Agricultural insurance data acquisition and management system based on satellite remote sensing

Abstract

The invention discloses an agricultural insurance data acquisition and management system based on satellite remote sensing. The system comprises a data acquisition module, a preprocessing module, a verification module and a management module. The data acquisition module covers basic, universal and vector data acquisition according to data types; the preprocessing module improves the quality of collected data; the verification module verifies a cultivated land vector and insured user information, and data accuracy and reliability are enhanced; the management module builds two libraries, and constructs an agricultural insurance data large model by utilizing knowledge graph modeling. In the technical application, on the basis of satellite remote sensing and A I, large-user plot data is collected by means of an image segmentation algorithm, and the cost is reduced. A full-coverage model is constructed through abstract materialization, and business application is supported. Technical means such as satellite remote sensing, weight dispatch data superposition and AI image recognition are used for verifying data, manual operation is reduced, verification accuracy is guaranteed, data value is mined, and the agricultural insurance data management level is improved.

Description

Technical Field

[0001] This invention belongs to the field of agricultural insurance technology, specifically an agricultural insurance data collection and management system based on satellite remote sensing. Background Technology

[0002] Agricultural insurance is an important means of supporting agricultural development in my country, providing trillions of yuan in risk protection to hundreds of millions of farmers annually. It is typically divided into crop insurance, livestock insurance, and other insurance based on different production types. Such a massive agricultural insurance business naturally generates a wealth of data; the diverse types and large volume of data necessitate the use of information technology and other scientific methods to collect and manage this data, thereby supporting the operation of higher-level agricultural insurance businesses.

[0003] Existing agricultural insurance data collection mainly includes basic data such as remote sensing base maps, cultivated land vectors, and insured information, as well as various insurance business data such as insurance policies and claims. At the same time, monitoring and analysis data related to insurance business, such as crop growth, crop diseases, and yield forecasts, are also necessary data support for the business.

[0004] Because agricultural insurance is a complex business with strong interrelationships between its operations, there is also a complex relationship between the business and the data. Only by efficiently collecting and effectively managing this data, and then deeply mining and applying it to agricultural insurance, can we maximize its value beyond the data itself.

[0005] Existing technical solutions typically use drone aerial imagery as the base imagery. Data acquisition methods are divided into two parts: field collection and data processing. Field collection utilizes GIS positioning software to locate farmers' plots on-site, or uses a mobile app to record the spatial location of the plots. Subsequently, professional GIS personnel, referring to the drone imagery and spatial positioning information, complete the delineation of plot boundaries and the entry and binding of corresponding insured information in the data processing. How to quickly and efficiently complete farmland data collection is a key challenge that current technical solutions urgently need to overcome.

[0006] Existing technical solutions typically manage data using databases, including relational databases, non-relational databases, and spatial databases. Data sources are diverse, including basic imagery data, farmland vector data, insured information data, data from various business operations, crop monitoring and analysis data, and IoT data. The data sources are extensive, and the data formats are numerous. How to effectively aggregate massive amounts of multi-source data and form an organic whole for better application in insurance business is a key research issue in existing technical solutions.

[0007] Existing technical solutions typically focus on optimizing insurance business services, with extensive research conducted on end-stage business methods and systems such as underwriting, product verification, and claims settlement. However, basic data collection and management are the foundational support for backend operations. Optimizing the entire agricultural insurance data collection, management system, and methods to expand the service scenarios and improve service quality thresholds is a weakness of existing technical solutions.

[0008] The patent with publication number CN118822746A, titled "Agricultural Insurance Claims Judgment Method and System Based on Big Data," describes how to use big data to collect environmental, meteorological, and crop growth data from multiple days in the planting area. The system then uses vector encoding, correlation analysis, image feature analysis, and feature embedding analysis to estimate crop yield and determine whether an accident has occurred.

[0009] The patent with publication number CN115187413A, entitled "A Method for Agricultural Insurance Underwriting Based on Land Title Confirmation Information," describes a process where features extracted from graphic data are fed back to the land title confirmation system to confirm title information. Then, the system is connected to the agricultural insurance system, and the title information is entered into the user's insurance list. Once the verification is successful, the insurance entry is completed. The agricultural insurance system automatically generates a list of insured land title confirmations for each household and assigns values ​​to the content, thus completing the agricultural insurance underwriting.

[0010] The patent with publication number CN118470550A, entitled "A Method and Platform for Acquiring Natural Resource Asset Data",...

[0011] Based on UAV or satellite imagery, vector element data, including points, lines, and polygons, is generated manually using specialized software, and related attribute description information is recorded for subsequent continuous data monitoring, management, and analysis.

[0012] The patent CN114610829A, titled "A Land Information Management Method Based on Smart Terminal and Remote Sensing Intelligent Identification," describes a method where a user walks around with a mobile terminal equipped with the aforementioned app to collect boundary information of a land parcel and generate the first measured area information of the land parcel. The app acquires the location information of the mobile terminal in real time. Through the attribute collection and editing functions of the app, information such as the insured party information, land transfer information, management information, and administrative division information of the land parcel can be edited.

[0013] The patent CN117057936A, titled "An Agricultural Insurance Underwriting System," describes a system that collects required policy information via a data acquisition module when a user applies for insurance. This spatial information is then integrated with and imported into a geographic information system (GIS) map to generate target information with spatial data, ultimately leading to electronic application forms and electronic policies. This system requires minimal maintenance modules and backend support, accurately mapping the customer's target information without being affected by map data updates. Summary of the Invention

[0014] The purpose of this invention is to provide an agricultural insurance data acquisition and management system based on satellite remote sensing, in order to solve the problems in the prior art mentioned in the background, such as high data acquisition efficiency and cost with incomplete coverage, high barriers to data storage and management, inability to organically integrate and utilize multi-source data, and difficulty in ensuring accuracy due to lack of data verification.

