Full life cycle dynamic monitoring system and method for comprehensive land management

Abstract

This invention provides a dynamic monitoring system and method for the entire lifecycle of comprehensive land consolidation. First, a multi-source data fusion module collects and integrates heterogeneous data related to comprehensive land consolidation across the entire region, constructing a data base with unified spatiotemporal references and semantic associations. Then, an intelligent monitoring module, based on this data base, utilizes preset business rules and machine learning models to identify land use type changes, track project progress, and dynamically monitor indicators, acquiring monitoring data in real time and generating early warning information. Finally, a decision support module visualizes the monitoring data and early warning information, while providing interactive services and decision support for users. This invention can efficiently integrate heterogeneous multi-source data, enabling dynamic monitoring of the entire process of comprehensive land consolidation projects, intelligent effectiveness evaluation, and risk early warning, providing strong support for management decisions.

Description

Technical Field

[0001] This invention relates to the field of land resource management technology, and more specifically, to a dynamic monitoring system and method for the entire life cycle of land consolidation across the entire region. Background Technology

[0002] Comprehensive land consolidation is a comprehensive project guided by national land spatial planning, coordinating the consolidation of agricultural land and construction land with ecological restoration. It effectively improves land use efficiency, restores the ecosystem, improves the living environment, promotes industrial integration, and contributes to rural revitalization and urban-rural integration, thereby achieving a win-win situation for economic, social, and ecological benefits. Therefore, efficient, accurate, and dynamic monitoring and supervision of the entire process of comprehensive land consolidation projects are crucial.

[0003] Currently, the comprehensive land consolidation project faces the following problems in terms of process supervision:

[0004] 1) Data fragmentation and silos: The remediation involves multiple stages such as planning, implementation, acceptance, and maintenance, and involves multiple departments such as natural resources, agriculture and rural affairs, forestry, environmental protection, water conservancy, and housing and construction. The data sources are scattered and have different formats, making it difficult to effectively integrate them into a full-domain and full-cycle view.

[0005] 2) Poor monitoring timeliness: Existing technologies mostly rely on a combination of text reports, tables, pictures and manual on-site spot checks for data monitoring. This method is time-consuming, costly, and has limited coverage, making it difficult to detect deviations (such as illegal occupation, project delays, and substandard quality) and potential risks in the rectification process in a timely manner.

[0006] 3) The analysis and evaluation are highly subjective: The evaluation of the effectiveness of the remediation (such as the increase in the amount of arable land and the improvement of land use efficiency) relies heavily on simple quantitative indicators and lacks the support of objective, qualitative and multi-dimensional intelligent analysis models for the middle and later stages of project implementation.

[0007] 4) Insufficient early warning capabilities: It is difficult to proactively issue early warnings based on real-time or near-real-time data regarding issues such as delayed progress of remediation projects, abnormal use of funds, negative impacts on the ecological environment, and deviations from planning implementation.

[0008] 5) Weak decision support: There is a lack of effective tools to transform monitoring data into intuitive and actionable decision-making information (such as optimizing resource allocation, adjusting remediation strategies, and identifying demonstration areas). Summary of the Invention

[0009] To overcome the shortcomings of existing technologies in achieving efficient and high-precision monitoring of the entire process of comprehensive land consolidation, this invention provides a dynamic monitoring system and method for the entire life cycle of comprehensive land consolidation. This system can efficiently integrate multi-source heterogeneous data, realize dynamic monitoring of the entire process of comprehensive land consolidation projects, intelligent effectiveness evaluation and risk warning, and provide strong support for management decisions.

[0010] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0011] A dynamic monitoring system for the entire life cycle of comprehensive land consolidation includes, in sequence: a multi-source data fusion module, an intelligent monitoring module, and a decision support module;

[0012] The multi-source data fusion module includes, in sequence, a data access layer, a cleaning and treatment layer, a fusion processing layer, and an application interface layer. The multi-source data fusion module is used to collect and fuse multi-source heterogeneous data related to comprehensive land consolidation across the entire region, and to construct a data base with unified spatiotemporal benchmarks and semantic associations.

[0013] The intelligent monitoring module comprises, in sequence: a feature index extraction layer, a dynamic change analysis layer, an evaluation and early warning layer, and a feedback optimization layer. The intelligent monitoring module is used to identify land use type changes, track project progress, and dynamically monitor indicators based on the data base and using preset business rules and machine learning models, thereby acquiring monitoring data in real time and generating early warning information.

[0014] The decision support module comprises, in sequence, a data link layer, a visualization rendering layer, an interactive service layer, and a decision support layer; the data link layer is connected to the dynamic change analysis layer and the evaluation and early warning layer respectively; the decision support module is used to visualize the monitoring data and early warning information, and at the same time provide users with interactive services and decision support.

[0015] Preferably, in the multi-source data fusion module,

[0016] The data collected by the data access layer includes at least one or more of the following: satellite remote sensing imagery, UAV aerial survey data, ground sensor data, business database and document / table data, and publicly available Internet data; the data collected by the data access layer is classified, and the classified data is distributed and stored across multiple nodes using DSF / HDSF distributed storage technology;

[0017] The cleaning and governance layer is used to clean the data collected by the data access layer based on predefined cleaning rules. The data cleaning includes at least one or more of the following: outlier marking and correction, missing value imputation, and duplicate value removal.

[0018] The fusion processing layer is used to sequentially perform spatiotemporal benchmark unification, semantic association modeling, and business scenario adaptation processing on the cleaned data to generate the data baseboard. The spatiotemporal benchmark unification includes unifying and correcting the time and spatial coordinates of multi-source heterogeneous data. The semantic association modeling includes constructing a knowledge graph based on the spatiotemporally benchmark-unified data and performing entity semantic association. The business scenario adaptation includes dynamically packaging the semantically association-modeled data into several datasets adapted to different stages according to the data differences required at different stages of comprehensive land consolidation, and jointly saving the datasets corresponding to all stages as the data baseboard.

[0019] The application interface layer is used to output the data baseboard to the intelligent monitoring module through preset multi-interfaces.

[0020] Preferably, in the intelligent monitoring module,

[0021] The feature index extraction layer is used to extract several monitoring indicators from the data base and perform feature quantification processing based on the pre-constructed monitoring indicator system for comprehensive land consolidation.

[0022] The dynamic change analysis layer is used to automatically identify land parcels based on a preset machine learning model, further identify changes in land use types, and extract change data; at the same time, it is used to track project progress and dynamically monitor indicators based on preset business rules, and acquire monitoring data in real time.

[0023] The assessment and early warning layer is used to compare the monitoring data of each indicator with the preset threshold to achieve graded early warning and generate early warning information;

[0024] The feedback optimization layer is used to receive the actual rectification situation, compare it with the early warning information, and optimize the machine learning model and business rules based on the comparison results.

[0025] Preferably, in the feature index extraction layer, a comprehensive land consolidation monitoring index system is constructed based on the analytic hierarchy process (AHP), and the index system includes at least: a target layer, a criterion layer, and an index layer.

