Multi-source marine data intelligent supervision method, system, device and storage medium

Abstract

The present invention relates to a multi-source sea area data intelligent supervision method, system, device and storage medium. The method comprises: obtaining sea area geospatial data and historical sea area supervision data of a target sea area, and performing monitoring construction to obtain an island monitoring grid; obtaining multi-source monitoring data of the target sea area based on the island monitoring grid, and performing grid data recognition to obtain a global dynamic data set; extracting coordinates of suspected specific target areas from the global dynamic data set and calculating coordinate specific target confidence; performing range comparison on the coordinates of the suspected specific target areas, and performing behavior evaluation in combination with the coordinate specific target confidence to obtain a sea area specific target early warning report.

Description

Technical Field

[0001] The present invention relates to the technical field of sea area supervision, and in particular to a method, system, device and storage medium for intelligent supervision of multi-source sea area data. Background Art

[0002] With the rapid development of multi-sensor technologies such as satellite remote sensing and drone aerial photography, the amount of multi-source, heterogeneous marine geospatial data has exploded, providing a data foundation for intelligent supervision. However, existing technologies often rely on static thresholds, which can easily lead to false positives and false negatives. Furthermore, existing technologies struggle to distinguish between legitimate construction activities and marine uses. Given the increasing international emphasis on marine ecological protection, the ability to accurately identify and provide early warnings for specific target behaviors through multi-source data fusion has become a key issue in marine resource management. Summary of the Invention

[0003] The main purpose of the present invention is to provide a multi-source sea area data intelligent supervision method, system, device and storage medium, which can effectively distinguish between legal construction and specific target behaviors, reduce supervision misjudgments, and improve the accuracy of aerial photography decisions.

[0004] To achieve the above objectives, the present invention provides a method for intelligent monitoring of multi-source marine data, comprising:

[0005] Obtain the maritime geospatial data and historical maritime supervision data of the target sea area, conduct monitoring and construction, and obtain the island monitoring grid;

[0006] Acquire multi-source monitoring data of the target sea area based on the island monitoring grid, perform grid data recognition, and obtain a global dynamic data set;

[0007] Extracting the coordinates of the suspected specific target area of ​​the global spot dynamic data set and calculating the specific target confidence of the coordinates;

[0008] The coordinates of the suspected specific target area are compared in range, and a behavior assessment is performed in combination with the specific target confidence of the coordinates to obtain a sea area specific target early warning report.

[0009] Furthermore, the acquisition of the target sea area's geospatial data and historical sea area supervision data, and the construction of monitoring to obtain an island monitoring grid, includes:

[0010] Acquiring sea area boundary data and geographic information based on the target sea area, and performing spatial construction to obtain the sea area geographic spatial data;

[0011] Obtaining marine aerial photography records and manual inspection reports of the target sea area, conducting sea use behavior analysis, and obtaining the historical sea area supervision data;

[0012] Spatially superimposing the sea area geographic spatial data according to the historical sea area supervision data, and performing initial grid setting in combination with a preset grid division threshold to obtain an initial monitoring grid unit;

[0013] The initial monitoring grid unit is subjected to graph iterative optimization according to the resource deployment constraint conditions of the target sea area to obtain the island monitoring grid.

[0014] Furthermore, the multi-source monitoring data of the target sea area is obtained based on the island monitoring grid, and grid data recognition is performed to obtain a global dynamic data set, including:

[0015] Planning multi-source monitoring points for the island monitoring grid to obtain topological locations of monitoring nodes;

[0016] Acquire multispectral remote sensing images, drone aerial photography data, and nearshore monitoring data of the target sea area according to the topological positions of the monitoring nodes, and perform multi-source alignment and integration to obtain the multi-source monitoring data;

[0017] Performing ground object classification processing on the multi-source monitoring data based on a preset ground object classification rule library to obtain a gridded ground object label distribution map;

[0018] Performing regional edge detection and segmentation on the gridded feature label distribution map to obtain independent patch boundary vector data;

[0019] Global matching is performed on the independent patch boundary vector data according to the historical patch database to obtain the global patch dynamic data set.

[0020] Furthermore, the multispectral remote sensing images, drone aerial data and nearshore monitoring data of the target sea area are obtained according to the topological positions of the monitoring nodes, and multi-source alignment and integration are performed to obtain the multi-source monitoring data, including:

[0021] remote sensing image acquisition is performed according to the topological position of the monitoring node, and atmospheric and terrain errors are eliminated to obtain the multispectral remote sensing image;

[0022] Performing UAV aerial photography planning based on the topological positions of the monitoring nodes to obtain the UAV aerial photography data;

[0023] Deploy nearshore monitoring nodes according to the topological positions of the monitoring nodes, collect wave monitoring data, tide level monitoring data, and water quality monitoring data, and integrate the nearshore data to obtain the nearshore monitoring data;

[0024] Performing aerial triangulation and position registration on the UAV aerial data based on the topological positions of the monitoring nodes to obtain an aerial image dataset;

[0025] Multi-source registration is performed on the multispectral remote sensing image, the aerial image dataset, and the nearshore monitoring data to obtain the multi-source monitoring data.

[0026] Furthermore, extracting the coordinates of the suspected specific target area of ​​the global spot dynamic data set and calculating the specific target confidence of the coordinates includes:

[0027] According to the marine ecological protection rules, the dynamic dataset of the global map patches is subjected to a spatial overlay analysis of the protected areas to obtain an initial set of specific target areas;

[0028] Performing regional change detection on the initial specific target area set based on the global patch dynamic data set, and identifying abnormal areas to obtain the outline of the suspected specific target area;

[0029] Performing spatial coordinate conversion on the outline of the suspected specific target area according to the sea area geographic spatial data to obtain the coordinates of the suspected specific target area;

[0030] Calculating the probability of a specific target in the area of ​​the suspected specific target based on the historical sea area supervision data to obtain a preliminary confidence level;

[0031] The specific target level of the preliminary confidence is calculated according to the marine ecological protection rules to obtain the specific target confidence of the coordinates.

[0032] Furthermore, the coordinates of the suspected specific target area are compared, and a behavior assessment is performed in combination with the specific target confidence of the coordinates to obtain a sea area specific target early warning report, including:

[0033] The legal sea use spatial range analysis is performed on the coordinates of the suspected specific target area according to the preset sea use project database to obtain the abnormal coordinate set of the permitted coverage area and the specific target coordinate set of the unlicensed area;

[0034] Based on the sea use permit information in the sea use project database, regional sea use type matching is performed on the abnormal coordinate set of the permit coverage area to obtain a sea use compliance determination result;

[0035] Performing regional coordinate screening on the unlicensed area specific target coordinate set according to a preset specific target confidence threshold to obtain a high-confidence specific target coordinate set;

[0036] performing specific target behavior pattern matching on the high-confidence specific target coordinate set according to the historical sea area supervision data to obtain a specific target behavior type level;

[0037] A specific target assessment is performed on the sea use compliance determination result, the high-confidence specific target coordinate set and the specific target behavior type level to obtain the sea area specific target early warning report.