[0015] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0016] An agricultural insurance data acquisition and management system based on satellite remote sensing includes a data acquisition module, a data preprocessing module, a data verification module, and a data management module.

[0017] The data acquisition module is designed for different data types and includes basic data acquisition, general data uploading, and vector data acquisition.

[0018] The data preprocessing module performs preprocessing operations on the basic data and vector data collected by the data acquisition module to improve data quality.

[0019] The data verification module improves the accuracy and reliability of basic data by verifying cultivated land vector data and insured household information data;

[0020] The data management module builds a large-scale agricultural insurance data model by establishing a basic database and a business database, and mapping business objects through knowledge graph modeling.

[0021] According to the above technical solution, the basic data acquisition module acquires remote sensing image data of the agricultural insurance coverage area by using remote sensing satellites, including but not limited to high-resolution images and multispectral images.

[0022] The general data upload of the data collection module includes insured household information data, insurance company data, basic data, and business forms; among them, insured household information data includes farmer name, insurance information, and planting information; insurance company data includes company information, salesperson information, and assistant information; basic data includes insurance type information, subsidy information, and administrative division; business forms include insurance application form, household list, and claim form;

[0023] The vector data acquisition module includes the spatial location and boundary vector data of the insured's land plots. Using satellite remote sensing imagery as the base map and combining it with intelligent boundary extraction algorithms, the module completes the land plot data acquisition for the insured. The acquired land plot data is then combined with the insured's basic information to establish the human-land relationship within the agricultural insurance data system.

[0024] Based on the above technical solution, the intelligent boundary extraction algorithm is as follows:

[0025] First, the remote sensing image data collected by the data acquisition module is locked and captured. The captured remote sensing image data is then sent to the image encoder deployed on the backend server to interpret the image and obtain the feature map of the remote sensing image data.

[0026] The image encoder acquires the location information from remote sensing image data and feeds it back to the prompt encoder, which converts the coordinates into feature vectors.

[0027] The intermediate decoder combines the feature vector with the feature map, inputs it into the mask decoder to obtain the returned result set, scales and maps it according to the original image size, and finally outputs the result with the highest score, which is the extracted land data of the insured.

[0028] According to the above technical solution, the data preprocessing module performs preprocessing operations on the collected data, including geometric correction, radiometric correction, noise reduction, cropping and mosaicking, color balancing and format conversion of remote sensing image data, to improve data quality and provide a single image to support all modules of the system.

[0029] According to the above technical solution, geometric correction: eliminates the geometric distortion of the image, matches the pixel position with the real geographical location on the Earth's surface, and thus gives the image clear geographical reference coordinates;

[0030] Radiometric correction: used to eliminate sensor errors and atmospheric interference, and restore the true radiation or reflection characteristics of ground objects;

[0031] Noise filtering: Eliminates noise and artifacts in images, improving image clarity and interpretability;

[0032] Cutting and inlaying: Adjusting the scope and seamlessly splicing data according to research needs;

[0033] Color balancing: Eliminates color differences between multiple images to ensure overall consistency;

[0034] Data format conversion: Converting data into a standard data format to facilitate subsequent analysis and processing.

[0035] According to the above technical solution, the data preprocessing module also combines Geographic Information System (GIS) and Artificial Intelligence (AI) to extract farmland boundaries, classify and identify crops, assess growth status, and analyze disasters from the collected remote sensing image data, providing data support for subsequent routine monitoring.

[0036] According to the above technical solution, the data verification module includes verification of cultivated land vector data for large-scale farmers and verification of data for small-scale farmers.

[0037] Specifically, the verification of large account data includes:

[0038] After the boundary data of large-scale farmland is identified through on-site communication and intelligent boundary drawing, it is compared with the crop classification data in the system based on satellite remote sensing image preprocessing. If the insurance information of the large-scale farmer matches the system data, that is, the type of the insured crop is the same as the crop identified in the system preprocessing, and the difference between the drawn large-scale farmer's plot area and the crop area identified by the system is within the specified error range, then the verification is considered successful; otherwise, on-site business personnel need to manually confirm and complete the verification.

[0039] The specific steps for verifying retail investor data are as follows:

[0040] First, a circular buffer zone of 100m is drawn from the spatial point P where the individual investor takes photos, and the vector range Pi is obtained. The formula is Pi = (P:d(Pi,Oi)≤100). Then, using the GIS overlay analysis capability, the Pi vector and the weighted vector data are overlaid and analyzed. If the weighted vector patch of the individual investor's ownership exists in the overlay analysis result, the verification is passed.

[0041] Then, AI image recognition technology is used to identify the crop type in the photos taken by individual farmers and compare and verify it with the crop information insured by those farmers.

[0042] According to the above technical solution, the construction of the basic database in the data management module is specifically as follows:

[0043] Based on the different data formats stored, basic databases are divided into relational databases, spatial databases, and file-based databases, which can be abstracted into three databases: human-land, insurance, and remote sensing.

[0044] Among them, the human-land database is used to store data on core participants in agricultural insurance and data on insurable arable land planted by farmers. These two are established through a strong one-to-many relationship, which forms the basis of the "human-land relationship" and supports the entire agricultural insurance data system.

[0045] The insurance database is used to store business data for the entire agricultural insurance process, including data, forms, and various attachments during the insurance application, inspection, claims, and settlement processes, in order to meet the application needs of agricultural insurance business.

[0046] The remote sensing database is used to store basic satellite remote sensing images as well as preprocessed satellite remote sensing monitoring and analysis data on farmland boundaries, crop classification, growth status assessment, and disaster analysis, providing data support for high-quality agricultural insurance services.

