An Automatic InSAR Data Processing Method Based on Natural Language Interaction

Abstract

This application discloses an automatic InSAR data processing method based on natural language interaction, belonging to the field of Synthetic Aperture Radar Interferometry (InSAR) data processing technology. This method receives natural language requests input by the user, uses a large language model to generate intent analysis results and control commands, and then the system selects the target processing path from multiple preset processing paths based on these commands, entering multiple rounds of reasoning-action loops to dynamically generate a structured processing configuration. During processing, the system sequentially calls InSAR processing tools according to the control commands, verifies the preconditions and execution results of each processing step, and automatically advances the subsequent processing flow through a state update mechanism until the final processing result is output. This method achieves end-to-end automated processing from natural language requests to InSAR processing results, lowering the barrier to entry for InSAR data processing and improving the automation and reproducibility of the processing flow.

Description

Technical Field

[0001] This application relates to the field of synthetic aperture radar interferometry (InSAR) data processing technology, and in particular to an automatic InSAR data processing method based on natural language interaction. Background Technology

[0002] Synthetic Aperture Radar Interferometry (InSAR) is an active remote sensing technology that uses radar phase difference to acquire information on land surface deformation. It is widely used in areas such as land subsidence monitoring, crustal movement analysis, and geological hazard assessment. Small Baseline Set Time Series InSAR (SBAS-InSAR) has become the mainstream land surface deformation monitoring technology by constructing a redundant network composed of small spatial baselines and temporal baseline interferometric pairs and using the least squares method to invert the deformation time series. Existing time series InSAR data processing typically involves multiple steps, including data preparation, interferogram generation, phase unwrapping, error correction, and time series inversion. It involves complex parameter settings and algorithm selection, requiring a high level of expertise from the user.

[0003] Current InSAR processing methods mainly fall into two categories: script-based open-source processing tools and commercial graphical interface software. The former relies on users writing or modifying Python scripts and using tools such as PyGMTSAR, ISCE, and SNAP for processing. This approach not only requires users to possess advanced professional knowledge and programming skills but also necessitates manually setting numerous parameters, resulting in a complex processing flow. Furthermore, it lacks the ability to understand natural language requirements and cannot achieve intelligent process management and result retrieval. The latter, such as commercial software like SARscape, while offering graphical interfaces, is costly and still requires users to possess certain professional knowledge and manually configure parameters and download raw data step-by-step. Similarly, it cannot automatically recognize user needs based on natural language and lacks intelligent processing tool integration and semantic retrieval capabilities.

[0004] In summary, existing InSAR processing systems generally suffer from problems such as high operational barriers, insufficient automation, difficulty in integrating multiple processing tools, and lack of intelligent retrieval and interaction capabilities, making it difficult to meet the needs of non-professional users or high-time-sensitivity application scenarios. Summary of the Invention

[0005] This invention addresses the technical problems existing in the background art by proposing an automated InSAR data processing method based on natural language interaction. It constructs an automated InSAR data processing system with natural language as the interactive interface, LLM as the intelligent hub, and domain tools as the execution units. The aim is to transform complex, multi-step professional data processing workflows into intelligent services that users can easily drive through natural language dialogue, thereby achieving end-to-end automation from research ideas to output results.

[0006] To solve the technical problem, the technical solution of the present invention is as follows:

[0007] An automatic InSAR data processing method based on natural language interaction, the method comprising:

[0008] S1: Receive natural language requests from the user;

[0009] S2: Input the natural language request, the available InSAR processing toolset, and the current dialogue history context into the Large Language Model (LLM). The LLM generates intent analysis results, and the system selects the target processing path from multiple preset processing paths based on the intent analysis results.

[0010] S3: In the target processing path, enter a multi-round reasoning-action loop based on the reasoning results generated by the large language model to generate a structured processing configuration;

[0011] S4: Based on the structured processing configuration, according to the control instructions generated by the large language model, dynamically plan and sequentially call subsequent InSAR data processing tools; wherein, before calling each processing tool, verify whether the preconditions corresponding to the processing tool are met;

[0012] S5: After each InSAR data processing tool has completed its execution, the execution results are verified and the system processing status is updated. The processing status is then fed back to the large language model as input.