[0026] The target layer includes: basic bottom-line constraints, improvement of farmland ecosystem service functions, ecological protection and restoration and improvement of rural landscape, project and fund coordination and project supervision;

[0027] The basic bottom-line constraints in the aforementioned criteria layer include: quantity control, quality control, and other controls; improvement of farmland ecosystem service functions include: the degree of farmland concentration and contiguousness, the degree of intensive and economical use of construction land, and land transfer status; ecological protection and restoration and rural landscape improvement include: the status of ecological protection and restoration and the status of rural landscape improvement; project and funding coordination includes: the status of funding guarantee for improvement, the progress of project completion, and the project quality objectives; project supervision includes: innovation and illegal monitoring.

[0028] In the indicator layer, quantity control includes: the proportion of newly added arable land area, the proportion of newly added permanent basic farmland area, and the surplus construction land indicator; quality control includes: improving the quality grade of arable land; other controls include: the area conflicting with ecological protection red lines and the area protected for historical and cultural heritage; the degree of farmland concentration and contiguousness includes: the area of ​​newly added high-standard farmland, the area of ​​developed and supplemented arable land, the area of ​​reclaimed paddy fields, the number of small arable land plots converted into large arable land plots, and the degree of increase in the concentration and contiguousness of permanent basic farmland; the degree of intensive and economical use of construction land includes: the proportion of land use quota linked to land increase and decrease used for infrastructure construction, the scale of redevelopment of inefficient construction land, the scale of demolition and reclamation of rural construction land, the rate of decrease in the average household homestead area, and the scale of reduction in village construction land; land transfer includes: the area of ​​arable land used for grain cultivation after remediation, the introduction of agricultural enterprises, the number of large grain growers, and the proportion of land management rights transferred; ecological The protection and restoration status includes: the area of ​​comprehensive mine management, the area of ​​comprehensive water environment management, the area of ​​forest land transformation, the area of ​​mangrove protection and restoration, the area of ​​coastal zone remediation, the number of kilometers of newly added ecological corridors, and the soil and water conservation rate; the rural landscape improvement status includes: the number of rural living environment improvements, the coverage rate of sewage treatment facilities, the coverage rate of garbage disposal, and the popularization rate of sanitary toilets; the remediation funding guarantee status includes: the proportion of integrated agricultural funds, the completion rate of budgeted investment funds, the actual expenditure rate of funds, and the proportion of social capital investment; the project completion progress status includes: the completion rate of agricultural land remediation projects, the completion rate of construction land remediation projects, the completion rate of ecological protection and restoration projects, the completion rate of cultural protection and rural landscape improvement projects, and other projects; the project quality target status includes: the project quality pass rate; innovation and illegal monitoring includes: the number of innovative systems and the area of ​​illegal land use.

[0029] Preferably, in the dynamic change analysis layer, the preset machine learning model is specifically an improved PiDiNet model, used for plot edge extraction;

[0030] The improved PiDiNet model structure includes convolutional blocks 1, 2, 3, and 4 connected in sequence. Each convolutional block outputs feature maps for different channels. After channel-wise transformation of each feature map, they are input into an attention layer to obtain corresponding attention features. The attention features corresponding to convolutional blocks 2-3 are then subjected to feature difference and concatenated with the attention features corresponding to convolutional block 1 to obtain concatenated features. The concatenated features are then channel-wise transformed again, and finally, the Sigmoid function is used to obtain the detection results of the plot edges.

[0031] Convolutional block 1 and convolutional block 2 have the same structure, both including a Gabor convolutional layer, a batch normalization layer, a max pooling layer and an activation layer connected in sequence; convolutional block 3 and convolutional block 4 have the same structure, both including a convolutional layer, a batch normalization layer, a max pooling layer and an activation layer connected in sequence.

[0032] Preferably, the Gabor convolutional layer includes at least one Gabor filter and a learnable convolutional kernel.

[0033] Preferably, in the decision support module,

[0034] The data link layer is used to collect the monitoring data and early warning information in real time;

[0035] The visualization rendering layer is used to realize three-dimensional terrain visualization, two-dimensional thematic visualization, and dynamic timeline visualization based on monitoring data and early warning information. The three-dimensional terrain visualization includes: constructing a three-dimensional real-scene model of the entire land consolidation project area, overlaying monitoring data and early warning information of each plot, and realizing three-dimensional information browsing; the two-dimensional thematic visualization uses an external GIS system to overlay and analyze monitoring data and early warning information to generate multiple thematic layers; the dynamic timeline visualization uses the timeline of the entire land consolidation project as the axis, combined with monitoring data and early warning information, to realize dynamic playback and future projection of the consolidation process.

[0036] The interaction service layer is used to provide interaction services to users, including general services and customized services;

[0037] The decision support layer is used to provide decision support for users.

[0038] Preferably, in the visualization rendering layer, a three-dimensional real-scene model of the entire land consolidation project area is constructed based on WebGL or Unity 3D engine.

[0039] Preferably, in the interactive service layer, the general services include: intelligent data annotation, spatial measurement, comparative analysis, and intelligent search;

[0040] The customized services include: for managers, a cockpit view with dynamic dashboards integrating core indicators; for technical personnel, a professional analysis view with overlaid engineering drawings and sensor network topology diagrams for data interface debugging and model parameter adjustment; and for grassroots users, a simplified interface that allows them to intuitively view the remediation effects of their own land through a 3D real-scene model and provides online feedback functionality.

[0041] This invention also provides a dynamic monitoring method for the entire life cycle of comprehensive land consolidation, based on the above-mentioned system, comprising the following steps:

[0042] S1: The multi-source data fusion module collects and merges multi-source heterogeneous data related to comprehensive land consolidation across the entire region to construct a data base with unified spatiotemporal benchmarks and semantic associations;

[0043] S2: The intelligent monitoring module, based on the data base, uses preset business rules and machine learning models to identify changes in land use types, track project progress, and dynamically monitor indicators, acquire monitoring data in real time, and generate early warning information;

[0044] S3: The decision support module visualizes the monitoring data and early warning information, and provides users with interactive services and decision support.

[0045] This invention provides a dynamic monitoring system and method for the entire lifecycle of comprehensive land consolidation. First, a multi-source data fusion module collects and integrates heterogeneous data related to comprehensive land consolidation across the entire region, constructing a data base with unified spatiotemporal references and semantic associations. Then, based on this data base, an intelligent monitoring module uses preset business rules and machine learning models to identify land use type changes, track project progress, and dynamically monitor indicators, acquiring monitoring data in real time and generating early warning information. Finally, a decision support module visualizes the monitoring data and early warning information, while providing interactive services and decision support for users.

[0046] Compared with the prior art, the beneficial effects of the technical solution of the present invention are:

[0047] 1) Full-area and full-cycle coverage: Achieve dynamic tracking of all elements of land consolidation projects within a specific area, from planning to management.

[0048] 2) Deep integration of multi-source data: Break down data silos, form a unified and authoritative "data base" for governance, and improve information integrity.