[0047] Based on the above technical solution, the construction of the business database in the data management module is specifically as follows:

[0048] A business database is a data model that takes the data physically stored in the basic database, abstracts it into entity objects, models it using knowledge graphs, and maps it to business objects. This model is then stored in the computer as a business logic data model. This includes data on human-land relationships, target crops, insurance events, and situation monitoring.

[0049] Based on the above technical solution, the specific large-scale model of agricultural insurance data in the data management module is as follows:

[0050] The concept of agricultural insurance involves understanding all specific things and businesses, and classifying and abstracting related objects with common characteristics and strong business connections to form a conceptual entity model of five categories: people, places, things, and events.

[0051] The entity object model is then associated, decomposed, and refined, and mapped to four major categories of business objects: human-land relationships, target crops, insurance events, and situation monitoring, thereby supporting the overall operation of agricultural insurance.

[0052] Compared with the prior art, the present invention has the following beneficial effects:

[0053] In this invention, satellite remote sensing and AI are used as technical support. Data on large plots of land are collected through image segmentation algorithms to reduce technical difficulty and cost. Real-world objects are abstracted and materialized to form "people, land, things, events" entity objects. Relationships between entities are constructed and mapped to four major business objects to achieve full-coverage modeling of users, space, business, and time, providing support for business applications.

[0054] By leveraging knowledge graph technology, a basic database and business database are established to form a big data system covering the entire physical space and all business scopes, thereby enhancing basic data management capabilities and unlocking data value. Addressing the different data collection characteristics of large-scale and small-scale farmers, three technical methods are employed: satellite remote sensing crop type monitoring and comparison, weighted data overlay analysis, and AI-powered intelligent crop recognition, to reduce manual operations and ensure the accuracy of verification results. Attached Figure Description

[0055] Figure 1 This is a schematic diagram of the farmland boundary extraction results of the present invention;

[0056] Figure 2 This is a schematic diagram of the data collection results from large-scale farmland in this invention;

[0057] Figure 3 This is a schematic diagram of the data collection results from individual farmers' cultivated land in this invention;

[0058] Figure 4 This is a schematic diagram of the model design of the present invention;

[0059] Figure 5 This is a schematic diagram of the knowledge graph structure of the present invention;

[0060] Figure 6 This is a key flowchart of the system of the present invention;

[0061] Figure 7 This is a system framework diagram of the present invention. Detailed Implementation

[0062] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0063] Example 1

[0064] like Figure 7 As shown, an agricultural insurance data acquisition and management system based on satellite remote sensing includes a data acquisition module, a data preprocessing module, a data verification module, a data management module, and an application service module.

[0065] The data acquisition module is designed for different data types and includes basic data acquisition, general data uploading, and vector data acquisition.

[0066] The data preprocessing module performs preprocessing operations on the basic data and vector data collected by the data acquisition module to improve data quality.

[0067] The data verification module improves the accuracy and reliability of basic data by verifying cultivated land vector data and insured household information data;

[0068] The data management module builds a large-scale agricultural insurance data model by establishing a basic database and a business database, and mapping business objects through knowledge graph modeling.

[0069] The system in this invention implements the construction of two databases: a database and a business database. Targeting the entire lifecycle of agricultural insurance business, it abstracts real-world entities into five core entities: "people, land, things, and events." Furthermore, through knowledge graph modeling, it forms a large-scale agricultural insurance data model integrating business objects such as "people-land relationships," "target crops," "insurance events," and "situation monitoring," to support routine data asset accumulation and continuous business application services. The system includes a data acquisition module, a data preprocessing module, a data verification module, a data management module, an application service module, and a system security module. (e.g.) Figure 6 (As shown)

[0070] I. Data Acquisition Module

[0071] The data acquisition module provides various capabilities for different data types, including basic data acquisition, general data uploading, and vector data acquisition.

[0072] 1. The basic data is obtained by technology companies using remote sensing satellites to acquire remote sensing image data of the agricultural insurance coverage area, including but not limited to high-resolution imagery and multispectral imagery.

[0073] 2. General data upload includes farmer information (such as farmer name, insurance information, planting information, etc.), insurance company data (such as company information, salesperson information, and insurance agent information, etc.), basic data (including insurance type information, subsidy information, administrative divisions, etc.), and business forms (including insurance application forms, household lists, and claim forms, etc.). The system supports uploading and entering data in various formats, including strings, timestamps, numbers, images, files, and compressed packages. The system also provides batch upload templates, allowing users to organize and upload data in batches, supporting up to 20,000 data entries at a time.

[0074] 3. Vector data acquisition focuses on the core spatial location and boundary vector data of insured property plots. Using satellite remote sensing imagery as the base map and combining it with intelligent boundary extraction algorithms, two methods are provided for land plot boundary acquisition: PC and mobile, catering to both large and small-scale insured properties. The operation procedures are as follows:

[0075] A. Acquisition of large-scale land parcels:

[0076] First, high-resolution satellite imagery is acquired and preprocessed accordingly before being integrated into the system. Then, insurance company agents upload relevant information about major policyholders to the system. The system automatically creates land parcel mapping tasks, and agents can distribute the data collection tasks to field personnel based on the region and personnel distribution.

[0077] After receiving the task, the field staff will confirm the location and time of the centralized operation with the large households through grassroots organizations such as town governments and village committees, and carry out the centralized operation using PCs. The operation site is usually a location with PC working conditions, such as a town government or village committee.