[0013] S6: When the system meets the preset termination condition according to the processing status, it outputs the final InSAR data processing result.

[0014] Furthermore, the multiple preset processing paths include: a result viewing path, a local InSAR processing path, and an online InSAR processing path; the system selects the target processing path from the multiple preset processing paths according to preset priority rules based on the intent analysis results generated by the large language model.

[0015] When the natural language request contains keywords indicating result viewing, the system selects the result viewing path; when the natural language request contains trigger keywords indicating online processing, the system selects the online InSAR processing path; otherwise, the system selects the local InSAR processing path.

[0016] Furthermore, the multi-round reasoning-action loop continues until the system determines that a complete structured processing configuration has been generated based on preset completion conditions. The multi-round reasoning-action loop adopts a reasoning-action-observation interaction paradigm, wherein the reasoning phase is used to generate control commands, the action phase is used to execute the InSAR processing tool corresponding to the control commands, and the observation phase is used to feed back the tool execution results to the large language model.

[0017] Furthermore, the structured processing configuration includes: study area information, time range information, study target information, and InSAR data processing parameter information, wherein the study area information includes a geographic bounding box or polygon range.

[0018] Furthermore, when generating the structured processing configuration, the system performs range and type checks on the InSAR data processing parameters. When the parameters meet the preset physical constraints, the parameters are written into the hookup structured processing configuration.

[0019] Furthermore, the InSAR data processing tools include: a study scope determination tool, a parameter configuration tool, a data download tool, and an InSAR processing tool; before calling any processing tool, the system verifies whether the preconditions corresponding to that processing tool are met;

[0020] After any processing tool completes its execution, the system updates the current processing status based on the execution result of that tool, and automatically triggers the invocation of subsequent processing tools when the processing status meets the triggering conditions for the next processing stage.

[0021] Furthermore, the system persistently stores the structured processing configuration in the form of a standardized configuration file, and reproduces the corresponding InSAR data processing flow based on the configuration file in subsequent processing.

[0022] Furthermore, when the online InSAR processing path is selected, the system converts the structured processing configuration into cloud processing task parameters, submits them to the remote InSAR processing platform for execution, and obtains the processing results after the cloud processing task is completed.

[0023] Furthermore, when the result viewing path is selected, the system retrieves the processing result file corresponding to the natural language request from the existing processing results based on semantic matching rules, and outputs the matching InSAR result image.

[0024] Furthermore, in the multi-round reasoning-action loop of step S3, each round of the loop includes:

[0025] S301: Input the current natural language request, historical tool execution results and current processing status into the large language model, and the large language model generates control instructions for instructing the next InSAR processing tool and its input parameters to be executed;

[0026] S302: According to the control command, call the corresponding InSAR processing tool and execute it to obtain the tool execution result;

[0027] S303: Verify the validity of the tool's execution result, and feed the verified result back to the large language model as observation information to generate the next round of control instructions.

[0028] This application has the following advantages:

[0029] Significantly lowers the barrier to entry: Through natural language interaction, users do not need to learn complex command-line operations. The system can automatically configure parameters, reducing the reliance on professional knowledge and enabling non-professional users to complete complex InSAR data processing tasks.

[0030] Significantly improves processing efficiency: Automated process organization reduces human intervention, intelligent tool selection avoids invalid operations, and status management ensures the correct execution of processes, thereby freeing users from tedious repetitive tasks and allowing them to focus on the research itself.

[0031] Supports flexible expansion: The modular design makes it easy to add new tools, the standardized interface facilitates the integration of third-party tools, and the configuration-driven mechanism supports different processing strategies, providing a good architectural foundation for the continuous development of the system.