[0038] Furthermore, the legal sea use spatial range analysis is performed on the coordinates of the suspected specific target area according to the preset sea use project database to obtain the abnormal coordinate set of the permitted coverage area and the specific target coordinate set of the unlicensed area, including:

[0039] Performing polygonal spatial boundary calculation on the coordinates of the suspected specific target area based on the legal sea use scope data in the sea use project database to obtain a preliminary screening coordinate set;

[0040] Performing spatial relationship and time validity filtering on the preliminary screening coordinate set to obtain an expired permission coordinate set and a valid permission coverage coordinate set;

[0041] Performing a comparison of the permitted activity type attributes on the valid permitted coverage coordinate set to obtain the permitted coverage area abnormal coordinate set;

[0042] The preliminary screening coordinate set and the expired license coordinate set are clustered into non-licensed areas according to the legal sea use range data to obtain the non-licensed area specific target coordinate set.

[0043] The present invention further provides a multi-source marine data intelligent monitoring system, which is applied to any of the multi-source marine data intelligent monitoring methods described above, comprising:

[0044] The acquisition module is used to obtain the sea area geospatial data and historical sea area supervision data of the target sea area, and to perform monitoring construction to obtain the island monitoring grid;

[0045] An analysis module, configured to obtain multi-source monitoring data of the target sea area based on the island monitoring grid, and perform grid data recognition to obtain a global dynamic data set;

[0046] An association module, the association module is used to extract the coordinates of the suspected specific target area of ​​the global spot dynamic data set and calculate the specific target confidence of the coordinates;

[0047] A processing module is used to perform range comparison on the coordinates of the suspected specific target area, and to perform behavior evaluation based on the confidence level of the specific target of the coordinates to obtain a sea area specific target early warning report.

[0048] The present invention also provides a multi-source sea area data intelligent monitoring device, comprising:

[0049] Memory, used to store programs;

[0050] The processor is used to execute the program to implement each step of the multi-source sea area data intelligent supervision method described in any one of the above.

[0051] The present invention also provides a storage medium storing computer instructions, wherein the computer instructions are used to enable a computer to execute any of the above methods.

[0052] The present invention provides a method, system, device, and storage medium for intelligent monitoring of multi-source maritime data, which have the following beneficial effects:

[0053] By constructing an island monitoring grid and integrating multi-source monitoring data, dynamic perception of changes in the entire sea area and islands is achieved, overcoming the lag problem of traditional single remote sensing or manual inspections and significantly improving the ability to detect behavior early. Confidence calculations are performed on suspected specific target areas to avoid the misjudgment and omission problems of static threshold judgments, making the identification of specific target behavior more scientific and reliable. By combining the sea use project database with real-time monitoring data for range comparison, it is possible to effectively distinguish between legal construction and specific target behavior, reduce regulatory misjudgments, and improve the accuracy of aerial photography decisions. By comprehensively analyzing the specific target confidence and behavior assessment results, an accurate sea area specific target early warning report is generated, providing regulatory authorities with a data-driven decision-making basis, improving regulatory response speed and governance effectiveness. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 This is a flow chart of a multi-source sea area data intelligent supervision method provided by the present invention;

[0055] Figure 2 This is a structural diagram of a multi-source marine data intelligent monitoring system provided by the present invention;

[0056] Figure 3 This is a structural diagram of a multi-source sea area data intelligent monitoring device provided by the present invention.

[0057] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0058] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0059] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0060] Reference Figure 1 As shown, the present invention provides a multi-source marine data intelligent supervision method, comprising:

[0061] Step S101: Acquire the sea area geospatial data and historical sea area supervision data of the target sea area, and perform monitoring construction to obtain an island monitoring grid;

[0062] Step S201: Acquire multi-source monitoring data of the target sea area based on the island monitoring grid, perform grid data recognition, and obtain a dynamic dataset of global patches;

[0063] Step S301: extracting the coordinates of the suspected specific target area of ​​the global patch dynamic data set and calculating the specific target confidence of the coordinates;

[0064] Step S401: performing range comparison on the coordinates of the suspected specific target area, and performing behavior evaluation based on the specific target confidence of the coordinates to obtain a sea area specific target warning report.

[0065] Based on the above steps, the detailed steps are as follows:

[0066] Step S101:

[0067] Geospatial data includes vector or raster data such as island topography, shoreline changes, and current sea area use, typically derived from satellite remote sensing, drone aerial photography, or marine surveying and mapping. Historical sea area supervision data is extracted from aerial photography records, supervision information, or manual inspection reports, marking the spatial location and timestamp of events such as specific target sea use and illegal land reclamation. The construction of the monitoring grid uses geographic gridding technology to divide the target sea area into regular or irregular grid cells, each grid associated with geographic attributes, ecological sensitivity indicators, and the probability of historical specific targets. Grid division must consider the balance between data resolution and computational efficiency, such as using quadtree or hexagonal grids to accommodate analysis at different scales. Through spatial overlay analysis, historical hotspots are associated with grid attributes to form an island monitoring grid with risk identification, providing a spatial benchmark for subsequent dynamic monitoring. The grid data output by this step implicitly reflects the regional risk level, which directly affects the collection priority and recognition accuracy of the global dynamic dataset.

[0068] Step S201:

[0069] Based on the spatial framework of the island monitoring grid, this step focuses on the real-time collection and dynamic identification of multi-source monitoring data. Multi-source monitoring data includes heterogeneous sources such as satellite remote sensing imagery, ship AIS tracks, offshore radar signals, and buoy sensor data. Through spatiotemporal alignment and grid mapping, the raw data is converted into feature vectors within grid cells. For example, satellite imagery uses semantic segmentation to extract island shoreline change patterns, AIS data analyzes unusual ship berthing coordinates, and radar signals detect unreported offshore structures. Grid data recognition uses deep learning models (such as U-Net or YOLOv5) to automatically classify patterns, distinguishing between legitimate maritime activities and potential specific targets. The output is a dynamic dataset containing spatiotemporal coordinates, type labels, and intensity of change. The dynamic nature of global patterns is reflected in the detection of differences in time series, such as identifying newly reclaimed areas by comparing previous and later images. This step relies on the appropriateness of the grid division. A coarse-grained monitoring grid can lead to missed detection of small-scale specific target patterns. The resulting dynamic dataset provides an initial screening result for suspected specific target areas, which is then used to calculate confidence scores for the suspected specific target areas.