[0078] The process begins with large-scale landowners identifying the spatial location of their plots using high-resolution satellite imagery. Field personnel then confirm this with the landowners on-site and delineate the boundaries of the insured plots at the corresponding locations. The boundaries of cultivated land can be intelligently extracted through simple interactive operations, improving operational efficiency and reducing costs. (e.g.) Figure 2 (As shown)

[0079] Specifically, the intelligent boundary extraction algorithm mentioned in this invention adopts the existing SAM image segmentation model and utilizes information integration to transform user operations into positive and negative feedback inputs for the algorithm. After calculation by the SAM model, the results are converted into vector outputs, completing the entire boundary extraction process. The specific calculation steps of the model include:

[0080] a. First, the image currently displayed on the screen is locked through the front-end page, and the captured image is sent to the image encoder deployed on the back-end to decode the image and obtain the feature map of the image;

[0081] b. The front-end page obtains the screen position information of the user's click (including positive and negative samples) and feeds it back to the prompt encoder, which converts the coordinates into feature vectors;

[0082] c. The front-end deployed decoder combines the feature vector with the feature map returned from the back end, inputs it into the mask decoder to obtain the returned result set (i.e., obtain), scales and maps it according to the original image size, and finally outputs the result with the highest score.

[0083] B. Acquisition of land parcels from individual buyers:

[0084] like Figure 3 As shown, high-resolution satellite images are first acquired and preprocessed with corresponding geometric correction, radiometric correction, denoising, cropping and mosaicking, color balancing, and format conversion before being integrated into the system. Also integrated into the system is farmland boundary data intelligently extracted through image segmentation and boundary extraction algorithms. Then, agents from insurance companies or third-party companies upload relevant information of individual insured households by village to the system. The system automatically creates land parcel drawing tasks and automatically distributes them to the corresponding individual insured households.

[0085] Individual farmers, within the designated task time period, access the mobile system and take photos of their insured plots as required. The photos and corresponding location information are automatically transmitted back to the system and automatically overlaid and bound to the system's built-in farmland boundary data. It should be noted that due to limitations in rural development conditions in my country, the data collection work can be completed by grassroots insurance coordinators or agricultural workers. Furthermore, the verification and confirmation process after the data is generated is described in detail in the relevant section of this invention. Ultimately, by combining the basic information of large-scale and individual farmers, the core "human-land relationship" foundation of the agricultural insurance data system (such as...) is established. Figure 4 (As shown).

[0086] II. Data Preprocessing Module

[0087] The preprocessing module performs basic preprocessing operations on the acquired remote sensing image data, including geometric correction, radiometric correction, denoising, cropping and mosaicking, color balancing, and format conversion, to improve data quality and provide image data support for various modules of the system.

[0088] Geometric correction: Eliminates geometric distortion of the image and matches pixel positions with the actual geographical locations on the Earth's surface, thereby giving the image clear geographic reference coordinates.

[0089] Radiometric correction: used to eliminate sensor errors and atmospheric interference, and restore the true radiation or reflection characteristics of ground objects.

[0090] Noise filtering: Eliminates noise and artifacts in images, improving image clarity and interpretability.

[0091] Cutting and tiling: Adjusting the scope and stitching seamless data according to research needs.

[0092] Color uniformity: Eliminates color differences between multiple images to ensure overall consistency.

[0093] Data format conversion: Converting data into a standard data format to facilitate subsequent analysis and processing.

[0094] The preprocessing module also integrates Geographic Information System (GIS) and Artificial Intelligence (AI) technologies to process and analyze the collected remote sensing data, including farmland boundary extraction, crop classification and identification, growth status assessment, and disaster analysis, providing data support for subsequent routine monitoring. Among these:

[0095] Farmland boundary extraction employs a deep learning-based semantic segmentation and edge detection model, which enhances both the accuracy of farmland target segmentation and the usability of the boundary extraction results (e.g., Figure 1 As shown). Specifically:

[0096] The first step is to use an Encoder-Decoder structured semantic segmentation model to extract the extent of the fields and obtain the field extent segmentation results.

[0097] The second step, in boundary post-processing, employs adaptive thresholding and skeletalization methods to obtain collinear field boundaries. Closed boundaries are then obtained through breakpoint connection and line extension. Finally, the closed boundaries are converted into vector surfaces and masked using the extracted field extent results. This process ultimately yields more accurate field boundaries that are collinear and separated.

[0098] Crop classification and identification employs a deep learning crop classification model based on multi-temporal imagery. This model integrates RGB texture and spectral NDVI (Normalized Difference Vegetation Index) from the images to construct a time series of crop growth, thereby analyzing its growth cycle and variation patterns to achieve accurate identification of crop species. Specifically:

[0099] First, using multi-temporal high-resolution remote sensing imagery and UAV-sampled imagery, various remote sensing indices were calculated, and RGB (red, green, and blue) visible light bands and remote sensing spectral index bands were fused to construct a deep learning crop classification model based on a ResNet backbone network. Second, through visual interpretation and field surveys, prior knowledge of the category attributes of image features in certain sample areas of the remote sensing images was obtained. A certain number of training samples were selected for each category, and the model was trained using these seed categories to meet the requirements for classifying various subcategories. Finally, the trained classification model was used to achieve fine classification of crops.

[0100] The growth status assessment uses vegetation indices (NDVI and LAI). By establishing empirical regression models of growth indices for different crops, the growth and health status of crops are quantitatively assessed.

[0101] Specifically:

[0102] First, empirical regression models were established for vegetation indices (NDVI, LAI) and growth parameters (leaf area index, chlorophyll content, nitrogen content, and aboveground biomass), mainly including five types of models: power function, linear function, quadratic function, exponential function, and polynomial function. Next, the coefficient of determination, root mean square error, and estimation accuracy were used to evaluate the inversion models, and the optimal growth monitoring model was selected. Finally, based on the optimal growth monitoring model, monitoring results for rice, maize, wheat, and rapeseed were obtained.

[0103] Among them, the vegetation index is the normalized vegetation index (NDVI). The changes in NDVI are closely related to crop growth status and development stage. It can accurately reflect vegetation greenness, photosynthetic intensity, vegetation metabolic intensity and their seasonal and interannual variations, and is widely used in large-scale vegetation dynamic monitoring, crop growth monitoring and crop yield prediction.

