language model identifies the term "FVC" for "vegetation cover" and maps it to the "mcpFvc" service in the toolset. This service's parameter specifications, such as the requirement for a region input in GeoJSON format, are used as part of the system prompt. Using a place name database, the spatial coordinates for "Southern Mountain Area" are converted, with polygon boundaries described in WKT format.

[0028] The resulting structured tool call instructions are in JSON format, including the tool name "mcpFvc" and the parameters {"roi":"POLYGON((x1 y1,x2 y2,...))"}. This process converts unstructured natural language into standardized, machine-executable instructions.

[0029] Step 102: According to the tool name in the tool calling instruction, the corresponding MCP remote sensing tool set service is called to perform auxiliary remote sensing indicator calculation.

[0030] In one embodiment of this application, an MCP service toolset is determined based on the remote sensing metrics to be calculated, and a remote sensing application LLM prompt is written. These two components serve as input to the LLM, which returns the result information to the MCP client. The MCP client then calls the tools based on the list of tools returned by the LLM. If no tool list is available, the loop is not entered.

[0031] In this embodiment, after receiving the tool call, the MCP client checks whether the ROI parameter is a closed polygon, illustratively by calculating the consistency of the coordinates of the first and last points. This verifies that the polygon area is within a reasonable threshold to avoid excessive computational complexity caused by misoperation. If verification fails, {status:400,error:"Region format invalid"} is returned to the LLM.

[0032] In this embodiment, during the data retrieval phase, spatiotemporally matched source data is screened from the image library based on the ROI range, with Sentinel-2 images with cloud cover below a preset value being preferentially selected. During the indicator calculation decision, if the target FVC indicator has not been precalculated, the (NIR-R) / (NIR+R) algorithm is called to generate raster data. Spatial processing operations crop the resulting image according to the ROI boundary and store it in blocks in the object storage service, returning a structured response {status:200, data_url:"s3: / / bucket / fvc.tif"}. It should be noted that if data retrieval fails, such as if there is no image covering the area, {status:404, error:"No source data available"} is returned.

[0033] Step 103: Recombine the tool execution results and the remote sensing toolset description information into new prompt words, and input the new prompt words into the large language model.

[0034] The tool call results are organized into prompt words and input into the LLM again along with the toolset information. The LLM then returns the result information until the LLM returns no more tool list information. The information returned by the LLM at the end is the result information.

[0035] In this embodiment, the status code returned by the tool is mapped to natural language, exemplarily converting 200 to "success" and 500 to "service exception", and embedding data_url into an actionable tag, such as "download link: [s3: / / bucket / fvc.tif]".

[0036] Get the latest MCP toolset description in real time. For example, pull the newly added tool "mpcRender" and its parameter description through the registration center API and append it to the prompt word.

[0037] The combination generates a message such as: "FVC calculation for the southern mountainous area completed (status: success). Download link: [s3: / / bucket / fvc.tif]. Is further rendering required? Available tool: mpcRender(tif_path)". It is understandable that this design enables the LLM to make subsequent decisions based on real-time context, such as calling a rendering tool to process the downloaded tif file.

[0038] Step 104 : cyclically execute the tool calling and result feedback operations until the large language model no longer generates new tool calling instructions, so as to output the final response data generated by the large language model to the client.

[0039] In this embodiment, when the LLM response no longer contains the tool_calls field, it indicates that the task has been determined to be completed, and the loop is terminated.

[0040] Aggregate multiple rounds of output, for example, package the FVC grid and rendering into a ZIP, generate a new download link, and extract error codes from the execution log and convert them into natural language suggestions. Error codes such as cropping timeout error and natural language suggestions such as "It is recommended to reduce the area range."

[0041] Before the first loop, the LLM context is dynamically constructed, prompt word templates are loaded, such as the fixed instruction: "You are responsible for coordinating the remote sensing tool call", and the real-time toolset description is bound, such as injecting "Currently available tools: mcpFvc, mpcRender...", and the LLM initial system role is set, for example, "Role: Remote Sensing Analysis Assistant".

[0042] Then, a natural language report is generated and returned to the client. The natural language report is such as "FVC calculation of the southern mountainous area has been completed [download link], and the rendering image generation failed." It is recommended that the area is too large and needs to be split. The client will parse it into an interactive interface, which includes a download button and an error prompt box.

[0043] In one embodiment of the present application, there is an MCP service that can calculate the NDVI normalized vegetation index, named mcpNdvi. The input is the region of interest ROI, and the output is the download address of the NDVI image within the cropped ROI.

[0044] Calculate the NDVI for the Southern Mountainous Region: The LLM response contains a function called "tool_calls" in the "choices" section. Within tool_call, the tool name is mcpNdvi and the parameter "Southern Mountainous Region." The client calls the mcpNdvi function and enters the ROI parameter "Southern Mountainous Region." mcpNdvi internally locates the remote sensing image containing the ROI and determines whether the NDVI has been calculated. If not, it calculates the NDVI and then crops the image based on the ROI. If so, it crops the image directly. The resulting cropping result is then downloaded.

[0045] The above is an embodiment of the method proposed in this application. Based on the same inventive concept, this application embodiment also provides a large language model assisted remote sensing index calculation device, whose structure is as follows Figure 2 shown.

[0046] Figure 2 This is a schematic diagram of the internal structure of a large language model assisted remote sensing index calculation device provided in an embodiment of the present application. Figure 2 As shown, the equipment includes: at least one processor; and, a memory communicatively coupled to the at least one processor; The memory stores instructions that can be executed by at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to: Input the user's computing request and predefined system prompt words into the large language model to analyze the user's intention and generate tool call instructions; According to the tool name in the tool call instruction, call the corresponding MCP remote sensing tool set service to perform auxiliary remote sensing indicator calculation; Recombining tool execution results and remote sensing toolset description information into new prompt words, and inputting the new prompt words into the large language model; The tool call and result feedback operations are executed cyclically until the large language model no longer generates new tool call instructions, so as to output the final response data generated by the large language model to the client.