[0049] 3) Real-time and automated monitoring: Significantly reduces reliance on manual labor, increases monitoring frequency and efficiency, and enables timely detection and management of problems.

[0050] 4) Objective and intelligent assessment: Based on model-driven quantitative assessment, subjectivity is reduced and the effectiveness of the rectification is measured in a comprehensive and scientific manner.

[0051] 5) Precise and proactive early warning: This invention can transform the traditional post-event handling plan into a pre-event early warning and in-event intervention plan, effectively preventing risks and improving management initiative.

[0052] 6) Scientific and Visualized Decision Making: Provides intuitive and comprehensive information views and auxiliary analysis tools to significantly improve the scientific nature and efficiency of user decision making.

[0053] 7) Standardization and scalability: The system architecture and method of this invention are universal and can provide a reference for land consolidation monitoring in different regions, with a wide range of applicability. Attached Figure Description

[0054] Figure 1 This is a structural diagram of a full life-cycle dynamic monitoring system for comprehensive land consolidation provided in Example 1.

[0055] Figure 2 This is an overall architecture diagram of a full life-cycle dynamic monitoring system for comprehensive land consolidation provided in Example 2.

[0056] Figure 3 This is a schematic diagram of the DSF / HDSF distributed storage technology provided in Example 2.

[0057] Figure 4 This is a schematic diagram of the construction of the monitoring indicator system for comprehensive land consolidation provided in Example 2.

[0058] Figure 5 This is a schematic diagram of the land parcel identification process provided in Example 2.

[0059] Figure 6 This is a schematic diagram of the Gabor convolutional layer provided in Example 2.

[0060] Figure 7 This is a schematic diagram of the detection effect of the homogeneous block provided in Example 2.

[0061] Figure 8 This is a schematic diagram of the detection effect of the non-homogeneous block provided in Example 2.

[0062] Figure 9 This is a schematic diagram of the detection effect of dense plots provided in Example 2.

[0063] Figure 10 This is a flowchart of a full life-cycle dynamic monitoring method for comprehensive land consolidation provided in Example 3. Detailed Implementation

[0064] The accompanying drawings are for illustrative purposes only and should not be construed as limiting the scope of this application.

[0065] To better illustrate this embodiment, some parts in the accompanying drawings may be omitted, enlarged, or reduced, and do not represent the actual product dimensions;

[0066] It will be understood by those skilled in the art that certain well-known structures and their descriptions may be omitted in the accompanying drawings.

[0067] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0068] Example 1

[0069] like Figure 1 As shown, this embodiment provides a dynamic monitoring system for the entire life cycle of comprehensive land consolidation, which includes, in sequence: a multi-source data fusion module, an intelligent monitoring module, and a decision support module;

[0070] The multi-source data fusion module includes, in sequence, a data access layer, a cleaning and treatment layer, a fusion processing layer, and an application interface layer. The multi-source data fusion module is used to collect and fuse multi-source heterogeneous data related to comprehensive land consolidation across the entire region, and to construct a data base with unified spatiotemporal benchmarks and semantic associations.

[0071] The intelligent monitoring module comprises, in sequence: a feature index extraction layer, a dynamic change analysis layer, an evaluation and early warning layer, and a feedback optimization layer. The intelligent monitoring module is used to identify land use type changes, track project progress, and dynamically monitor indicators based on the data base and using preset business rules and machine learning models, thereby acquiring monitoring data in real time and generating early warning information.

[0072] The decision support module comprises, in sequence, a data link layer, a visualization rendering layer, an interactive service layer, and a decision support layer; the data link layer is connected to the dynamic change analysis layer and the evaluation and early warning layer respectively; the decision support module is used to visualize the monitoring data and early warning information, and at the same time provide users with interactive services and decision support.

[0073] In the specific implementation process, this embodiment constructs a dynamic monitoring and decision support platform based on multi-source data fusion, with an intelligent analysis engine at its core, and oriented towards the entire life cycle of comprehensive land consolidation; the system architecture is as follows:

[0074] (1) Multi-source data fusion module (data end);

[0075] Comprehensive land consolidation is a multifaceted undertaking with diverse stakeholders, a wide coverage, and complex planning-engineering attributes. It involves business data from multiple departments such as natural resources, agriculture and rural affairs, ecology and environment, and water conservancy, as well as heterogeneous data from multiple sources, including satellite remote sensing, drone inspections, ground sensors, and mobile devices (such as spatiotemporal images, sensor time-series data, business ledgers, planning maps, and policy texts). This data suffers from pain points such as heterogeneous formats (structured / semi-structured / unstructured), inconsistent spatiotemporal benchmarks, uneven quality, and weak semantic connections, making it difficult to directly use the data for intelligent assessment and early warning. Therefore, this module adopts a layered and decoupled architecture, consisting of a data access layer, a cleaning and governance layer, a fusion processing layer, and an application interface layer, from the bottom layer to the application layer. Each layer works together to achieve a closed loop of "collection, management, fusion, and use" of multi-source data.

[0076] 1) Data Access Layer: Through microservice architecture design, it is compatible with mainstream protocols such as MQTT, HTTP / HTTPS, FTP, OGC WMS / WFS, JDBC / ODBC, and supports full compatibility access of multi-source heterogeneous data, including access to satellite remote sensing imagery (optical / radar / SAR), UAV aerial survey data, ground sensor networks, business databases (land and space planning, land use status, farmland occupation and compensation balance, etc.), documents / tables (project ledgers, approval documents, policies and regulations), and publicly available Internet data (meteorological disasters, regional population and economy), etc.

[0077] 2) Data Cleaning and Governance Layer: Addressing the issues of "dirtiness, disorder, and fragmentation" in multi-source data, this layer, based on predefined business rules (such as consistency of land use classification codes and logical verification of area values), uses structured data governance to automatically correct or mark outliers, and supports functions such as missing value imputation and duplicate value removal. Simultaneously, for remote sensing imagery (radiometric correction, geometric registration) and documents (OCR recognition to extract key information), adaptive algorithms are employed to optimize data quality.

[0078] 3) Fusion Processing Layer: Through three-level fusion—spatiotemporal benchmark unification, semantic association modeling, and business scenario adaptation—a unified data foundation for comprehensive land consolidation across the entire region is constructed. Spatiotemporal benchmark unification uses a unified coordinate system for the project area as a reference to correct the spatiotemporal coordinates of multi-source data. For data of different resolutions, methods such as sub-pixel decomposition and spatial interpolation are used to achieve multi-scale fusion. Semantic association modeling is based on knowledge graph technology to construct an ontology for the land consolidation domain (including core entities and relationships such as "project-plot-problem-measure-subject"), forming "plot-attribute-behavior" triples from scattered data through entity association, thus solving the problem of data semantic fragmentation. Business scenario adaptation dynamically assembles and fuses datasets to meet the business needs of the entire land consolidation cycle (planning-implementation-acceptance-maintenance). For example, in the "implementation phase," remote sensing imagery, sensor data, and mobile terminal reported data are fused to generate a multi-dimensional dynamic dataset of "project progress-engineering quality-ecological impact."