[0070] Step S301:

[0071] Pre-set rules include spatial constraints such as ecological redline boundaries, restrictions on sea use (e.g., prohibited aquaculture zones), and the extent of core nature reserves. GIS spatial analysis is used to extract coordinates of patches that conflict with these rules. For example, construction patches appearing within a mangrove reserve are flagged as high-suspicion targets. Confidence calculations incorporate multiple lines of evidence: the overlap ratio between the patch and the reserve, the historical frequency of similar specific targets, and the reliability of sensor data (e.g., cross-validation of radar and optical imagery). A Bayesian network or random forest model is used to quantify the probability of a specific target and output a specific target confidence score for the coordinate. High-scoring areas are associated with illegal land reclamation or illegal sand mining, while low-scoring areas require further manual verification. The accuracy of this step is affected by the quality of the dynamic dataset. If the data is noisy (e.g., due to cloud cover), the confidence level must be corrected using time-series data backtracking. The output high-confidence coordinates serve as direct input for the marine specific target early warning report.

[0072] Step S401:

[0073] The marine project database stores the approved coordinates, uses, and validity periods of legal marine use projects. Spatial connectivity analysis (such as overlay analysis or buffer analysis) verifies whether suspected coordinates fall within the approved scope. Unmatched coordinates are graded based on their confidence level: high-confidence coordinates trigger a direct red alert, while medium- and low-confidence coordinates are linked to vessel registration information or corporate credit records for secondary filtering. Behavioral assessment incorporates a temporal dimension. For example, the association of short-term high-frequency AIS signals with image changes suggests poaching. Warning reports contain specific target types (such as illegal aquaculture), spatial locations, confidence levels, and optimization recommendations, and are automatically pushed to aerial photography terminals. This step relies on confidence calibration; setting the threshold too high can result in missed alerts. The frequency of dynamic data updates determines the timeliness of warnings. Once generated, the reports are fed back to the monitoring grid to optimize historical distribution data, creating a closed-loop management system.

[0074] The present invention provides a multi-source sea area data intelligent supervision method, which realizes the dynamic perception of the changes in the entire sea area and islands by constructing an island monitoring grid and integrating multi-source monitoring data, overcomes the lag problem of traditional single remote sensing or manual inspection, and significantly improves the early detection capability of behavior. Confidence calculation is performed on suspected specific target areas to avoid the misjudgment and omission problems of static threshold judgment, making the identification of specific target behavior more scientific and reliable. By combining the sea use project database with real-time monitoring data for range comparison, it is possible to effectively distinguish between legal construction and specific target behavior, reduce regulatory misjudgments, and improve the accuracy of aerial photography decisions. By comprehensively analyzing the specific target confidence and behavior assessment results, an accurate sea area specific target early warning report is generated, providing data-driven decision-making basis for regulatory authorities, improving regulatory response speed and governance efficiency.

[0075] In one embodiment, the sea area geospatial data and historical sea area supervision data of the target sea area are obtained, and monitoring is constructed to obtain an island monitoring grid, including:

[0076] Maritime boundary data and geographic information are fundamental elements in constructing geospatial data. Maritime boundary data is derived from official nautical charts, remote sensing imagery, or electronic navigational charts provided by marine surveying and mapping agencies. It includes vector data such as coastlines, island outlines, and territorial sea baselines. Geographic information encompasses dynamic hydrological data such as seabed topography, water depth, tides, and ocean currents, as well as fixed features such as reefs, waterways, and anchorages. This data is spatially registered and coordinate-system standardized using a geographic information system (GIS) platform, ensuring that data from different sources is aligned within the same projection coordinate system.

[0077] The original sea area vector data is topologically verified to eliminate geometric errors and generate a continuous sea area surface layer and a discrete island point layer. The water depth sampling point data is converted into a regular grid surface to create a 3D water depth digital elevation model. The sea area surface layer and the water depth digital elevation model are spatially overlaid, and the island point layer is integrated to form a 3D terrain model. The generated 3D terrain model data is stored hierarchically by boundary layer, terrain layer, and feature layer, creating a structured database that supports spatial analysis.

[0078] Marine aerial photography records come from the case database, including time, location, type (such as illegal fishing, dumping, and cross-border navigation) and penalty results. Manual inspection reports are filled out by island personnel or patrol boats to record non-case incidents such as suspicious activities and equipment damage. The two types of data are cleaned through spatiotemporal matching and standardization to eliminate duplicate or invalid records. Behavioral analysis uses the kernel density estimation method to convert discrete case points into a continuous spatial probability distribution surface to identify high-frequency hotspot areas. For different types of behavior, their spatial autocorrelation is calculated separately to verify whether the events show clustering characteristics. The historical sea area supervision data is finally output as a raster layer. Each pixel value represents the risk intensity of the location, while retaining the type classification attributes to provide a risk weight basis for subsequent grid division.

[0079] The spatial overlay operation is achieved through the weighted overlay tool, which performs algebraic operations on the distribution grid and the boundary, terrain and other layers in the geospatial data. High-risk areas are given higher weights and are superimposed with dangerous sea areas with water depths exceeding the warning value to generate a comprehensive risk score layer. The grid division threshold is set according to the regulatory accuracy requirements, such as a fixed size of 1km×1km, or a variable size that is dynamically adjusted according to the area of ​​the sea area. The initial monitoring grid unit is generated using the regular quadrilateral subdivision method to ensure full coverage and no overlap. Each grid cell inherits the underlying geographic attributes (such as average water depth, main landform type) and risk level label to form a spatial management unit with multi-dimensional attributes. The grid boundary is topologically consistent with the administrative jurisdiction boundary of the sea area to avoid cross-jurisdictional division.

[0080] Resource deployment constraints include physical limitations such as patrol vessel range radius, drone flight time, and monitoring equipment coverage. Graph iterative optimization uses the initial monitoring grid units as nodes to construct an adjacency graph, which is then iterated through the following steps: calculating the regulatory demand index (comprehensive area, risk level, and accessibility) for each grid; evaluating the actual coverage capacity of existing resources for that grid; and merging or splitting grids where regulatory gaps exceed a threshold. The merge operation prioritizes grids with similar risk levels and spatial proximity, while the split operation subdivides large, high-demand grids based on terrain features. The iterative termination condition is that the resource supply-demand ratio of all grids reaches a balanced threshold. The final output island monitoring grid not only meets the needs of key hotspot monitoring but also conforms to actual equipment and manpower deployment capabilities, forming an operational regulatory unit system.

[0081] This embodiment constructs an island monitoring grid with dynamic optimization capabilities by integrating geospatial data with historical sea area supervision data. The spatial construction based on sea area boundaries and geographic information ensures accurate coverage of the supervision scope and avoids the arbitrariness of traditional manual demarcation. Through in-depth analysis of marine aerial photography records and manual inspection reports, scientific identification of hotspots is achieved, significantly improving the targeting of supervision resources. The combination of spatial superposition and grid division thresholds enables the initial monitoring grid to reflect the natural characteristics of the sea area and match the risk distribution, solving the problem that static grids cannot adapt to dynamic supervision needs. The graph iterative optimization process incorporates resource deployment constraints into grid adjustment, so that the final generated island monitoring grid not only meets supervision needs, but also takes into account the actual operating capabilities of patrol boats, drones and other equipment, greatly improving supervision efficiency and resource utilization.