[0032] Enhanced user experience: Through intelligent image matching technology, users can quickly locate processing results through natural language queries, the web interface provides a user-friendly interactive experience, and intelligent error prompts reduce debugging difficulty. Attached Figure Description

[0033] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0034] Figure 1 This application provides a technical roadmap for an automatic InSAR data processing method based on natural language interaction, which is an embodiment of the present application. Detailed Implementation

[0035] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0036] Example 1:

[0037] This application provides an automated InSAR data processing method based on natural language interaction. The InSAR agent is an agent framework for InSAR data processing. The system takes natural language requirements as input, uses a language model to complete intent recognition and tool invocation planning, and ensures that each processing stage meets dependencies and preconditions through execution constraints. It aims to achieve end-to-end automation from research requirements to processing results through natural language interaction. The InSAR agent provides an integrated execution path for InSAR processing tasks: natural language requirements → structured processing configuration → automated processing pipeline → result visualization / retrieval; such as... Figure 1 As shown, it mainly includes:

[0038] (1) Interaction layer;

[0039] Functionality: Provides a conversational UI, handling text responses, form requests, and image display requests, and maintains the conversation (chat history, processing status, form visibility). Input: User text, form field values. Output: Conversation message stream, configuration / search result summary display, and image display components.

[0040] (2) Tool layer;

[0041] Tool 1: Data Range Determiner;

[0042] Tool 2: Parameter Configurator;

[0043] Tool 3: Data Downloader;

[0044] Tool 4: InSAR Processor;

[0045] Tool 5: MintPy Online Processor;

[0046] Tool 6: Image Viewer.

[0047] (3) LLM layer: InSAR agent core orchestrator;

[0048] Functionality: Determines user intent and routes tools based on user input, executes multi-step tool calls and converges to the final output; maintains the dialogue context to support continuous interaction.

[0049] Input: user_input: string (natural language / control commands / status notifications / view requests / online trigger words, etc.). Output: result: string | dict (text results or structured request objects, such as form requests or image display requests).

[0050] Key processing logic: 1. Path selection is based on intent priority: Result viewing > Online processing > Local workflow. 2. Within the selected path, the LLM (Large Language Model) is responsible for planning tool calls (selecting tools and constructing inputs), while execution constraints are responsible for validation and constraint execution. 3. The reasoning-action loop is driven by an "intermediate step list (action, observation)," and tool outputs are used as the basis for subsequent decisions. Dependencies: LLM call interface, tool registry, dialogue memory module. Constraints: When structured outputs such as "form request / image request" appear, the system uses them as part of the interaction protocol and renders them at the interface layer; after the user submits parameters or completes the display, the Agent can continue to receive subsequent inputs and plan the next tool call.

[0051] The InSAR agent system employs an agent architecture based on a large language model to achieve end-to-end automated processing from natural language research needs to InSAR processing results. The system abstracts user input into a unified request object, and through intent classification and routing mechanisms, distributes the requests to different processing paths, ultimately completing data processing and returning the results.

[0052] The system workflow follows a closed-loop control model of "intent recognition → path selection → toolchain execution → result return". Upon receiving user input, the system first performs intent classification, categorizing the request into image viewing, online processing, or local processing requests based on keywords and semantic features in the user input. The system employs a three-tiered priority routing strategy: the highest priority image viewing requests directly enter the result retrieval and display path; the next highest priority online processing requests enter the cloud processing and time-series analysis path; and the lowest priority local processing requests enter the standard processing workflow path.

[0053] Intent Classification and Routing Mechanism: The intent classification module identifies user input intent through keyword matching and semantic analysis. When user input contains both viewing verbs (such as "view," "display," "output," etc.) and image-related nouns (such as "image," "rate," "displacement," "coherence," etc.), the system determines it as an image viewing request and directly enters the image viewing processing path. When user input contains online processing trigger words (such as "mintpy," "online," "hyp3," "cloud processing," etc.), the system determines it as an online processing request and enters the cloud processing path. Other requests are determined as local processing requests and enter the standard processing flow path.

[0054] Image viewing and processing path; the image viewing and processing path serves as a shortcut in the system, used to retrieve and display target images from existing processing results. The execution flow of this path includes the following steps:

[0055] The first step is keyword extraction: The system extracts image type keywords from user input, such as "rate", "displacement", "coherence", "DEM", etc., as well as modifiers, such as "filtered", "original", "cumulative", etc.