[0079] 4) Application Interface Layer: Through diverse interfaces such as RESTful APIs, spatial data services (WMS / WFS), and message queues (Kafka), data services are provided to upper-layer applications such as intelligent assessment and early warning monitoring, project management collaboration, and public services. This includes providing a unified data catalog (such as "Basic Project Information," "Real-time Monitoring Data," and "Historical Remediation Case Library"), metadata descriptions (data source, accuracy, update time), and access control. Simultaneously, it can output labeled sample datasets for intelligent assessment model training needs, push real-time or near-real-time dynamic data streams (such as "Project progress is lagging") for early warning monitoring needs, and output lightweight data packages (such as satellite image thumbnails of the project area) for mobile applications.

[0080] (2) Intelligent monitoring module (monitoring terminal);

[0081] To address the challenges faced by project managers in timely monitoring the continuous changes in land use, project progress, and funding during comprehensive land consolidation, as well as the "dynamic anomalies" in some projects, this module focuses on "full-element perception, full-cycle tracking, and intelligent analysis." It employs a closed-loop architecture of "perception-analysis-early warning-feedback," progressively capturing, analyzing, predicting, and intervening in dynamic changes from underlying data processing to upper-level intelligent applications. The module's technical architecture is divided into a feature indicator extraction layer, a dynamic change analysis layer, an assessment, early warning, and decision-making layer, and a feedback optimization layer. Each layer collaboratively supports the dynamic monitoring needs of the consolidation process.

[0082] 1) Feature Indicator Extraction Layer: Targeting core elements such as "land type change, project progress, ecological indicators, and problem risks" in the land consolidation process, this layer transforms project data from farmland consolidation, construction land consolidation, rural ecological protection and restoration, and rural landscape improvement into analyzable dynamic features. Specifically, based on the nature and objectives of the monitoring indicators, the indicator factors related to the objectives are aggregated and combined at different levels according to their affiliation, thereby forming a multi-objective, multi-level indicator system. Combining suggestions from experts in land, agriculture, forestry, and water resources departments, a comprehensive monitoring indicator system for land consolidation is constructed from aspects such as baseline evaluation, basic evaluation, and auxiliary evaluation. The selected feature indicators are quantified based on remote sensing image monitoring technology, 3D auxiliary technology, close-range photogrammetry monitoring technology, GNSS monitoring technology, field survey technology, automatic land type identification technology, and survey statistical analysis technology.

[0083] 2) Dynamic Change Analysis Layer: This layer is the core of this module. It primarily upgrades the remediation process from "phenomenon perception" to "causal analysis" through automated identification of land type changes and embedding business rules. Automated identification of land type changes is based on a pre-set deep learning network model to achieve pixel-level classification of land types (cultivated land, forest land, construction land, etc.). Then, it automatically identifies and extracts change features based on the time sequence before, during, and after remediation. Business rule embedding analysis mainly uses a built-in land remediation policy standard library (such as the "Land Remediation Project Management Measures" and "Cultivated Land Quality Grades") to transform business rules into analytical logic, enabling the monitoring of feature indicators extracted from the previous layer. For example, when it detects that "the cultivated land restoration area in a certain project area has not reached 80% of the planned area," it automatically associates data such as "construction machinery input" and "capital investment" to pinpoint the cause of the delay.

[0084] 3) Assessment and Early Warning Decision-Making Level: Combining dynamic analysis results with preset thresholds, a three-tiered early warning system ("yellow-orange-red") is constructed to achieve "early detection and rapid response" to problems. Anomaly warnings (yellow) primarily target short-term sudden changes, such as "a plot of land suddenly changing to construction land" or "sensor data jumping by more than 20%." Warnings are triggered through real-time calculations and pushed to the project manager, requiring verification within 48 hours. Risk warnings (orange) primarily target trend deterioration, such as "ecological benefit indicators falling below expected values ​​by 15%." The risk level is assessed based on business rules and pushed to the management department, requiring the development of intervention plans, such as adjusting irrigation methods or strengthening protective engineering. Major Risk Warnings (red) target systemic problems (such as "the overlap between the remediation area and permanent basic farmland exceeds 5%" or "project progress lagging by more than 30% may affect acceptance"). Alternative solutions are generated, and construction progress is accelerated through adjustments to land use layout, etc., and simultaneously pushed to the relevant authorities.

[0085] 4) Feedback Optimization Layer: Through the data feedback mechanism, the monitoring capabilities are continuously optimized. The early warning results are compared with the actual rectification situation (such as "illegal buildings on a certain plot of land have been demolished") to evaluate the accuracy of the early warning. At the same time, based on the verification results, the AI ​​model parameters (such as adjusting the feature weights of the land classification model) or the rule base (such as supplementing the identification rules for "new types of illegal behaviors") are updated, and typical problems (such as "the restoration of farmland in a certain area is slow due to slope factors") are entered into the case library to provide experience reference for subsequent projects.

[0086] (3) Decision support module (management end);

[0087] This module, centered on "data visualization, scenario-based decision-making, and collaborative operation," constructs a visualization platform and intelligent decision support toolset covering the entire lifecycle of "planning-implementation-acceptance-maintenance." It transforms abstract data into intuitive graphical language and decision logic into actionable scenario guidance, providing project managers, technical teams, and regulatory departments with "visualized, simulated, and traceable" decision-making tools. It serves as the application hub for the system to achieve "scientific management and precise policy implementation." The module adopts a four-layer architecture of "data-rendering-interaction-decision-making." By integrating full-cycle data assets, integrating multi-dimensional visualization engines, and constructing a human-machine collaborative decision-making mechanism, it achieves dynamic visualization and management support for the entire lifecycle of remediation projects. The module architecture is divided into a data link layer, a visualization rendering layer, an interaction service layer, and a decision support layer, with each layer collaboratively supporting visualization and management decision-making needs.

[0088] 1) Data Link Layer: Seamlessly connects standardized data assets output by the "Intelligent Monitoring Module" and external extended data to generate "lightweight datasets" suitable for visualization (such as vector boundaries, raster images, time-series animations, and statistical charts). Simultaneously, it provides data subscription services through API interfaces, supporting the visualization module to obtain data within specific ranges, at specific times, and for specific indicators on demand, enabling unified access to multi-source visualization data.