[0104] NDVI = (NIR - R) / (NIR + R)

[0105] Wherein, NIR represents the near-infrared band of the satellite image, and R represents the red band of the satellite image.

[0106] Growth parameters include leaf area index, chlorophyll content, nitrogen content, and aboveground biomass.

[0107] (1) Leaf area index (LAI)

[0108] The leaf area index reflects the total area of ​​plant leaves per unit land area. It is related to both individual and population characteristics of plants and is a key parameter for characterizing vegetation growth status, as well as a commonly used parameter for monitoring crop growth.

[0109] (2) Chlorophyll content (Soil and plant analyzer development, SPAD)

[0110] Chlorophyll is the main substance for crops to absorb light and is one of the important pigments for light energy utilization. Plant photosynthesis relies on chlorophyll to absorb and convert light energy, and the chlorophyll content of vegetation plays a dominant role in the photosynthetic rate, thus directly affecting crop growth and yield.

[0111] (3) Plant nitrogen concentration (PNC)

[0112] Nitrogen is an essential nutrient element for crop growth. It is a basic component of chlorophyll, protein, genetic material, and other organic molecules in plants, participating in all stages of plant growth and development. The sufficiency of nitrogen concentration in plants significantly affects crop growth and development and is closely related to grain quality and yield.

[0113] (4) Aboveground biomass (AGB)

[0114] Aboveground biomass refers to the dry mass of living organic matter per unit area at a certain moment. It is an important biochemical parameter used to characterize the growth and development of crops. It has a significant impact on light energy utilization and dry matter yield formation, and is an important parameter for characterizing crop growth status and closely related to yield.

[0115] Empirical regression analysis was employed, with the vegetation index NDVI as the independent variable and crop growth parameters (leaf area index, chlorophyll content, nitrogen content, and aboveground biomass) as the dependent variables, to establish a growth monitoring model. The model types used included exponential, linear, polynomial, logarithmic, and power function models. The model with the highest coefficient of determination (R²) was selected, and the performance of each regression model was compared using the coefficient of determination, root mean square error, and estimation accuracy. The model with the best performance was then used for growth monitoring.

[0116] Disaster analysis employs commonly used large-scale disaster models, combined with meteorological data, to simulate and predict disasters; after a disaster occurs, the model is used to conduct a post-disaster assessment of the scope and severity of the disaster's impact.

[0117] Specifically, disaster analysis and simulation models are based on mathematical and statistical methods. They calculate the probability and severity of disasters by establishing models that require substantial data and prior expertise to provide relatively accurate and reliable assessment results. This invention primarily employs the following two models to achieve disaster analysis and simulation:

[0118] Regression models: These are mathematical models built by analyzing historical data and relevant factors to predict the likelihood and extent of disasters. For example, by analyzing historical flood data and rainfall, predictive models can be built to estimate the probability of flooding and the extent of flood inundation.

[0119] Neural network models are models that simulate the workings of the human nervous system, possessing powerful nonlinear mapping capabilities and self-learning abilities. In disaster analysis and simulation, neural network models, through their self-learning capabilities, use large datasets of past disasters for training and validation, thereby predicting the probability of disaster occurrence, the extent of impact, and post-disaster recovery.

[0120] III. Data Validation Module

[0121] Farmland vector data and farmer information are the two most important types of data in the entire system, forming the core foundation of the "human-land relationship." Ensuring the accuracy of farmer information is crucial to ensuring farmers can successfully enroll in insurance and enjoy insurance coverage, and to guaranteeing their legitimate rights and interests through timely compensation after disasters. Ensuring the accuracy of farmland information is also essential to preventing false reporting, concealment, and other violations, promptly detecting and stopping insurance fraud, and maintaining the healthy and stable development of the agricultural insurance market.

[0122] By verifying cultivated land vector data and farmer information data, the accuracy and reliability of basic data on the "human-land relationship" can be improved, enabling more effective application in insurance business. The insurance industry can then more accurately grasp the market demand for agricultural insurance, optimize resource allocation, improve operational efficiency, and provide strong support for the sustainable and healthy development of agricultural insurance. This has profound significance for protecting farmers' rights and interests and promoting stable agricultural development.

[0123] This invention employs two targeted methods for data verification based on different types of farmers, and utilizes satellite remote sensing monitoring and AI image recognition technology to improve verification efficiency, reduce manual work, and ensure the accuracy of verification results.

[0124] 1. Large account data verification

[0125] After the large-scale farmland boundary data is identified and intelligently drawn through on-site communication by the sales staff, it will be compared with the crop classification data in the system based on satellite remote sensing image preprocessing. If the large-scale farmer's insurance information matches the system data (the insured crop type is the same as the crop identified in the system preprocessing, and the drawn large-scale farmer's plot area differs from the crop area identified by the system within the specified error range), the verification is considered successful; otherwise, on-site sales staff need to conduct manual confirmation, and the large-scale farmer can complete the verification after providing supporting documents.

[0126] 2. Retail investor data verification

[0127] The data collection of farmland boundaries from individual farmers is done independently by them through a system mobile app, which raises ethical concerns regarding the accuracy of the results. Furthermore, the large volume of data and fragmented vector boundaries from individual farmers' plots require a greater margin of error and tolerance compared to the data from large-scale farmers. To address this issue, this invention combines two methods—overlay comparison of land ownership data and photo recognition comparison—for preliminary verification.

[0128] First, agricultural land ownership survey data is used to overlay and compare the spatial location of cultivated land photographed and returned to the system by individual farmers with the land ownership survey data within a certain buffer range to verify whether the information of the cultivated land owner in the land ownership survey data matches. It should be noted that because the land ownership survey data has poor timeliness, it is usually 5 to 10 years behind the actual situation, and land ownership has changed significantly. Therefore, the matching results are poor. Thus, a failure to match this information does not necessarily mean that the data verification has failed; conversely, if the match is successful, the verification is successful.