[0056] The second step is candidate image scanning: The system automatically scans the output directory and collects all available processed image files to form a candidate image set.

[0057] The third step is similarity matching and ranking: the system scores and ranks candidate images based on filename keyword matching and semantic similarity calculation to determine the image that best matches the user's query.

[0058] The fourth step is result return: The system returns the image object with the highest matching degree, including information such as file path, display name, and matching criteria, which is then handed over to the interface layer for display.

[0059] The local processing workflow path is the core processing path of the system, used to execute the complete InSAR data processing workflow. This path adopts an intelligent planning mechanism based on a large language model, dynamically selecting and executing processing tools through multiple rounds of reasoning-action loops.

[0060] Intelligent planning mechanism: The system maintains a list of execution trajectories, recording the sequence of executed tool calls. Each tool call includes an action and an observation. The system sets a maximum iteration limit (usually 5 times) to prevent infinite loops.

[0061] In each iteration, the system performs the following operations:

[0062] First, the large language model performs reasoning and planning: The system takes the current user input, the history of executed steps, the dialogue memory context, and the list of available tools as input, and calls the large language model for reasoning. The large language model is based on the ReAct (Reasoning and Acting) paradigm. It first performs a thought process, analyzing the current task state and the goals to be achieved, and then decides on the next action, including selecting the tool to be invoked and constructing the tool's input parameters.

[0063] Second, tool execution: Based on the decisions of the large language model, the system searches for the corresponding tool function in the tool registry and dynamically calls the tool, passing in the constructed input parameters. After the tool is executed, it returns observation results, including execution status, output data, error messages, etc.

[0064] Third, execution constraint verification: The system performs precondition verification on the tool execution results to ensure that the results meet the expected format and constraints. For the study scope determination tool, the verification results must contain valid regional and temporal information; for the parameter configuration tool, the verification parameter values ​​must be within a reasonable range (e.g., spatial baseline threshold greater than 0, coherence threshold between 0 and 1, and thread count greater than or equal to 1); for the data download tool, the verification configuration status must be "parameters ready," and the download results must contain valid files; for the InSAR processing tool, the verification data status must be "data ready."

[0065] Fourth, execution trajectory recording: The system records the tool call actions and execution results in the execution trajectory list, providing historical information for the next round of inference.

[0066] Fifth, automatic continuation determination: The system checks the current tool execution result to determine whether it needs to automatically continue to the next processing stage. When the research scope determination tool completes and returns "configuration saved" information, the system automatically generates a "configuration parameters" status notification as input for the next round of large language model inference; when the data download tool completes and returns "download completed" information, the system automatically generates a "data download completed" status notification, triggering subsequent processing flows.

[0067] Toolchain execution flow; the local processing flow path contains four core tools, which are executed sequentially according to logical order:

[0068] Tool 1: Data Range Determiner. This tool extracts key information from the user-input natural language research description, including the research area (converting place names into precise geographic boundaries through geocoding, or parsing user-provided coordinate ranges, POLYGON format, etc.), time range (intelligently parsing relative time expressions such as "last two years" or absolute time expressions such as "YYYY-MM-DD" format date ranges), and research objectives (identifying research types such as "subsidence" and "deformation" through keyword matching). The tool outputs a structured research configuration object, containing core fields such as spatial range, time range, and objective type, and persistently saves the configuration information as a JSON file.

[0069] Tool 2: Parameter Configurator. This tool generates or modifies data search and processing parameters based on research needs. Data search parameters include polarization mode (VV / VH / HH / HV / auto) and orbital orientation (A ascending / D descending / auto); data processing parameters include spatial baseline threshold, coherence threshold, number of parallel threads, and multi-look filtering window. The tool supports two working modes: when the user explicitly requests parameter configuration, a form request object is returned, and the interface layer collects the user input before executing the configuration; when the user chooses to use default values, the system's preset default parameters are directly applied. The tool performs rigorous type and range validation on key parameters to ensure they conform to the physical constraints of InSAR processing.