[0089] 2) Visualization Rendering Layer: Based on high-performance rendering technologies such as WebGL and Unity 3D, combined with GIS spatial analysis capabilities, a 3D visualization system of "3D terrain + 2D thematic + timeline" is constructed, supporting multi-granularity display from macro to micro and from static to dynamic. Specifically, it includes 3D terrain visualization, 2D thematic visualization, and timeline dynamic visualization. 3D terrain visualization is mainly based on oblique photogrammetry or laser point cloud data to generate a high-precision 3D real-world model of the project area, overlaying information such as current land use status and planning layout to achieve a "bird's-eye view" and "drill-down to specific plots" three-dimensional browsing experience. Clicking on a plot allows viewing its "current land type, remediation goals, and project progress." 2D thematic visualization generates multiple thematic layers (such as "farmland restoration potential map," "ecologically sensitive area distribution map," and "project progress Gantt chart") through GIS overlay analysis, supporting dynamic layer switching and overlay display (e.g., simultaneously displaying conflicting areas between "planned farmland area" and "current construction land"). The timeline dynamic visualization uses the project timeline as the axis, integrating satellite image time-series comparison, sensor time-series data animation, and project progress animation (such as the full-process simulation of "land leveling - irrigation facility construction - crop planting"), to achieve "dynamic playback" and "future projection" of the remediation process.

[0090] 3) Interaction Service Layer: Configure general interaction functions and role customization functions to meet the needs of different roles such as managers, technical personnel and grassroots workers, so as to improve the usability and practicality of the visualization system. The general interactive tools include intelligent annotation (such as supporting manual / automatic annotation of key points, adding text, images, and video notes), spatial measurement (such as providing distance, area, and volume measurement tools, with results automatically linked to the database), comparative analysis (such as supporting visual comparison of multiple schemes and multiple time phases (before and after remediation), and intelligent search (such as supporting natural language search, "find farmland with a slope > 15° in the project area, the system automatically locates and highlights the plots that meet the criteria") and other toolsets. Simultaneously, this module also supports role customization. For managers, it provides a "dashboard" view, integrating dynamic dashboards of core indicators (such as "farmland restoration rate," "ecological compliance rate," and "fund utilization rate"), and supports one-click generation of report PPTs; for technical personnel, it provides a "professional analysis" view, overlaying engineering drawings (such as CAD design drawings) and sensor network topology diagrams, supporting data interface debugging and model parameter adjustment; for grassroots users, it provides a "simplified" mobile interface, allowing them to intuitively view the effects of their own land remediation through a 3D reality model, and supporting online feedback submission.

[0091] 4) Decision Support Layer: This layer deeply integrates visualization and intelligent analysis to build a closed loop of "visualized analysis conclusions, visualized decision-making paths, and visualized implementation results" to support full-cycle management decision-making.

[0092] During the planning phase of the remediation project, the decision support layer provides visualization for scheme comparison and compliance verification. This layer displays multiple planning alternatives (such as "Scheme A: Reclamation Area X + Ecological Zone Y", "Scheme B: Reclamation Area M + Ecological Zone N") in a three-dimensional overlay, simultaneously labeling indicators such as "arable land retention," "ecological red line conflict area," and "cost budget" for each scheme. Spatial overlay analysis is used to verify the compliance of the planning schemes, automatically marking conflict areas between the schemes and project boundaries to help quickly identify compliance risks.

[0093] During the implementation phase of the remediation project, the decision support layer provides visualization for progress tracking and problem handling. The project plan (such as milestones in the BIM model) is overlaid with actual progress (such as completed areas as shown by UAV aerial surveys), using colors to distinguish between "completed (green), in progress (yellow), and lagging (red)" processes (e.g., "Land leveling is 90% complete, irrigation facilities are only 50% complete"). For warnings of "illegal occupation of farmland" and "ecological damage," the system automatically links to a historical case database, marking "similar problem locations" on a 3D map (e.g., "Project XX was previously penalized in this area for temporary construction"), and pushing out a handling flowchart (e.g., "Issuance of rectification notice, demolition within 3 days, restoration of farmland").

[0094] During the project acceptance phase, the decision support layer facilitates the visualization of results presentation and benefit evaluation. Satellite imagery, sensor data, and villager feedback before and after the remediation are integrated into an "electronic archive of project results," supporting 360° panoramic viewing (e.g., "click on a plot to view the entire process from wasteland before remediation to construction during remediation and paddy fields after remediation"). Simultaneously, dynamic charts (e.g., "Trend Chart of Cultivated Land Area Growth" and "Bar Chart of Ecosystem Service Value Enhancement") and 3D models (e.g., "Simulation of Rice Planting on Newly Added Cultivated Land") visually demonstrate core achievements such as "cultivated land restoration rate," "ecological restoration area," and "farmer income increase ratio."

[0095] During the project maintenance phase, the decision support layer facilitates risk warning and visualization of maintenance strategies. "High-risk areas" (e.g., "a blocked irrigation canal causing farmland drought") are marked on a 3D map. Interactive tools recommend maintenance solutions (e.g., "canal repair cost of 50,000 yuan vs. loss of 100,000 yuan from abandonment") and suggest optimal maintenance plans (e.g., prioritizing the repair of high-value farmland facilities).

[0096] This system efficiently integrates multi-dimensional heterogeneous data from remote sensing, IoT, business management, and social sensing to construct a multi-source heterogeneous land consolidation data fusion solution for a full-domain, full-cycle thematic database. Simultaneously, the system introduces a real-time dynamic monitoring mechanism for comprehensive land consolidation projects across the entire region. By combining time-series remote sensing image analysis and real-time IoT data, it achieves automated and high-frequency monitoring of consolidation project progress, land use changes, and ecological environment elements. Furthermore, the system spatially and visually displays monitoring, assessment, and early warning results on a unified platform and provides specific decision support tools, thereby significantly improving the scientific rigor and efficiency of user decision-making.

[0097] Example 2

[0098] This embodiment provides a dynamic monitoring system for the entire life cycle of land consolidation across the entire region, comprising, in sequence: a multi-source data fusion module, an intelligent monitoring module, and a decision support module;

[0099] The multi-source data fusion module includes, in sequence, a data access layer, a cleaning and treatment layer, a fusion processing layer, and an application interface layer. The multi-source data fusion module is used to collect and fuse multi-source heterogeneous data related to comprehensive land consolidation across the entire region, and to construct a data base with unified spatiotemporal benchmarks and semantic associations.

[0100] The intelligent monitoring module comprises, in sequence: a feature index extraction layer, a dynamic change analysis layer, an evaluation and early warning layer, and a feedback optimization layer. The intelligent monitoring module is used to identify land use type changes, track project progress, and dynamically monitor indicators based on the data base and using preset business rules and machine learning models, thereby acquiring monitoring data in real time and generating early warning information.

[0101] The decision support module comprises, in sequence, a data link layer, a visualization rendering layer, an interactive service layer, and a decision support layer; the data link layer is connected to the dynamic change analysis layer and the evaluation and early warning layer respectively; the decision support module is used to visualize the monitoring data and early warning information, and at the same time provide users with interactive services and decision support.

[0102] In the specific implementation process, such as Figure 2 The diagram shown is the overall architecture of the system provided in this embodiment. This example is based on a pilot project for comprehensive land consolidation in a town. Based on the project characteristics and monitoring needs, the system monitors the three stages of the comprehensive land consolidation project: before, during, and after the consolidation. At the same time, it introduces automatic land parcel identification technology based on deep learning algorithms to solve key and difficult technical problems in the dynamic monitoring of farmland changes, providing good technical support for the real-scene 3D visualization monitoring work of the business department.