[0129] Specifically: First, a circular buffer zone of 100m is drawn from the spatial point P where the individual investor takes photos, and the vector range Pi is obtained. The formula is Pi = (P:d(Pi,Oi)≤100); then, using the GIS overlay analysis capability, the Pi vector and the weighted vector data are overlaid and analyzed. If the weighted vector patch of the individual investor's ownership exists in the overlay analysis result, then the verification is passed.

[0130] Second, photo recognition and comparison: using mature AI image recognition technology, the crop type in the photos taken by individual farmers is identified and compared with the crop information insured by the individual farmers.

[0131] Compared to large investors, the data verification of individual investors involves a larger volume of data and is affected by various real-world factors. Therefore, it is necessary to achieve large-scale data verification through methods such as manual review, grassroots outreach, and annual updates.

[0132] IV. Data Storage Management Module

[0133] Data storage management builds two databases: a basic database and a business database. Based on the "human-land relationship," it abstracts the entity objects of agricultural insurance and maps the business objects through knowledge graph modeling, thereby constructing a large-scale agricultural insurance data model to support insurance application services.

[0134] 1. Basic database and entity objects

[0135] Basic databases are categorized based on their data storage formats into relational databases (such as Oracle, SQL Server, DB2, PostgreSQL, and MySQL), spatial databases (such as PostGIS, Oracle Spatial, and Spatial Lite), and file-based databases (such as MongoDB and SequoiaDB). These can be abstracted into three databases: human-land, insurance, and remote sensing.

[0136] The human-land database is responsible for storing data on key participants in agricultural insurance, including farmers, insurance company personnel, and grassroots insurance assistants, as well as data on insurable arable land cultivated by farmers. These two data form the foundation of the "human-land relationship" through a strong one-to-many relationship, supporting the entire agricultural insurance data system.

[0137] The insurance database is responsible for storing business data for the entire agricultural insurance process, including data, forms, and various attachments from the processes of insurance application, verification, claims settlement, and other related procedures, in order to meet the application needs of agricultural insurance business.

[0138] The remote sensing database is responsible for storing basic satellite remote sensing images, as well as pre-processed satellite remote sensing monitoring and analysis data such as farmland boundaries, crop classification, growth status assessment, and disaster analysis, providing data support for high-quality agricultural insurance services.

[0139] After abstraction and classification, it specifically includes five major entity objects: "people, place, thing, and event".

[0140] People: This refers to the core group of people involved in agricultural insurance, centered around the insured farmers. It also includes insurance company agents, grassroots insurance coordinators, and third-party personnel, who together constitute the main actors in agricultural insurance.

[0141] Land: refers to the cultivated land insured by participating farmers. There is a one-to-many relationship between the insured farmers and the farmers, which provides physical space support for agricultural insurance.

[0142] "Insured property" refers to the crops insured by the insured household, i.e., the insured object, which is the core property of agricultural insurance.

[0143] Events: refers to various business events in the stages of underwriting, inspection, loss assessment, and claims settlement.

[0144] "Situation" refers to routine monitoring data and analysis conclusions on crop growth, yield, diseases, etc., which can provide scientific data indicators to support operations.

[0145] 2. Business database and business objects

[0146] A business database is a business logic data model stored in a computer after the data physically stored in the basic database is abstracted into entity objects, modeled with knowledge graphs, and mapped to business objects.

[0147] Specifically, it includes four main business areas:

[0148] The relationship between people and land: the fundamental basis of agricultural insurance. People and land are interconnected and together form the core of insurance business.

[0149] Underlying crops: the core of agricultural insurance business.

[0150] Insurance event: A business segment related to agricultural insurance.

[0151] Situation monitoring: Agricultural insurance provides business empowerment.

[0152] By establishing a foundational framework for "human-land relationship" through the connection between people and land, the corresponding target crops are solidified; with these target crops as the core of the business, business is carried out throughout the entire life cycle; in this process, satellite remote sensing is used to conduct routine monitoring and analysis of crop conditions at each business node, so as to realize data support for business and technology empowerment for agricultural insurance.

[0153] 3. Large-scale data model for agricultural insurance:

[0154] Model design ideas (such as) Figure 4 As shown):

[0155] This process conceptualizes all specific aspects and business involved in agricultural insurance, categorizing and abstracting related objects with common characteristics and strong business connections to form a conceptual entity model encompassing five major categories: "people, land, things, and events." Further, this entity model is linked, decomposed, and refined, mapping it to four major business objects: "people-land relationship," "target crop," "insurance event," and "situation monitoring," thereby supporting the overall operation of agricultural insurance (e.g.,...). Figure 5 (As shown).

[0156] Specifically:

[0157] Taking a participating farmer as an example, a one-to-many relationship between people and land is established based on the amount of cultivated land. On this basis, a one-to-many crop association is established for each plot of land according to the crop type. A single crop is taken as the core point of an insurance business, which is linked to the subsequent underwriting, verification, claims and other insurance business nodes. At the same time, satellites continuously and routinely monitor crops and feed the monitoring results back to the insurance business at the corresponding stage, providing scientific data support for the insurance business.

[0158] The summary has the following characteristics:

[0159] Full user coverage: Covering all participants in agricultural insurance business, including the subjects, objects, and third parties involved in existing agricultural insurance business.

[0160] Full spatial coverage: The relationship between people and land can clearly show the spatial location and vector information of all insured farmers' land plots, truly achieving full spatial coverage.

[0161] Full business coverage: It can describe the entire lifecycle of business such as underwriting, inspection, loss assessment, and claims settlement, and provide corresponding business application support for each key business node.

[0162] Full time coverage: Through satellite remote sensing monitoring, we have truly achieved normalized monitoring of crop targets at all times, making them traceable and predictable.