[0070] Tool 3: Data Downloader. This tool automatically searches for suitable Sentinel-1 data based on the generated study and parameter configurations, and downloads data files from data sources such as ASF. Before execution, the tool verifies that the configuration status must be "Parameters Ready," and after execution, it verifies that the download results must contain valid files. After the download is complete, the tool updates the configuration status to "Data Downloaded" and returns information such as the download status, data list, and file paths.

[0071] Tool 4: InSAR Processor. This tool, assuming the data is ready, executes the complete SBAS-InSAR processing workflow, including core steps such as interferogram generation, phase unwrapping, time-series inversion, and result generation. Before execution, the tool verifies that the data status must be "data ready." During execution, it performs error detection and reporting for missing data or empty scene sets. After processing, the tool outputs a summary of the processing results, including the locations of key products and a summary of the processing log.

[0072] State-driven process control: The system employs a state-driven mechanism to automate the process. Before each tool is executed, the system performs rigorous precondition checks to ensure that execution conditions are met. After a tool completes its execution, the system automatically triggers the next stage. When the research scope determination tool is complete, the system automatically generates a "configuration parameters" status notification, triggering the execution of the parameter configuration tool. When the parameter configuration tool completes, the system automatically generates a "parameter configuration complete" status notification, triggering the execution of the data download tool. When the data download tool completes, the system automatically generates a "data download complete" status notification, triggering the execution of the InSAR processing tool. This state-driven mechanism ensures that the processing flow is executed automatically in the correct order, reducing manual intervention.

[0073] Online processing workflow path; the online processing workflow path is used to perform online InSAR processing based on the HyP3 cloud platform and MintPy time series analysis. When the system detects that the user input contains online processing trigger words, it automatically enters this path.

[0074] The execution flow of this path includes the following steps:

[0075] The first step is parameter acquisition: In a web interaction scenario, the system first returns a form request object, and the interface layer collects the parameters input by the user, including project name, time range, research area boundary, orbit number, reference point coordinates, tropospheric correction method, etc.

[0076] The second step is cloud task submission: Based on the collected parameters, the system automatically searches for Sentinel-1 data that meets the requirements and submits the InSAR processing task to the HyP3 cloud platform.

[0077] The third step is task status monitoring: the system monitors the processing status of tasks in the cloud and waits for the tasks to complete.

[0078] Step 4, Result Download: Once the cloud task is completed, the system will automatically download the processing result file.

[0079] Step 5, Time Series Analysis: The system uses MintPy to perform time series analysis on the downloaded processing results, generating analysis results such as deformation rate plots and time series deformation plots.

[0080] Step 6, Results Return: The system returns a summary of the processing results and key product location information.

[0081] Execution constraints and security mechanisms: The system has established a comprehensive execution constraint mechanism to ensure the correct execution of the processing flow and the reliability of the results. Execution constraints include three types:

[0082] First, timing constraints: ensuring that processing steps are executed in the correct logical order. The scope of the study must be determined before parameter configuration, parameter configuration must be completed before data download, and data download must be completed before InSAR processing. The system uses a precondition check mechanism to ensure that the preconditions for each tool are completed and meet the execution conditions before execution.

[0083] Second, parameter constraints: Ensure that all parameter values ​​conform to physical meaning and scientific standards. The spatial baseline threshold must be greater than 0, the coherence threshold must be between 0 and 1, the temporal baseline limit must be greater than 0, and the number of threads must be greater than or equal to 1. The system performs strict type and range validation on all parameters in the parameter configuration tool, rejecting parameters that do not meet the constraints and returning an error message.

[0084] Third, data constraints: ensuring that data dependencies are met. Before the data download tool executes, it must verify that the configuration status is "parameters ready"; before the InSAR processing tool executes, it must verify that the data status is "data ready"; before the image viewing tool executes, it must verify that the processing result directory exists and contains valid files. The system uses a status verification mechanism to ensure that the required data and configuration are ready when each tool executes.

[0085] Dialogue memory and context management: The system maintains a dialogue memory module to record the user's interaction history with the system, including user input, system responses, tool call sequences, and other information. Dialogue memory provides contextual information for subsequent interactions, enabling the system to understand the user's subsequent requests and supporting multi-turn dialogues and context-sensitive processing.