[0082] In one embodiment, multi-source monitoring data of the target sea area is obtained based on the island monitoring grid, and grid data recognition is performed to obtain a global dynamic data set, including:

[0083] The planning of the island monitoring grid is based on the principle of geographic spatial gridding, dividing the target sea area into regular or irregular geographic units. The grid size is determined by the island area, coastline complexity, and regulatory accuracy requirements. Gridding is performed using the spatial grid generation tool of the Geographic Information System (GIS). After inputting island vector boundary data, a base grid layer is generated based on the principle of equal area or equal latitude and longitude intervals. The topological location of monitoring nodes must meet the spatial coverage constraints of multi-source data acquisition equipment: satellite remote sensing monitoring requires cloud cover within the grid and satellite transit time windows; drone monitoring must ensure that route planning complies with air traffic control requirements; and nearshore monitoring equipment deployment must avoid tidal erosion areas. The spatial distribution of monitoring nodes follows a gradient priority strategy, with differentiated node density configurations implemented in areas of high human activity, ecological protection areas, and historically targeted areas. For grids with undulating terrain, 3D terrain modeling is used to calculate the equipment's field of view. Node coordinates are adjusted to ensure that the monitoring equipment's coverage area forms a continuous, blind-spot-free monitoring network within the grid. The topological location of the monitoring node is ultimately output as a metadata file containing device type, spatial coordinates, coverage radius, and acquisition frequency, providing a spatial index basis for subsequent multi-source data acquisition.

[0084] Multi-source data collection relies on spatial indexing of the topological locations of monitoring nodes to trigger the corresponding data acquisition process. Satellite remote sensing data is acquired through ground receiving stations, either by accessing archived data or by issuing programmed tasks to obtain multispectral images of a specified grid. The spatial resolution of the images is at least 10 meters, and the wavelength range covers the visible to near-infrared spectrum. Drone aerial data collection utilizes an automatic route planning module to generate a terrain-mimicking flight path based on the coordinates of the monitoring nodes. Equipped with high-resolution optical cameras and lidar equipment, it acquires orthophotos and 3D point cloud data with centimeter-level ground resolution. Nearshore monitoring data is collected using high-definition cameras and hydrological sensors deployed at fixed facilities such as docks and observation towers, collecting real-time video streams and physical and chemical parameters such as water temperature and salinity. Multi-source alignment and integration involves two core steps: spatial reference unification and temporal synchronization. Spatial reference unification utilizes control point registration, using permanent landmarks on islands as reference points. Satellite imagery, drone data, and nearshore video frames are converted to the same plane coordinate system. Temporal synchronization utilizes the Global Navigation Satellite System (GNSS) timing module of the data acquisition equipment to normalize the time tags of the multi-source data, eliminating time deviations caused by device response delays or transmission lags. The integrated multi-source monitoring data is stored as a spatiotemporally consistent raster-vector hybrid dataset, in which satellite images provide large-scale spectral information, drone data supplement detailed textures, and nearshore monitoring data embeds dynamic change features to form a multidimensional data cube.

[0085] The feature classification rule library consists of three components: a spectral feature library for typical marine and island features, a texture feature library, and morphological rules. The spectral feature library defines reflectance threshold ranges for features such as vegetation, water bodies, beaches, and man-made structures in each band of multispectral imagery. The texture feature library extracts parameters such as contrast and entropy based on the gray-level co-occurrence matrix of drone data to distinguish features with similar surface textures. The morphological rules define area thresholds, shape indices, and spatial distribution patterns for different features. The classification process utilizes a hierarchical decision-making mechanism: first, pixel-level classification is performed on satellite multispectral imagery, using the Normalized Difference Water Index and the Normalized Vegetation Index to separate water and vegetation areas. High-resolution texture features from drone aerial data are then incorporated, using a sliding window analysis method to identify detailed features such as building edges and traces of reclamation projects. Nearshore monitoring data utilizes HSV color space analysis of video keyframes to assist in identifying dynamic intertidal zones. Spatial context verification is implemented during the classification process. When conflicting classification results for adjacent pixels are found, logical corrections are made based on the spatial distribution of features (e.g., buildings are discontinuously distributed in deep water). The classification results are spatially aggregated according to the original monitoring grid, generating a statistical table of the proportion of each feature type within each grid and a spatial distribution raster map. The classification confidence level for each pixel is also recorded. The gridded feature label distribution map is stored in GeoTIFF format. The attribute table contains feature codes, area statistics, and data source identification, providing semantic input for subsequent image segmentation.

[0086] The feature classification results of the gridded feature label distribution map contain discrete pixel-level label information, which requires spatial aggregation and boundary extraction to form independent patches with geographical significance. Regional edge detection and segmentation are based on the spatial continuity characteristics of feature categories. A method combining morphological filtering and edge tracking is used to eliminate classification noise and extract precise boundaries. A closing operation is performed on each feature category layer, using circular structuring elements to fill small holes, smooth irregular edges, and enhance the spatial connectivity of pixels in the same category. Edge detection uses a multi-scale gradient detection strategy: for large, continuous features such as vegetation and water, operators are used to extract weak edges, and false boundaries are filtered out in combination with feature area thresholds. For areas with complex details such as buildings and land reclamation projects, the Laplacian of Gaussian operator is used to enhance the response of corner points and linear features. The detected edge lines are converted into continuous polygon outlines, and contour lines with intersections or breakpoints are topologically repaired to ensure the geometric integrity of each closed polygon.

[0087] During the vector data generation phase, the outline polygons are converted into vector boundaries, and the number of redundant vertices is reduced while maintaining the shape features. The simplification tolerance is set to 0.5 times the image spatial resolution. A topological consistency check is performed on the simplified vector boundaries to eliminate boundary overlaps or gaps between adjacent patches, and the attribution relationship between the patches and the original grid is established through spatial indexing. The independent patch boundary vector data records the identifier, feature type, area, and grid number of each patch in a layer attribute table. The output format uses a standardized Shapefile or GeoPackage to support spatial query and overlay analysis. The boundary accuracy of this step is directly affected by the resolution and edge detection parameters of the feature classification results. Areas with a classification confidence level below 85% trigger a manual review mechanism during segmentation.

[0088] The historical patch database stores the geometric boundaries, land feature attributes, and change event records of all patches in previous monitoring cycles in a spatiotemporal cube structure. Global matching identifies the evolutionary relationship between the current patch and historical data through spatiotemporal sequence comparison and spatial relationship reasoning. The matching process is divided into two stages: spatial superposition and attribute association. In the spatial superposition stage, the current patch boundary vector data is superimposed layer by layer with the patches of the same geographic grid in the historical database to calculate the spatial intersection area, overlap rate, and shape similarity index between the two. In the attribute association stage, the land feature type, area change rate, and spatial distribution pattern of the current patch and the historical patch are compared, and the patch status is determined based on the preset change threshold.