[0163] V. Application Service Module

[0164] Based on human-land relationship data, combined with satellite remote sensing imagery and data on farmland distribution and crop types, an analysis model for pre-insurance verification is established and compared with the insurance policy to achieve precise agricultural insurance coverage.

[0165] Based on remote sensing satellite data and related analysis results, and combined with the specific types of agricultural insurance and risk factors, a risk assessment model is established. Based on the risk assessment results, agricultural insurance premium rates and claims standards are formulated.

[0166] Remote sensing data is used for disaster early warning and monitoring to promptly identify potential risks and anomalies. After a disaster occurs, image data of the affected area is quickly acquired to assess the extent and scope of damage, providing a basis for claims processing.

[0167] VI. System Security Module:

[0168] Collaborative work: Build a cloud service platform that supports multi-user online access, data exchange, and collaborative work.

[0169] Network security design: Employing network security devices and technologies such as firewalls and intrusion detection / prevention systems (IDS / IPS) to build a secure network topology. Preventing external attacks and unauthorized access, protecting the cloud platform from network threats.

[0170] Identity authentication and access control: Establish a strict identity authentication mechanism to ensure that only authorized users can access cloud platform resources. Implement fine-grained access control policies to restrict users' access to and operations on resources based on their roles and permissions.

[0171] Data encryption and key management: Encrypt sensitive data in the cloud platform to ensure data confidentiality and integrity. Establish a key management system to achieve secure key generation, storage, distribution, and destruction.

[0172] Security Audit and Logging: Establish a security audit mechanism to record and analyze operations, access, and events on the cloud platform. Through log recording and analysis, security issues can be identified and addressed promptly, providing a basis for tracing and investigating security incidents.

[0173] Existing data acquisition methods typically use UAV aerial imagery as the base imagery, requiring cumbersome and highly specialized methods such as field collection and indoor processing. This invention, based on satellite remote sensing imagery and incorporating an AI image segmentation model, along with a mobile app, provides methodological and systemic support for the acquisition of arable land spatial vector data for agricultural insurance, enabling a wider range of basic data to be aggregated into the system in a low-cost and efficient manner.

[0174] Existing conventional database storage cannot effectively aggregate and form a cohesive whole from massive amounts of diverse, multi-source data in various formats. This invention proposes a "two-database" approach—a basic database and a business database—which, through the abstraction of entity objects and the mapping of business objects, solves the problem of data silos and achieves efficient data management.

[0175] Existing data models cannot adequately reflect the complexities of agricultural insurance operations. The agricultural insurance big data model proposed in this invention effectively establishes the connection between business and data through knowledge graph modeling, forming a comprehensive agricultural insurance big data system covering all aspects of the business. This achieves efficient data collection, clear business processes, routine process monitoring, and intelligent analysis and reasoning, enabling a deeper exploration of the value of agricultural data.

[0176] In this invention, satellite remote sensing and AI are the main technologies used. The image segmentation algorithm is used to collect large-scale land plots on satellite imagery, reducing technical difficulty and cost. At the same time, a one-click photo identification function is provided for individual farmers on mobile devices. This expands the scope of basic data collection for agricultural insurance with the same cost investment, and to some extent solves the problem of small coverage of agricultural insurance in the past.

[0177] This invention proposes to abstract entity objects, abstracting and materializing real-world things into "people, places, things, events" entity objects, constructing the relationship between entities, and then mapping them to four major business objects to achieve full-coverage modeling of users, space, business, and time, supporting business applications.

[0178] The agricultural insurance data big data model proposed in this invention uses knowledge graph technology to build two databases: a basic database and a business database. This forms a big data system covering the entire physical space and the entire business scope, which can better manage basic data and more deeply explore the value of the data.

[0179] The data verification module proposed in this invention comprehensively adopts three technical means—satellite remote sensing crop type monitoring and comparison, weighted data overlay analysis, and AI image intelligent crop recognition—to address the different data collection characteristics of large-scale farmers and small-scale farmers. This reduces manual work while ensuring the accuracy of the verification results.

[0180] This invention features a user-friendly interface and a simple workflow, making it easy for farmers and other non-professionals to use. It employs a secure and reliable cloud computing architecture to ensure the secure storage and transmission of confidential data such as satellite remote sensing imagery, farmer information, farmland vector data, and insurance policies.

[0181] The data verification method proposed in this invention focuses on data quality and accuracy, which are generally not addressed in existing technical solutions. It uses an efficient and low-manual approach to complete data verification, providing the most basic guarantee for further data applications.

[0182] This invention utilizes the all-time monitoring capabilities of satellite remote sensing to provide agricultural insurance with more multi-dimensional data references and analytical verification, which is more in line with the current development trend of technology empowering agricultural insurance.

[0183] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0184] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An agricultural insurance data acquisition and management system based on satellite remote sensing, characterized in that: It includes a data acquisition module, a data preprocessing module, a data verification module, and a data management module; The data acquisition module is designed for different data types and includes basic data acquisition, general data uploading, and vector data acquisition. The data preprocessing module performs preprocessing operations on the basic data and vector data collected by the data acquisition module to improve data quality. The data verification module improves the accuracy and reliability of basic data by verifying cultivated land vector data and insured household information data; The data management module builds a large-scale agricultural insurance data model by establishing a basic database and a business database, and mapping business objects through knowledge graph modeling.