[0103] This system adopts a hybrid database architecture optimized for land consolidation and management scenarios. It fully leverages the advantages of different storage technologies to classify and summarize multi-dimensional data collected from the national land spatial information platform, remote sensing imagery, IoT sensors, drone aerial photography, business management systems, mobile patrol apps, and internet public opinion, based on data characteristics and application scenarios. Then, using DSF / HDSF distributed storage technology, the classified data is distributed and stored across multiple nodes, utilizing its high scalability, fault tolerance, and read / write performance advantages to achieve efficient data management and rapid retrieval. Figure 3 The diagram shown is a schematic of DSF / HDSF distributed storage technology.

[0104] This system constructs a comprehensive land consolidation thematic model library in a structured manner, including a basic data entity library, an indicator library with entity feature indicators, a rule library with operational rules, and a cognitive reasoning intelligent model. It dynamically integrates and logically associates multi-source data to form a "single map" data base with "consolidation project / plot" as the core, which is used to store the core business data of the entire life cycle that has been standardized, associated and integrated.

[0105] like Figure 4 As shown, during dynamic monitoring, this system, based on literature review, human-land coupling theory, field surveys, and expert consultation, focuses on farmland consolidation, construction land consolidation, rural ecological protection and restoration, and rural landscape improvement within land consolidation projects. It aims to improve the monitoring efficiency and effectiveness of the entire lifecycle of land consolidation projects. The system constructs a scientific, reasonable, objective, and comprehensive monitoring indicator system for the entire land consolidation area, with a focus on monitoring infrastructure conditions, natural resource conditions and land type distribution, the scope of the consolidation area, the area of ​​engineering construction, the completion status of engineering tasks, indicator status, and analyzing the achievement status of planning objectives.

[0106] Based on the nature and objectives of the monitoring indicators, the relevant indicator factors are aggregated and combined at different levels according to their hierarchical relationships, thus forming a multi-objective, multi-level indicator system. This system selects 41 indicators in 5 categories to construct a comprehensive land consolidation monitoring indicator system. Through remote sensing image monitoring technology, 3D-assisted technology, close-range photogrammetry monitoring technology, GNSS monitoring technology, field survey technology, automatic identification of cultivated land plots, survey statistical analysis technology, ArcGIS database technology, and Fragstats analysis technology, real-time monitoring of each indicator throughout its entire lifecycle is conducted. This monitoring spans the entire process of comprehensive land consolidation project monitoring, effectively controlling changes in project implementation progress, temporal and spatial aspects, quality, economic, social, and ecological benefits. This allows for a comprehensive and systematic reflection, analysis, and evaluation of land consolidation indicator content, timely identification of deviations, and correction of development directions, thereby improving the effective implementation rate of planning and promoting the achievement of monitoring objectives. The comprehensive land consolidation monitoring indicator system is shown in Table 1.

[0107] Table 1 Summary Table of Technical System for Comprehensive Land Consolidation Indicators

[0108]

[0109] To enable analysis of land use changes at different points in time across any time dimension, this system combines the advantages of deep convolutional neural networks and utilizes enhanced boundary extraction methods to detect the edges of land use parcels. This allows for efficient and accurate detection of land use parcel edges, improving the ability to automatically identify different land use types.

[0110] like Figure 5 As shown, to ensure the detection effect of small plots and to eliminate interference from irrelevant data such as roads, residential areas, and woodlands, paddy field patches are first superimposed as input. Then, the data of large plots are input into a deep learning edge detection network based on pixel interpolation to predict the edges of small plots within the large plots. Finally, based on the prediction results, image processing and computer vision algorithms are applied to gradually obtain a fused edge map, a binary edge map, an edge dilation map, an edge refinement map, and a closed region map, ultimately obtaining the complete small plot region and number.

[0111] In this embodiment, an improved PiDiNet model is used to extract land parcel edges. The structure of the improved PiDiNet model includes convolutional blocks 1, 2, 3, and 4 connected in sequence. Each convolutional block outputs feature maps for different channels. After channel-wise transformation of each feature map, they are input into an attention layer to obtain the corresponding attention features. The attention features corresponding to convolutional blocks 2-3 are then subjected to feature difference and concatenated with the attention features corresponding to convolutional block 1 to obtain the concatenated features. The concatenated features are then channel-wise transformed again, and finally, the Sigmoid function is used to obtain the detection results of the land parcel edges.

[0112] In this embodiment, to address the issue of complex and diverse edge directions of land parcels, Gabor convolutional layers are introduced into convolutional blocks 1 and 2 of the improved PiDiNet model. This structure mainly combines Gabor filters and learnable convolutional layers, and its structure is as follows: Figure 6As shown, the Gabor filter possesses the characteristic of achieving local optimization in both the spatial and frequency domains. Gabor convolutional layers efficiently and clearly extract multi-level features at different scales and directions, significantly enhancing the ability to perceive complex and diverse edges. Based on the Gabor filter's ability to extract edge features and its advantage in parameter update during backpropagation, this model can extract significant texture features of land parcel edges and geometric relationship features between different edges. By defining different convolutional channels for feature extraction, dimensionality reduction and encoding of high-dimensional features are performed, resulting in more discriminative edge features.

[0113] After obtaining the edge of the land parcel, a thresholding algorithm is used to make decisions on the pixels at the edge of the land parcel, directly removing some pixels that are below or above a certain value; then, morphological algorithms, such as dilation and erosion algorithms, are used to connect the discontinuous edge lines to obtain better recognition results; finally, skeleton extraction and contour assembly are performed. The above deep learning-based edge detection algorithm can detect the pixels of the contour boundary based on the differences between pixels, but it does not treat the contour as a whole, so it is still necessary to assemble these edge pixels into a contour.

[0114] Furthermore, to verify the beneficial effects of the improved PiDiNet model in this system, a verification experiment was conducted in this embodiment. First, farmland samples were collected for the experimental data, and ArcGIS software was used to draw the samples, resulting in 100 image tiles of 512×512 pixels each. Subsequently, data augmentation operations such as flipping and rotation were performed to expand the size to 500. 300 images were randomly selected as the training set, 100 as the validation set, and 100 as the test set. During the training phase, each high-resolution image had two labels: farmland texture and farmland edge. The texture label was a binary image, with 0 representing the background and 1 representing farmland. The farmland edge label was a single-pixel wide line. In the land parcel detection task, the optimized algorithm can identify most land parcels with generally accurate edges. The geometry of the identified land parcels basically matches the original image. It performs well in both homogeneous and heterogeneous land parcels and has strong recognition capabilities in densely populated land parcels. Figures 7-9 As shown.