[0086] In each round of reasoning using the large language model, the system incorporates dialogue memory as part of the input, enabling the large language model to reason and make decisions based on historical interaction information. For example, when a user says "continue processing," the system can understand the specific meaning of "continue" based on dialogue memory and automatically select the tool to be executed next.

[0087] Error Handling and Exception Recovery: The system has established a comprehensive error handling mechanism, encompassing three levels: input validation, workflow protection, and exception recovery. For input validation, the system checks the validity of user input formats, verifies the correctness of parameter ranges and types, and provides clear error messages to help users quickly locate and correct input problems. Regarding workflow protection, the system performs precondition checks before each critical step, handles runtime errors through an exception capture mechanism, and restores the system to a safe state when an error occurs through a state rollback mechanism, ensuring stable system operation. For exception recovery, the system provides clear error messages explaining the problem, offers specific repair suggestions to guide users in resolving the issue, and uses progress indicators to keep users informed of the processing status, improving the user experience.

[0088] The system ensures persistent and reproducible configuration. Through an explicit configuration mechanism, all processing parameters and configuration information are stored in a standardized JSON format, ensuring traceability and reproducibility of the processing. The configuration file includes research area information (area name, geographical boundary, area description), time range information (start date, end date, time description), research objective information (research type, research objective description), processing parameter information (polarization mode, orbital orientation, spatial baseline threshold, coherence threshold, etc.), and configuration metadata (configuration status, creation time, etc.).

[0089] The system automatically saves the configuration file every time there is a configuration change, ensuring real-time updates and persistence of the configuration. During configuration loading, the system performs rigorous parameter validation, including parameter range checks, dependency checks, and format validation, to ensure the validity of the configuration. Through the configuration persistence mechanism, users can completely reproduce previous processing steps, ensuring the verifiability of scientific research.

[0090] This application is based on a large language model-based intelligent planning mechanism. The system is the first to apply a large language model to the field of InSAR data processing. It realizes intelligent reasoning and tool selection through the ReAct paradigm, enabling the system to dynamically determine the next tool and input parameters to be executed based on user needs and the current state, thus achieving truly intelligent processing flow organization.

[0091] This application is based on intent-driven multi-path adaptive routing. The system achieves the ability to automatically select the processing path according to the user's intent through priority rules and intent classification mechanism. It supports multiple processing modes such as image viewing, online processing, and local processing to meet the needs of different users.

[0092] This application is based on state-driven automated process advancement. The system achieves automatic process advancement through a status notification mechanism. When a tool completes its execution, the system automatically generates a status notification, triggering the execution of the next stage, reducing manual intervention and improving processing efficiency.

[0093] Based on execution constraints and security mechanisms, this application ensures the correct execution of the processing flow and the reliability of the results through various constraint mechanisms such as precondition checks, parameter verification, and data dependency checks, thus avoiding processing failures caused by parameter errors or missing data.

[0094] Based on explicit configuration expression and reproducibility assurance, this application uses a standardized JSON configuration file to fully record all processing parameters and configuration information, achieving traceability and reproducibility of the processing process and providing assurance for the verifiability of scientific research.

[0095] In summary, the workflow of the InSAR intelligent agent system can be summarized as follows: After the user inputs their natural language research needs, the system first classifies the intent and selects a processing path based on the user's intent. For image viewing requests, the system directly enters the image retrieval and display path; for online processing requests, the system enters the cloud processing path; for local processing requests, the system enters the standard processing workflow path. In the standard processing workflow path, the system dynamically selects and executes processing tools through the intelligent planning mechanism of the large language model, including steps such as determining the research scope, configuring parameters, downloading data, and InSAR processing. Before each tool is executed, the system performs precondition checks; after execution, the system verifies the results and automatically triggers the next stage of execution. The entire processing process is guaranteed by an execution constraint mechanism to ensure the correct execution of the process and the reliability of the results. All configuration information is persistently saved in a standardized JSON format, achieving traceability and reproducibility of the processing process.