[0089] For patches with a spatial intersection area of ​​more than 90% and consistent feature types, they are marked as "stable patches"; patches with an intersection area of ​​less than 30% or a change in feature type are marked as "new patches" or "disappearing patches"; patches with an intersection area between 30% and 90% are further analyzed for their morphological change trajectories. If the area continues to grow and meets the characteristics of land reclamation projects, they are marked as "expanding patches." The construction of the dynamic dataset adopts an incremental update mechanism to extract the spatiotemporal fingerprint features of new or changed patches, including the coordinates of the center point, the change time window, the associated monitoring equipment type, and the original data traceability information, and perform similarity matching with the historical specific target case library. The matching results are stored as dynamic data of the global patches. Each patch record contains the change type, confidence score, and associated historical event number, supporting the playback of the feature evolution process along the timeline.

[0090] This embodiment realizes full coverage and three-dimensional monitoring of the target sea area through the multi-source monitoring point planning of the island monitoring grid, combined with the coordinated deployment of satellites, drones and near-shore equipment, and effectively improves the integrity of monitoring data in complex terrain and dynamically changing areas. Based on the multi-source alignment integration technology, the temporal and spatial benchmark differences of heterogeneous data are eliminated to ensure the accuracy of multi-dimensional data fusion of spectrum, texture and dynamic characteristics, and provide a highly consistent data foundation for land feature classification and evolution analysis. Relying on the vector boundary data generated by the gridded land feature label distribution map and edge detection segmentation, high-precision spatial expression of natural landforms and artificial buildings is achieved, supporting millimeter-level boundary recognition of dynamically changing patches. Through the global matching mechanism of the historical patch database and current data, a traceability chain of specific target behaviors associated with time and space is constructed, which significantly enhances the intelligent discovery and early warning capabilities of problems such as land reclamation and ecological destruction, and provides a full-cycle closed-loop decision-making basis for sea area and island supervision.

[0091] In one embodiment, multispectral remote sensing images, drone aerial data, and nearshore monitoring data of the target sea area are obtained based on the topological location of the monitoring nodes, and multi-source alignment and integration are performed to obtain multi-source monitoring data, including:

[0092] Remote sensing image acquisition uses the deployment coordinates and time window parameters of the satellite equipment at the topological location of the monitoring node, triggering the satellite task scheduling system to obtain raw multispectral data for the target grid. Once the raw data is collected, the historical remote sensing spectral database is accessed to extract spectral reflectance curves of objects at the same geographic location, under similar seasons, and under similar meteorological conditions as a reference. Spectral matching is performed by aligning the band response functions and convolving the sensor band range of the current image with the band parameters of the historical spectral library to generate comparable equivalent spectral signatures.

[0093] Atmospheric error correction utilizes a method based on a combined radiation transfer model and historical spectral inversion. After inputting real-time meteorological parameters (aerosol optical depth, water vapor content), atmospheric correction parameters are iteratively optimized by minimizing the difference between the current image spectrum and historical reference spectra for typical features (such as pure water bodies and bare soil), thereby improving the accuracy of surface reflectance inversion. Terrain error correction, supported by DEM data, incorporates the reflectance attenuation patterns of features with different aspects and slopes from the historical spectral database, and employs a dynamic normalization model to eliminate terrain shadowing effects. The corrected multispectral remote sensing imagery retains the spectral characteristics of features in the blue, green, red, and near-infrared bands, and the spectral curve consistency error with the historical spectral database is less than 5%, providing high-fidelity input for subsequent multi-source data fusion.

[0094] Drone aerial photography planning uses the coordinates of the drone and its coverage radius within the topological location of monitoring nodes as input parameters. This is combined with the drone's flight performance (endurance, climb rate) and camera parameters (focal length, pixel size) to generate a three-dimensional waypoint sequence. Waypoint coordinates are spatially encrypted based on the grid vertex coordinates, and longitudinal flight strip spacing is calculated based on the camera's lateral field of view, ensuring a minimum 70% overlap in the heading direction and 50% overlap in the sideways direction. During the aerial photography mission execution phase, the drone, equipped with a differential GNSS module and an inertial navigation system (INS), performs terrain-simulating flight along the planned waypoints, adjusting its altitude in real time to maintain consistent ground resolution.

[0095] The camera trigger mode uses equidistant intervals for capturing images, simultaneously recording the POS data (latitude, longitude, altitude, and attitude angle) for each image. Aerial data preprocessing includes lens distortion correction and radiometric consistency processing: The original images are geometrically corrected based on camera calibration parameters (radial and tangential distortion coefficients); a histogram matching algorithm is used to equalize the brightness and contrast of adjacent images to eliminate the effects of illumination variations. The preprocessed aerial data sets are output as georeferenced orthophoto mosaics and digital surface models (DSMs), with a spatial resolution better than 5 cm and sub-meter planar positioning accuracy.

[0096] Nearshore monitoring node deployment strictly adheres to the fixed device coordinates pre-set within the monitoring node topology. Wave radars, pressure-type tide sensors, and multispectral water quality meters are installed in key areas such as intertidal zones, reefs, and harbors. Wave monitoring data is collected by high-frequency radars transmitting electromagnetic waves and receiving surface echoes, which are then inverted to obtain wave height, direction, and period parameters. The sampling frequency is no less than 2 Hz, and the spatial coverage is aligned with the monitoring node's field of view. Tide monitoring data is collected using underwater pressure sensors. Through atmospheric pressure compensation and temperature drift correction, the raw pressure values ​​are converted to absolute water level elevations, and the timestamps are synchronized with the GNSS clock.

[0097] Water quality monitoring data, collected every 30 minutes using optical sensors, measures chlorophyll a concentration, turbidity, and dissolved oxygen content. This data is calibrated on-site to eliminate interference from ambient temperature and salinity. During the data integration phase, the multi-source sensor data undergoes spatiotemporal alignment and spatial interpolation. Temporal alignment uses a sliding window method to unify the time base of each device, eliminating timing misalignments caused by transmission delays. Spatial interpolation, based on the Kriging algorithm, converts discrete point data into a continuous raster surface covering the nearshore grid. Wave data generates contour maps of wave height distribution, tidal data constructs a water level elevation surface model, and water quality data forms a spatial distribution map of parameter concentrations. The integrated nearshore monitoring data is stored as a time series multidimensional raster dataset, supporting spatial overlay analysis with remote sensing imagery and drone data.