2. The agricultural insurance data acquisition and management system based on satellite remote sensing according to claim 1, characterized in that: The basic data acquisition module acquires data by using remote sensing satellites to obtain remote sensing image data of the agricultural insurance coverage area, including but not limited to high-resolution imagery and multispectral imagery. The general data upload of the data collection module includes insured household information data, insurance company data, basic data, and business forms; among them, insured household information data includes farmer name, insurance information, and planting information; insurance company data includes company information, salesperson information, and assistant information; basic data includes insurance type information, subsidy information, and administrative division; business forms include insurance application form, household list, and claim form; The vector data acquisition module includes the spatial location and boundary vector data of the insured's land plots. Using satellite remote sensing imagery as the base map and combining it with intelligent boundary extraction algorithms, the module completes the land plot data acquisition for the insured. The acquired land plot data is then combined with the insured's basic information to establish the human-land relationship within the agricultural insurance data system.

3. The agricultural insurance data acquisition and management system based on satellite remote sensing according to claim 2, characterized in that: The intelligent boundary extraction algorithm is as follows: First, the remote sensing image data collected by the data acquisition module is locked and captured. The captured remote sensing image data is then sent to the image encoder deployed on the backend server to interpret the image and obtain the feature map of the remote sensing image data. The image encoder acquires the location information from remote sensing image data and feeds it back to the prompt encoder, which then converts the coordinates into feature vectors. The intermediate decoder combines the feature vector with the feature map, inputs it into the mask decoder to obtain the returned result set, scales and maps it according to the original image size, and finally outputs the result with the highest score, which is the extracted land data of the insured.

4. The agricultural insurance data acquisition and management system based on satellite remote sensing according to claim 3, characterized in that: The data preprocessing module performs preprocessing operations on the collected data, including geometric correction, radiometric correction, noise reduction, cropping and mosaicking, color balancing, and format conversion of remote sensing image data, to improve data quality and provide a single image to support all modules of the system.

5. The agricultural insurance data acquisition and management system based on satellite remote sensing according to claim 4, characterized in that: Geometric correction: Eliminates geometric distortion of the image and matches the pixel position with the real geographical location on the Earth's surface, thereby giving the image a clear geographic reference coordinate; Radiometric correction: used to eliminate sensor errors and atmospheric interference, and restore the true radiation or reflection characteristics of ground objects; Noise filtering: Eliminates noise and artifacts in images, improving image clarity and interpretability; Cutting and inlaying: Adjusting the scope and seamlessly splicing data according to research needs; Color balancing: Eliminates color differences between multiple images to ensure overall consistency; Data format conversion: Converting data into a standard data format to facilitate subsequent analysis and processing.

6. The agricultural insurance data acquisition and management system based on satellite remote sensing according to claim 5, characterized in that: The data preprocessing module also combines Geographic Information System (GIS) and Artificial Intelligence (AI) to extract farmland boundaries, classify and identify crops, assess growth status, and analyze disasters from the collected remote sensing image data, providing data support for subsequent routine monitoring.

7. The agricultural insurance data acquisition and management system based on satellite remote sensing according to claim 6, characterized in that: The data verification module includes verification of cultivated land vector data for large-scale farmers and verification of data for small-scale farmers. Specifically, the verification of large account data includes: After the boundary data of large-scale farmland is identified through on-site communication and intelligent boundary drawing, it is compared with the crop classification data in the system based on satellite remote sensing image preprocessing. If the insurance information of the large-scale farmer matches the system data, that is, the type of the insured crop is the same as the crop identified in the system preprocessing, and the difference between the drawn large-scale farmer's plot area and the crop area identified by the system is within the specified error range, then the verification is considered successful; otherwise, on-site business personnel need to manually confirm and complete the verification. The specific steps for verifying retail investor data are as follows: First, a circular buffer zone of 100m is drawn from the spatial point P where the individual investor takes photos, and the vector range Pi is obtained. The formula is Pi = (P:d(Pi,Oi)≤100). Then, using the GIS overlay analysis capability, the Pi vector and the weighted vector data are overlaid and analyzed. If the weighted vector patch of the individual investor's ownership exists in the overlay analysis result, the verification is passed. Then, AI image recognition technology is used to identify the crop type in the photos taken by individual farmers and compare and verify it with the crop information insured by those farmers.

8. The agricultural insurance data acquisition and management system based on satellite remote sensing according to claim 7, characterized in that: In the data management module, the setup of the basic database is as follows: Based on the different data formats stored, basic databases are divided into relational databases, spatial databases, and file-based databases, which can be abstracted into three databases: human-land, insurance, and remote sensing. Among them, the human-land database is used to store data on core participants in agricultural insurance and data on insurable arable land planted by farmers. These two are established through a strong one-to-many relationship, which forms the basis of the "human-land relationship" and supports the entire agricultural insurance data system. The insurance database is used to store business data for the entire agricultural insurance process, including data, forms, and various attachments during the insurance application, inspection, claims, and settlement processes, in order to meet the application needs of agricultural insurance business. The remote sensing database is used to store basic satellite remote sensing images as well as preprocessed satellite remote sensing monitoring and analysis data on farmland boundaries, crop classification, growth status assessment, and disaster analysis, providing data support for high-quality agricultural insurance services.

9. The agricultural insurance data acquisition and management system based on satellite remote sensing according to claim 8, characterized in that: In the data management module, the specific setup of the business database is as follows: A business database is a data model that takes the data physically stored in the basic database, abstracts it into entity objects, models it using knowledge graphs, and maps it to business objects. This model is then stored in the computer as a business logic data model. This includes data on human-land relationships, target crops, insurance events, and situation monitoring.

10. The agricultural insurance data acquisition and management system based on satellite remote sensing according to claim 9, characterized in that: In the data management module, the agricultural insurance data big data model is specifically as follows: The concept of agricultural insurance involves understanding all specific things and businesses, and classifying and abstracting related objects with common characteristics and strong business connections to form a conceptual entity object model of five categories: "people, place, things, and events". The entity object model is then associated, decomposed, and refined, and mapped to four major categories of business objects: human-land relationships, target crops, insurance events, and situation monitoring, thereby supporting the overall operation of agricultural insurance.

Images