[0115] All accuracy metrics for boundary detection are calculated based on the confusion matrix. The confusion matrix represents the classification result of the statistical classification model, with the horizontal and vertical axes representing the number of pixels corresponding to the true category and the number of pixels corresponding to the predicted category. In the segmentation results, for the land parcel classification problem, categories are divided into positive and negative examples. Correct predictions by the classifier are denoted as true (T), and incorrect predictions are denoted as negative (F). These four basic combinations constitute the four basic elements of the confusion matrix. The accuracy calculation process is as follows:

[0116]

[0117]

[0118] Among them, TruePositive (TP) indicates that the model predicts a positive example and the ground truth is a positive example, representing the correctly identified boundary; FalsePositive (FP) indicates that the model predicts a positive example and the ground truth is a negative example, representing the erroneous part of the detection result; FalseNegative (FN) indicates that the model predicts a negative example and the ground truth is a negative example, representing the unidentified part of the true boundary; TrueNegative (TN) indicates that the model predicts a positive example and the ground truth is a negative example; F-score serves as a global boundary accuracy metric.

[0119] Furthermore, the calculation process of ODS (optimal dataset scale) is as follows: a fixed threshold β is selected and applied to all images to maximize the F-score across the entire dataset. This β is the ODS. OIS (optimal image scale) means that a different β is selected for each image to maximize the F-score of that image. This β is the OIS.

[0120] In this system, the improved algorithm shows significant improvements across four metrics, particularly in the average accuracy representing boundaries, which increased by 1.7 percentage points, and the accuracy of plot identification, which improved to 0.856. This enables accurate detection of plots in most cultivated land areas, as shown in Table 2.

[0121] Table 2 Comparison of Test Accuracy

[0122]

[0123] Based on the aforementioned machine learning algorithms, changes in natural resource categories within a specific area over a given time period can be quickly identified. This assists in the monitoring and evaluation of comprehensive land consolidation across the entire region, enabling regular monitoring of land use changes before and after consolidation. For example, by extracting changes in cultivated land in the consolidation area and controlling automatic camera patrols within the area for verification, "non-grain" plots identified through monitoring and analysis (including converted cultivated land to forest, grassland, construction land, and unused land) are marked, forming dynamic monitoring results of cultivated land for grain production. This provides an important reference for evaluating the effectiveness of comprehensive land consolidation projects across the entire region.

[0124] This system also compares the monitoring data of each indicator with the preset threshold to achieve tiered early warning and generate early warning information; after rectification is completed, it receives the actual rectification status and compares it with the early warning information, and optimizes the above machine learning model based on the comparison results.

[0125] This system also includes a decision support module, which enables real-time display and monitoring of functions such as comprehensive situation assessment, project management, dynamic monitoring, results presentation, and system management. By utilizing the various functional modules of this system, users can be provided with scientific and reasonable decision support.

[0126] Example 3

[0127] like Figure 10 As shown, this embodiment provides a dynamic monitoring method for the entire life cycle of comprehensive land consolidation, based on the system in Embodiment 1 or 2, and includes the following steps:

[0128] S1: The multi-source data fusion module collects and merges multi-source heterogeneous data related to comprehensive land consolidation across the entire region to construct a data base with unified spatiotemporal benchmarks and semantic associations;

[0129] S2: The intelligent monitoring module, based on the data base, uses preset business rules and machine learning models to identify changes in land use types, track project progress, and dynamically monitor indicators, acquire monitoring data in real time, and generate early warning information;

[0130] S3: The decision support module visualizes the monitoring data and early warning information, and provides users with interactive services and decision support.

[0131] In the specific implementation process, firstly, the multi-source data fusion module is used to collect and integrate multi-source heterogeneous data related to comprehensive land consolidation across the entire region, constructing a data base with unified spatiotemporal benchmarks and semantic associations; then, based on the data base, the intelligent monitoring module uses preset business rules and machine learning models to identify changes in land use types, track project progress, and dynamically monitor indicators, acquiring monitoring data in real time and generating early warning information; finally, the decision support module visualizes the monitoring data and early warning information, while providing users with interactive services and decision support.

[0132] This method can efficiently integrate multi-source heterogeneous data, enabling dynamic monitoring of the entire process of land consolidation projects across the entire region, intelligent effectiveness evaluation and risk warning, and providing strong support for management decisions.

[0133] The same or similar labels correspond to the same or similar parts;

[0134] The terms used to describe positional relationships in the accompanying drawings are for illustrative purposes only and should not be construed as limiting this application.