[0098] Aerial triangulation of drone-generated aerial data uses the coordinates of pre-set control points within the topological locations of monitoring nodes as geometric constraints. These control points can be permanent features (such as rock apex and pier corners) or artificially placed reflective targets. Control point coordinates are measured on-site using GNSS-RTK equipment with a planar accuracy of better than 2 cm and an elevation accuracy of better than 5 cm. These control points are strictly aligned with the spatial reference of the topological locations of the monitoring nodes. Aerial triangulation uses a regional network adjustment method, combining the POS data (positioning and attitude system data) recorded by the drone with automatically extracted image feature points (such as coastline inflection points and building outlines) for joint optimization: a scale-invariant feature matching algorithm is used to generate connecting points with the same name between adjacent images to ensure that each image contains no less than 20 evenly distributed feature points; using the coordinates of the control points as strong constraints and the connecting points as weak constraints, the image exterior orientation elements (latitude and longitude, altitude, roll angle, pitch angle, yaw angle) and the three-dimensional coordinates of the ground points are iteratively adjusted to minimize the sum of squared projection errors of all images. After the adjustment converges, the residual threshold is controlled within 1.5 times the ground resolution of the image. Exceeding the limit area triggers manual verification or additional control points.

[0099] During the registration phase, the adjusted image coordinate system is converted to the planar projection coordinate system (e.g., UTM) used for the topological locations of the monitoring nodes using a seven-parameter Helmert transformation model to eliminate translation, rotation, and scale errors caused by coordinate system differences. Thin plate splines (TPS) are used to compensate for residual geometric distortion by local nonlinear deformation of the image, using control point residuals as input parameters. The resulting planar positioning error is better than 0.3 times the image ground resolution (e.g., less than 1.5 cm for a 5 cm resolution image). The resolved aerial image is then tiled according to the monitoring grid number to generate orthophoto tiles and digital surface model (DSM) tiles. Each tile is associated with a metadata file recording acquisition time, sensor parameters, adjustment accuracy, and residual statistics. The dataset is stored in GeoTIFF format, embedding spatial reference and timestamp information. Planar errors are controlled to within 0.3 times the resolution, ensuring pixel-level spatial consistency with the multispectral remote sensing imagery and providing millimeter-level input data for multi-source registration.

[0100] Multi-source registration uses the spatial coordinate system of the topological location of monitoring nodes as a unified benchmark, and performs hierarchical processing on the spatial alignment and attribute association of heterogeneous data. In the spatial alignment stage, multispectral remote sensing images, aerial orthophoto blocks, and nearshore monitoring raster data are imported into the same geospatial platform, and geometric registration is performed using feature point matching and least squares transformation methods: for satellite and aerial images, stable feature points such as coastlines and building outlines are extracted and affine transformed; for nearshore wave and tide level raster data, a control point forced matching method is used to ensure that tide level contours coincide with the land and water boundaries of remote sensing images. Temporal alignment is achieved through a unified timestamp, interpolating and aligning the satellite transit time, drone aerial photography period, and nearshore monitoring data time series to generate a synchronized time slice dataset.

[0101] During the attribute association phase, semantic mapping relationships are established between multi-source data: textural features of aerial imagery are associated with spectral bands of remote sensing imagery, nearshore water quality parameters are associated with chlorophyll-a inversion results from remote sensing, and tide data are associated with inundated areas extracted from drone-derived dynamic surface mapping (DSM). After registration, multi-source monitoring data is stored as a temporally and spatially consistent, multi-layered raster-vector hybrid dataset containing multi-dimensional attribute fields such as spectrum, texture, wave, tide, and water quality. This supports rapid retrieval and overlay analysis by grid number or timestamp.

[0102] This embodiment significantly improves the accuracy of atmospheric correction and terrain correction by integrating the historical remote sensing spectral library with the real-time collected data for joint inversion, effectively overcomes the spectral distortion caused by complex meteorological conditions and terrain undulations, and ensures that the reflectance characteristics of the multispectral remote sensing images are true and reliable. The collaborative acquisition mechanism of multi-source data from the sky, land, and sea based on the topological position of the monitoring nodes realizes the complementary advantages of satellite wide-area coverage, high-precision detailed inspection by drones, and dynamic monitoring of the nearshore, forming a three-dimensional monitoring network with full spatial dimensions. The strict unification of spatiotemporal benchmarks and semantic association mapping in the multi-source registration process eliminates spatial dislocation and attribute deviation between heterogeneous data, providing a millimeter-level consistent data base for the evolution analysis of island features. The combination of aerial triangulation solution and closed-loop verification mechanism reduces the positioning error of drones while reversely optimizing the deployment strategy of monitoring nodes, forming an intelligent supervision closed loop with self-improvement of data quality, and greatly enhancing the credibility and timeliness of aerial photography evidence collection such as reclamation monitoring and ecological damage tracing.

[0103] In one embodiment, based on the preset marine ecological protection rules, the coordinates of the suspected specific target area are extracted from the global dynamic dataset and the confidence is calculated to obtain the specific target confidence of the coordinates, including:

[0104] The marine ecological protection rules are a series of predefined constraints used to regulate the development and utilization of marine resources, including but not limited to spatial boundaries and control requirements such as prohibited development zones, restricted development zones, and ecological red line zones. The global dynamic patch dataset is a spatiotemporal continuous data on sea and island surface cover changes acquired through multi-source remote sensing images, drone aerial photography, and ship monitoring. It is stored in the form of vector patches and contains patch boundaries, attribute labels, and timestamp information. The core of the spatial overlay analysis of protected areas is to perform a geometric intersection operation on the global dynamic patch dataset and the spatial control layer in the marine ecological protection rules to screen out patches that overlap or conflict with the boundaries of the protected areas.

[0105] Specifically, marine ecological protection rules usually exist in the form of polygon layers in geographic information systems, such as marine nature reserves, important fishing waters, or prohibited land reclamation areas. Each patch in the global patch dynamic dataset needs to be spatially related to these protected area layers. If the patch falls completely or partially within the protected area, a specific target condition is triggered. For example, if a patch shows newly reclaimed land, and the area is designated as a prohibited development area in the ecological protection rules, the patch will be marked as the initial specific target area. Spatial overlay analysis relies on the geometric calculation functions of GIS tools or spatial databases.

[0106] The generation of the initial set of specific target areas needs to be verified in conjunction with the time dimension. For example, if a certain patch already exists in historical data and has not changed, even if it is located in a protected area, it is a legal legacy project and needs to be excluded; while newly added or expanded patches need to be marked as highlights. In addition, attribute matching is also crucial. For example, if specific types of scientific research activities are allowed in a protected area, further filtering is required through patch attribute fields (such as "land use type"). The final output of the initial set of specific target areas contains the patch's unique identifier, spatial boundary, specific target type, and time information, providing a basis for subsequent change detection.

[0107] The initial set of specific target areas only reflects spatial location conflicts and does not distinguish between natural changes and human-specific target activities. The goal of regional change detection is to eliminate changes in patches caused by legitimate or natural factors through time series analysis, focusing on anomalous human activities. Global patch dynamic datasets contain multi-temporal data, such as patch sequences derived from daily, weekly, or monthly remote sensing imagery. Change detection is achieved by comparing patch attributes (such as area, shape, and cover type) within the same area at different times.