[0047] The present application also provides a non-volatile computer storage medium storing computer-executable instructions. When the computer-executable instructions are executed, they can: Input the user's computing request and predefined system prompt words into the large language model to analyze the user's intention and generate tool call instructions; According to the tool name in the tool call instruction, call the corresponding MCP remote sensing tool set service to perform auxiliary remote sensing indicator calculation; Recombining tool execution results and remote sensing toolset description information into new prompt words, and inputting the new prompt words into the large language model; The tool call and result feedback operations are executed cyclically until the large language model no longer generates new tool call instructions, so as to output the final response data generated by the large language model to the client.

[0048] The various embodiments in this application are described in a progressive manner. Similar portions between the various embodiments can be referred to in conjunction with each other. Each embodiment focuses on the differences between the other embodiments. In particular, the device and medium embodiments are generally similar to the method embodiments, so their descriptions are relatively simple. For relevant portions, refer to the descriptions of the method embodiments.

[0049] The devices and media provided in the embodiments of the present application correspond one-to-one to the methods. Therefore, the devices and media also have similar beneficial technical effects to their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the devices and media will not be repeated here.

[0050] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0051] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0052] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0053] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0054] In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0055] Memory may include non-permanent storage in a computer-readable medium, in the form of random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0056] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can be implemented using any method or technology for information storage. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change RAM (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices, or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory computer-readable media, such as modulated data signals and carrier waves.

[0057] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0058] The above are merely embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should all be included within the scope of the claims of the present application.

Claims

1. A large language model-assisted remote sensing index calculation method, characterized in that: The method comprises: Input the user's computing request and predefined system prompt words into the large language model to analyze the user's intention and generate tool call instructions; According to the tool name in the tool calling instruction, the corresponding MCP remote sensing tool set service is called to perform auxiliary remote sensing indicator calculation; Recombining the tool execution results and the remote sensing toolset description information into new prompt words, and inputting the new prompt words into the large language model; The tool calling and result feedback operations are executed cyclically until the large language model no longer generates new tool calling instructions, so as to output the final response data generated by the large language model to the client.

2. A large language model-assisted remote sensing index calculation method according to claim 1, characterized in that: The user's computing request and predefined system prompt words are input into the large language model, including: Receive natural language computing requests input by the user through the client, and determine the remote sensing indicators to be calculated in the computing requests; Determining remote sensing tool set description information corresponding to the remote sensing indicator to be calculated, and using the remote sensing tool set description information as a system prompt word; The calculation request and the system prompt word are input into a large language model.

3. The method for calculating remote sensing indicators assisted by a large language model according to claim 1, characterized in that: Analyze user intent and generate tool call instructions, including: Identifying a geographic region identifier in the computation request; The geographic area identifier is matched with the preset remote sensing data spatial range, the MCP tool name to be called and the tool calling area parameters are determined according to the matching result, and a tool calling instruction corresponding to the MCP tool name is generated.

4. The method for calculating remote sensing indicators assisted by a large language model according to claim 1, characterized in that: According to the tool name in the tool call instruction, the corresponding MCP remote sensing tool set service is called, including: Verify that the tool call area parameters comply with the parameter specifications of the corresponding MCP remote sensing tool set service; If parameter validation fails, an error message is returned to the large language model. If parameter validation succeeds, the tool execution process is triggered.

5. The method for calculating remote sensing indicators assisted by a large language model according to claim 1, characterized in that: Perform auxiliary remote sensing indicator calculations, including: According to the tool name, determine whether it contains the corresponding target remote sensing indicator; If not, then retrieving the corresponding remote sensing image data according to the tool calling parameters, and performing target remote sensing index calculation on the remote sensing image data; Perform spatial clipping on the calculation results according to the requested geographic area range corresponding to the tool call area parameter, and generate the calculation result storage path and download address.

6. A large language model-assisted remote sensing index calculation method according to claim 1, characterized in that: Execute tool calls and result feedback operations cyclically, including: Return the calculation result of each tool call in the data format of structured data; the structured data includes a status code, result file address and error description field; According to the status code in the structured data, the large language model is used to determine whether to initiate a call to a new tool.

7. The method for calculating remote sensing indicators assisted by a large language model according to claim 1, characterized in that: Outputting the final response data generated by the large language model to the client specifically includes: Automatically integrate intermediate results generated by multi-cycle step tool calls to generate a natural language description; the natural language description includes a link to a downloadable result file; Obtaining an execution status log of a multi-loop step tool call, and determining corresponding error handling suggestions based on the execution status log; The execution status log and the error handling suggestion are added to the natural language description to generate final response data corresponding to the large language model.

8. The method for calculating remote sensing indicators using a large language model according to claim 1, wherein: The method further comprises: Before calling the MCP Remote Sensing Toolset service for the first time, load the predefined prompt word template for remote sensing applications and obtain the real-time description information of the MCP Remote Sensing Toolset; The prompt word template is bound to the real-time description information, and an initialization context environment of a large language model is dynamically constructed.

9. A large language model-assisted remote sensing index calculation device, characterized in that: The device comprises: at least one processor; and, a memory communicatively coupled to the at least one processor; In which, the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute a large language model-assisted remote sensing indicator calculation method as described in any one of claims 1-8.

10. A non-volatile computer storage medium storing computer executable instructions, characterized in that: When the computer executable instructions are executed, a large language model-assisted remote sensing indicator calculation method as described in any one of claims 1 to 8 is implemented.

Images