[0096] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0097] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. An automatic processing method for InSAR data based on natural language interaction, characterized in that, The method includes: S1: Receive natural language requests from the user; S2: Input the natural language request, the available InSAR processing toolset, and the current dialogue history context into the Large Language Model (LLM). The LLM generates intent analysis results, and the system selects the target processing path from multiple preset processing paths based on the intent analysis results. S3: In the target processing path, enter a multi-round reasoning-action loop based on the reasoning results generated by the large language model to generate a structured processing configuration; S4: Based on the structured processing configuration, according to the control instructions generated by the large language model, dynamically plan and sequentially call subsequent InSAR data processing tools; wherein, before calling each processing tool, verify whether the preconditions corresponding to the processing tool are met; S5: After each InSAR data processing tool has completed its execution, the execution results are verified and the system processing status is updated. The processing status is then fed back to the large language model as input. S6: When the system meets the preset termination condition according to the processing status, it outputs the final InSAR data processing result.

2. The automatic InSAR data processing method based on natural language interaction according to claim 1, characterized in that, The various preset processing paths include: result viewing path, local InSAR processing path, and online InSAR processing path; the system selects the target processing path from the various preset processing paths according to the intent analysis results generated by the large language model and according to preset priority rules. When the natural language request contains keywords indicating result viewing, the system selects the result viewing path; when the natural language request contains trigger keywords indicating online processing, the system selects the online InSAR processing path; otherwise, the system selects the local InSAR processing path.

3. The automatic InSAR data processing method based on natural language interaction according to claim 1, characterized in that, The multi-round reasoning-action loop continues until the system determines that a complete structured processing configuration has been generated based on preset completion conditions. The multi-round reasoning-action loop adopts a reasoning-action-observation interaction paradigm, wherein the reasoning phase is used to generate control commands, the action phase is used to execute the InSAR processing tool corresponding to the control commands, and the observation phase is used to feed back the tool execution results to the large language model.

4. The automatic InSAR data processing method based on natural language interaction according to claim 3, characterized in that, The structured processing configuration includes: study area information, time range information, study target information, and InSAR data processing parameter information, wherein the study area information includes a geographic bounding box or polygon range.

5. The automatic InSAR data processing method based on natural language interaction according to claim 4, characterized in that, When generating the structured processing configuration, the system performs range and type checks on the InSAR data processing parameters. When the parameters meet the preset physical constraints, the parameters are written into the hookup structured processing configuration.

6. The automatic InSAR data processing method based on natural language interaction according to claim 1, characterized in that, The InSAR data processing tools include: a study scope determination tool, a parameter configuration tool, a data download tool, and an InSAR processing tool; before calling any processing tool, the system verifies whether the preconditions corresponding to that processing tool are met; After any processing tool completes its execution, the system updates the current processing status based on the execution result of that tool, and automatically triggers the invocation of subsequent processing tools when the processing status meets the triggering conditions for the next processing stage.

7. The automatic InSAR data processing method based on natural language interaction according to claim 1, characterized in that, The system persistently stores the structured processing configuration in the form of a standardized configuration file, and reproduces the corresponding InSAR data processing flow based on the configuration file in subsequent processing.

8. The automatic InSAR data processing method based on natural language interaction according to claim 2, characterized in that, When the online InSAR processing path is selected, the system converts the structured processing configuration into cloud processing task parameters, submits them to the remote InSAR processing platform for execution, and obtains the processing results after the cloud processing task is completed.

9. The automatic InSAR data processing method based on natural language interaction according to claim 2, characterized in that, When the result viewing path is selected, the system retrieves the processing result file corresponding to the natural language request from the existing processing results based on semantic matching rules, and outputs the matching InSAR result image.

10. The automatic InSAR data processing method based on natural language interaction according to claim 1, characterized in that, In the multi-round reasoning-action loop of step S3, each round includes: S301: Input the current natural language request, historical tool execution results and current processing status into the large language model, and the large language model generates control instructions for instructing the next InSAR processing tool and its input parameters to be executed; S302: According to the control command, call the corresponding InSAR processing tool and execute it to obtain the tool execution result; S303: Verify the validity of the tool's execution result, and feed the verified result back to the large language model as observation information to generate the next round of control instructions.

Images