[0108] The key to identifying anomalous areas lies in setting change thresholds and pattern rules. For example, if the area of ​​a patch within a protected area expands dramatically over a short period of time, it suggests illegal land reclamation; if the patch shape changes from a natural coastline to a regular rectangle, it points to the construction of artificial structures. Change detection algorithms can use simple interpolation methods (such as when the area change rate exceeds 20%) or complex models (such as time series-based mutation point detection), but it is important to avoid over-reliance on algorithms and instead set rules based on domain knowledge. For example, if the cover type of a patch within a mangrove reserve changes from "vegetation" to "bare land," it is directly marked as an anomaly. If the change occurs after the typhoon season, meteorological data must be combined to determine whether it is caused by a natural disaster.

[0109] Generating the outline of the suspected target area requires integrating geometric and semantic information. For example, after change detection, only areas with high levels of human intervention are retained from the initial target patch. These outlines are then generated using vector boundary thinning or raster-to-vector conversion. The outline data must be accompanied by evidence of change, such as screenshots of the preceding and following imagery, and statistical values ​​of change (e.g., percentage increase or decrease in area) for subsequent manual verification or confidence calculation. The output of this step represents a fusion of spatial and attribute information, providing a precise geometric foundation for coordinate transformation.

[0110] The outlines of suspected target areas are typically stored as vector polygons (e.g., WKT format) or raster masks, which need to be converted to standardized spatial coordinates for positioning and visualization by regulatory systems. Geospatial data includes coordinate system parameters, elevation datums, and projections (e.g., Mercator projection). The purpose of coordinate conversion is to unify the reference frames of different data sources and ensure that coordinates accurately correspond to actual locations.

[0111] Specifically, if the original contour data is inconsistent with the target coordinate system, projection transformation or datum conversion is required. For example, drone aerial photography data uses a local independent coordinate system, while maritime area supervision requires the use of the national geodetic coordinate system, which requires conversion through a seven-parameter or four-parameter model. For raster data (such as contours extracted from remote sensing images), pixel coordinates must be mapped to geographic coordinates through a geographic reference file. The converted coordinates must retain sufficient accuracy. For example, the boundaries of reclamation projects require sub-meter accuracy, while large-scale ecological monitoring can accept meter-level errors.

[0112] The output format of the coordinates of the suspected specific target area must be adapted to the needs of subsequent analysis. Common formats include center point coordinates (longitude, latitude), a sequence of boundary vertices, or minimum bounding rectangle coordinates. For example, for a specific target expansion in a circular aquaculture area, the center point and radius are output; for an irregular land reclamation area, the longitude and latitude sequences of all boundary points are output. Coordinate data must be accompanied by metadata, such as a coordinate system declaration, accuracy assessment value, and conversion method, to avoid misjudgments due to coordinate ambiguity in subsequent steps. The standardized output of this step provides a unified spatial reference for specific target probability calculations and confidence assessments.

[0113] Historical maritime surveillance data is a spatial dataset of past incidents involving specific targets in maritime areas. It typically includes information such as target type, occurrence time, handling outcome, and geographic location. This data can reveal specific target hotspots, high-incidence types, and temporal patterns in specific regions, providing a statistical basis for assessing the probability of currently suspected specific target areas. The goal of regional specific target probability calculation is to quantify the degree of match between current suspected specific target areas and historical specific target patterns through spatial correlation and statistical analysis, thereby generating a preliminary confidence level.

[0114] In practice, historical maritime surveillance data is typically stored as point or polygon layers, such as the location of illegal reclamation, targeted fishing, or sewage incidents. The calculation process first performs a spatial correlation analysis between the coordinates of suspected target areas and historical data. For example, buffer analysis or spatial joins are used to determine whether the current area is located within a historically high-incidence area for specific targets. If a suspected target area spatially overlaps or is adjacent to multiple illegal reclamation incidents that occurred within the past five years, the probability of it being a specific target increases significantly.

[0115] Statistical methods for calculating probability can use kernel density estimation to measure the frequency of regional specific targets, or adjust confidence levels based on prior probabilities using Bayesian methods. For example, if a certain sea area has experienced 10 illegal aquaculture incidents over the past three years, and the current suspected specific target area happens to be located in an area with high kernel density, the initial confidence level can be set to a higher value (such as 0.8). If the area has no historical record of specific targets but meets typical specific target characteristics (such as the sudden appearance of large-scale artificial structures), the probability needs to be adjusted based on other factors (such as regulatory intensity and ecological sensitivity).

[0116] The output of the preliminary confidence level should include a probability value (e.g., a range of 0 to 1) and supporting evidence. For example, a preliminary confidence level of 0.75 for a coordinate is based on three similar instances of specific target behavior in the area over the past two years, with the current pattern of change closely matching the historical cases. The results of this step provide data support for subsequent specific target level calculations, but they do not yet incorporate the stringency of specific protection rules and therefore require further refinement.

[0117] Marine ecological protection rules not only define the spatial scope of prohibited or restricted activities but also specify the control levels and severity of specific targets for different areas. For example, any human activity within the core protected area is considered a severe specific target, while some compliant development is permitted in the general control area. The purpose of calculating the specific target level is to weight the preliminary confidence level based on the stringency of the protection rules, ultimately generating a coordinate specific target confidence level that reflects the actual regulatory priority of specific target activities.

[0118] Specifically, marine ecological protection rules usually exist in the form of hierarchical control layers, for example:

[0119] Core protection area (level one control): any development is strictly prohibited, and the confidence level of specific target behaviors is directly adjusted to the highest level (such as 1.0).

[0120] Ecological restoration area (secondary control): restricted development. If the initial confidence level is 0.6, it will be increased to 0.8 based on the rules.

[0121] Moderate utilization zone (Level 3 control): Compliant activities are allowed. If the initial confidence level is low (e.g. 0.3), it will be reduced to 0.1.

[0122] Specific target level calculations can be performed using a rules engine or a decision matrix, for example:

[0123] If the area belongs to the core protection area and the preliminary confidence is ≥ 0.5, the coordinate specific target confidence = 1.0.

[0124] If the area belongs to the ecological restoration zone and the preliminary confidence level is ≥ 0.7, the coordinate-specific target confidence level = 0.9.

[0125] If the area belongs to the moderate utilization area and the preliminary confidence is ≤ 0.4, the coordinate specific target confidence = 0.2.

[0126] The final output, the coordinate-specific target confidence score, is a comprehensive assessment that takes into account both the historical probability of the specific target and the current protection level of the area. For example, if a reclamation plot is located in a historically high-incidence area (with an initial confidence score of 0.8) but is located in a general control zone, the final confidence score will be adjusted to 0.7. If the same plot is located in a core protection zone, the confidence score will be set directly to 1.0, triggering the highest priority regulatory response. This step ensures the rational allocation of regulatory resources, prioritizing the specific target behaviors with the most severe ecological impacts.