[0135] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the implementation of the present invention. Those skilled in the art can make other variations or modifications based on the above description. It is neither necessary nor possible to exhaustively describe all embodiments here. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A dynamic monitoring system for the entire lifecycle of comprehensive land consolidation, characterized in that, It includes, in sequence: a multi-source data fusion module, an intelligent monitoring module, and a decision support module; The multi-source data fusion module includes, in sequence, a data access layer, a cleaning and treatment layer, a fusion processing layer, and an application interface layer. The multi-source data fusion module is used to collect and fuse multi-source heterogeneous data related to comprehensive land consolidation across the entire region, and to construct a data base with unified spatiotemporal benchmarks and semantic associations. The intelligent monitoring module comprises, in sequence: a feature index extraction layer, a dynamic change analysis layer, an evaluation and early warning layer, and a feedback optimization layer. The intelligent monitoring module is used to identify land use type changes, track project progress, and dynamically monitor indicators based on the data base and using preset business rules and machine learning models, thereby acquiring monitoring data in real time and generating early warning information. The feature index extraction layer is used to extract several monitoring indicators from the data base and perform feature quantification processing based on the pre-constructed monitoring indicator system for comprehensive land consolidation. The dynamic change analysis layer is used to automatically identify land parcels based on a preset machine learning model, further identify changes in land use types, and extract change data; at the same time, it is used to track project progress and dynamically monitor indicators based on preset business rules, and acquire monitoring data in real time. The assessment and early warning layer is used to compare the monitoring data of each indicator with the preset threshold to achieve graded early warning and generate early warning information; The feedback optimization layer is used to receive the actual rectification situation, compare it with the early warning information, and optimize the machine learning model and business rules based on the comparison results; In the feature index extraction layer, a monitoring index system for comprehensive land consolidation is constructed based on the analytic hierarchy process. The index system includes at least: a target layer, a criterion layer, and an index layer. The target layer includes: basic bottom-line constraints, improvement of farmland ecosystem service functions, ecological protection and restoration and improvement of rural landscape, project and fund coordination and project supervision; The basic bottom-line constraints in the aforementioned criteria layer include: quantity control, quality control, and other controls; improvement of farmland ecosystem service functions include: the degree of farmland concentration and contiguousness, the degree of intensive and economical use of construction land, and land transfer status; ecological protection and restoration and rural landscape improvement include: the status of ecological protection and restoration and the status of rural landscape improvement; project and funding coordination includes: the status of funding guarantee for improvement, the progress of project completion, and the project quality objectives; project supervision includes: innovation and illegal monitoring. In the indicator layer, quantity control includes: the proportion of newly added arable land area, the proportion of newly added permanent basic farmland area, and the surplus construction land indicator; quality control includes: improving the quality grade of arable land; other controls include: the area conflicting with ecological protection red lines and the area protected for historical and cultural heritage; the degree of farmland concentration and contiguousness includes: the area of ​​newly added high-standard farmland, the area of ​​developed and supplemented arable land, the area of ​​reclaimed paddy fields, the number of small arable land plots converted into large arable land plots, and the degree of increase in the concentration and contiguousness of permanent basic farmland; the degree of intensive and economical use of construction land includes: the proportion of land use quota linked to land increase and decrease used for infrastructure construction, the scale of redevelopment of inefficient construction land, the scale of demolition and reclamation of rural construction land, the rate of decrease in the average household homestead area, and the scale of reduction in village construction land; land transfer includes: the area of ​​arable land used for grain cultivation after remediation, the introduction of agricultural enterprises, the number of large grain growers, and the proportion of land management rights transferred; ecological The protection and restoration status includes: the area of ​​comprehensive mine management, the area of ​​comprehensive water environment management, the area of ​​forest land transformation, the area of ​​mangrove protection and restoration, the area of ​​coastal zone remediation, the number of kilometers of newly added ecological corridors, and the soil and water conservation rate; the rural landscape improvement status includes: the number of rural living environment improvements, the coverage rate of sewage treatment facilities, the coverage rate of garbage disposal, and the popularization rate of sanitary toilets; the remediation funding guarantee status includes: the proportion of integrated agricultural funds, the completion rate of budgeted investment funds, the actual expenditure rate of funds, and the proportion of social capital investment; the project completion progress status includes: the completion rate of agricultural land remediation projects, the completion rate of construction land remediation projects, the completion rate of ecological protection and restoration projects, the completion rate of cultural protection and rural landscape improvement projects, and other projects; the project quality target status includes: the project quality pass rate; innovation and illegal monitoring includes: the number of innovative systems and the area of ​​illegal land use; In the dynamic change analysis layer, the preset machine learning model is specifically an improved PiDiNet model, used for plot edge extraction; The improved PiDiNet model structure includes convolutional blocks 1, 2, 3, and 4 connected in sequence. Each convolutional block outputs feature maps for different channels. After channel-wise transformation of each feature map, they are input into an attention layer to obtain corresponding attention features. The attention features corresponding to convolutional blocks 2-3 are then subjected to feature difference and concatenated with the attention features corresponding to convolutional block 1 to obtain concatenated features. The concatenated features are then channel-wise transformed again, and finally, the Sigmoid function is used to obtain the detection results of the plot edges. Convolutional block 1 and convolutional block 2 have the same structure, both including a Gabor convolutional layer, a batch normalization layer, a max pooling layer and an activation layer connected in sequence; convolutional block 3 and convolutional block 4 have the same structure, both including a convolutional layer, a batch normalization layer, a max pooling layer and an activation layer connected in sequence. The decision support module comprises, in sequence, a data link layer, a visualization rendering layer, an interactive service layer, and a decision support layer; the data link layer is connected to the dynamic change analysis layer and the evaluation and early warning layer respectively; the decision support module is used to visualize the monitoring data and early warning information, and at the same time provide users with interactive services and decision support.

2. The full life-cycle dynamic monitoring system for comprehensive land consolidation according to claim 1, characterized in that, In the multi-source data fusion module The data collected by the data access layer includes at least one or more of the following: satellite remote sensing imagery, UAV aerial survey data, ground sensor data, business database and document / table data, and publicly available Internet data; the data collected by the data access layer is classified, and the classified data is distributed and stored across multiple nodes using DSF / HDSF distributed storage technology; The cleaning and governance layer is used to clean the data collected by the data access layer based on predefined cleaning rules. The data cleaning includes at least one or more of the following: outlier marking and correction, missing value imputation, and duplicate value removal. The fusion processing layer is used to sequentially perform spatiotemporal benchmark unification, semantic association modeling, and business scenario adaptation processing on the cleaned data to generate the data baseboard. The spatiotemporal benchmark unification includes unifying and correcting the time and spatial coordinates of multi-source heterogeneous data. The semantic association modeling includes constructing a knowledge graph based on the spatiotemporally benchmark-unified data and performing entity semantic association. The business scenario adaptation includes dynamically packaging the semantically association-modeled data into several datasets adapted to different stages according to the data differences required at different stages of comprehensive land consolidation, and jointly saving the datasets corresponding to all stages as the data baseboard. The application interface layer is used to output the data baseboard to the intelligent monitoring module through preset multi-interfaces.

3. The full life-cycle dynamic monitoring system for comprehensive land consolidation according to claim 1, characterized in that, The Gabor convolutional layer includes at least one Gabor filter and a learnable convolutional kernel.

4. The full life-cycle dynamic monitoring system for comprehensive land consolidation according to claim 1, characterized in that, In the decision support module The data link layer is used to collect the monitoring data and early warning information in real time; The visualization rendering layer is used to realize three-dimensional terrain visualization, two-dimensional thematic visualization, and dynamic timeline visualization based on monitoring data and early warning information; The three-dimensional terrain visualization includes: constructing a three-dimensional real-scene model of the entire land consolidation project area, overlaying monitoring data and early warning information of each plot, and realizing three-dimensional information browsing; the two-dimensional thematic visualization uses an external GIS system to overlay and analyze monitoring data and early warning information to generate multiple thematic layers; the timeline dynamic visualization uses the timeline of the entire land consolidation project as the axis, combined with monitoring data and early warning information, to realize dynamic playback and future projection of the consolidation process. The interaction service layer is used to provide interaction services to users, including general services and customized services; The decision support layer is used to provide decision support for users.

5. A full life-cycle dynamic monitoring system for comprehensive land consolidation according to claim 4, characterized in that, In the visualization rendering layer, a three-dimensional real-scene model of the entire land consolidation project area is constructed based on WebGL or Unity 3D engine.

6. The full life-cycle dynamic monitoring system for comprehensive land consolidation according to claim 4, characterized in that, In the interactive service layer, general services include: intelligent data annotation, spatial measurement, comparative analysis, and intelligent search; The customized services include: for managers, a cockpit view with dynamic dashboards integrating core indicators; for technical personnel, a professional analysis view with overlaid engineering drawings and sensor network topology diagrams for data interface debugging and model parameter adjustment; and for grassroots users, a simplified interface that allows them to intuitively view the remediation effects of their own land through a 3D real-scene model and provides online feedback functionality.

7. A method for dynamic monitoring of the entire life cycle of land consolidation across the entire region, based on the system described in any one of claims 1 to 6, characterized in that, Includes the following steps: S1: The multi-source data fusion module collects and merges multi-source heterogeneous data related to comprehensive land consolidation across the entire region to construct a data base with unified spatiotemporal benchmarks and semantic associations; S2: The intelligent monitoring module, based on the data base, uses preset business rules and machine learning models to identify changes in land use types, track project progress, and dynamically monitor indicators, acquire monitoring data in real time, and generate early warning information; S3: The decision support module visualizes the monitoring data and early warning information, and provides users with interactive services and decision support.

Images