[0127] This embodiment is based on the combination of spatial overlay analysis of protected areas and regional change detection, which can effectively distinguish between natural changes and human-specific target activities, and greatly improve the accuracy of identifying specific target areas. The use of historical sea area supervision data for probability calculation enables the system to have learning capabilities, and can automatically adjust the monitoring focus according to the characteristics of regional specific targets, significantly improving supervision efficiency. By combining the preliminary confidence level with the sea area protection level, priority hierarchical management of specific target behaviors is achieved, so that limited supervision resources can be accurately invested in areas with the highest ecological risks. The entire method constructs a complete supervision closed loop from data collection to intelligent decision-making through multi-source data fusion and spatial analysis technology, providing a scientific and reliable intelligent supervision method for marine ecological protection.

[0128] In one embodiment, the coordinates of the suspected specific target area are compared with the sea project database, and the behavior is evaluated in combination with the confidence level of the specific target in the coordinates to obtain a sea area specific target warning report, including:

[0129] The coordinates of suspected specific target areas are derived from the coordinates of newly added or changed patch centers in the global dynamic patch dataset. Legal sea use data is extracted from the sea use project database and includes the vector boundaries of approved sea use blocks and their attribute information (such as sea use type, approval date, and validity period). Spatial analysis utilizes the spatial overlay analysis capabilities of the Geographic Information System (GIS), overlaying the suspected specific target coordinate point layer with the legal sea use vector area layer.

[0130] For coordinate points that fall completely within the legal sea use boundary, attribute conflict information with the approved sea use type and time validity period is extracted to generate an abnormal coordinate set for the permitted coverage area (for example, the approved type is wharf construction, but the actual monitoring is reclamation). For coordinate points outside the legal sea use boundary, a buffer zone analysis is performed to exclude a reasonable error range for the extension of the approved boundary (such as dynamic boundary fluctuations affected by tides). Coordinate points outside the buffer zone are marked as specific target coordinate sets without a permitted area. The attribute tables of both types of coordinate sets record the coordinate point's timestamp, source data (satellite, drone, or near-shore equipment), and original patch number, providing traceability for subsequent analysis.

[0131] Sea use permit information includes constraints such as the approved sea use type (land reclamation, channel dredging, wharf construction, etc.), permitted construction period, and upper limit on sea use area. Regional sea use type matching performs spatial joins and attribute associations on the set of abnormal coordinates within the permit coverage area. These spatial associations are performed between abnormal coordinate points and the corresponding legal sea use blocks to extract the approved sea use type. Coordinate points that have exceeded the approval validity period (e.g., expired construction permits) are removed through temporal validity verification. Furthermore, the actual monitored sea use type (determined by a land feature classification rule library) is semantically matched with the approved type (e.g., a temporary storage yard may be approved but a permanent structure may be monitored).

[0132] Matching results are categorized into three categories: consistent type and valid time (compliant), deviating type but within area limits (requiring manual review), and conflicting type or exceeding area limits (specific target). The results are written into the coordinate point attribute table and linked to the approval document number and responsible entity information in the sea use project database to form a complete sea use compliance determination result.

[0133] The confidence threshold for a specific target is determined based on the reliability of the monitoring data source, the frequency of occurrence of the coordinate point, and the proximity to the historical specific target area. Data source reliability weighting is calculated as follows: 0.7 for satellite monitoring coordinates, 0.9 for drone monitoring coordinates, and 0.8 for nearshore equipment coordinates. Coordinate point frequency is calculated through time series analysis, with a weight of 1.0 for coordinate points that appear three or more times in a row and 0.5 for a single occurrence. Distance to the historical specific target area is weighted using an exponential decay model, with the weight decreasing by 0.2 for every 100-meter increase in distance.

[0134] Confidence calculation combines the three weights, using the following formula: Confidence = Data Source Weight × Frequency Weight × Distance Weight. During the screening process, the threshold range (typically 0.6-0.9) is dynamically adjusted to retain only coordinates with confidence scores above the threshold. Spatial cluster analysis is performed on the filtered coordinates to eliminate isolated points (e.g., if there are no other specific target points within 500 meters of a single coordinate point). This ultimately generates a high-confidence set of specific target coordinates. The coordinate set attribute table records the confidence score, associated data source, and cluster number, providing input for behavioral pattern matching.

[0135] Historical maritime regulatory data stores the spatial coordinates, behavior types (such as illegal land reclamation, excessive aquaculture, and illegal sand mining), and processing results of all verified specific target events in past periods. Specific target behavior pattern matching utilizes a dual analysis of spatial association and behavioral characteristics: A high-confidence specific target coordinate set is spatially overlaid with a heat map of historical specific target events. The proximity of the current coordinate point to similar historical specific target areas is calculated (for example, coordinates within 500 meters of the nearest historical specific target point are marked as high-risk). The current coordinate point's location is also extracted from the feature type (obtained from a global dynamic dataset), its change trajectory (such as the monthly growth of land reclamation area), and the type of adjacent sea-use projects (such as proximity to ecological protection areas or waterways) to construct a specific target behavior feature vector.

[0136] Using a predefined rule base for specific target patterns (e.g., "continuous land reclamation adjacent to ecological zones" corresponds to "ecologically destructive specific targets"), we perform pattern matching on feature vectors and output specific target types (such as land reclamation, pollution discharge, illegal mining, etc.) and levels (low, medium, and high). This level is weighted based on the specific target area, duration, and ecological sensitivity, ultimately generating a structured specific target behavior type rating table containing type codes and risk levels.

[0137] The specific target assessment integrates the output data from the first three stages: compliance determination results for abnormal coordinate sets within the permitted coverage area (e.g., type conflicts or area exceeding limits), high-confidence specific target coordinate sets within unlicensed areas, and their specific target behavior type levels. The assessment process prioritizes and integrates the chain of evidence: Type-conflicting coordinate points within the permitted coverage area are associated with the sea use approval document number and responsible entity information and marked as "approved specific targets." High-risk coordinate points within the high-confidence specific target coordinate set are labeled as "repeated specific targets" by overlaying their specific target types (e.g., illegal sand mining) with historical aerial footage of similar incidents. For coordinate points adjacent to ecological red lines or sensitive areas, the ecological impact assessment model is combined to output "ecologically high-risk specific targets."

[0138] Maritime Target Warning Reports are categorized by urgency into three levels: Red (immediate action), Orange (limited rectification), and Yellow (focused monitoring). Each level is associated with a specific list of coordinate points, specific target type, legal basis, and optimization recommendations. Reports are generated using standardized templates, embedded with dynamic map visualization modules (such as heat maps of specific target points and historical comparison layers), and accompanied by data traceability information (monitoring equipment type, processing timestamp, and confidence score). These reports are pushed to aerial photography terminals via a secure interface, supporting on-site verification and access to case filing evidence.

[0139] This embodiment achieves accurate separation of abnormal behaviors inside and outside the approval area through spatial superposition analysis of legal sea use ranges and suspected specific target coordinates, effectively improving the positioning efficiency and spatial accuracy of specific target screening. A multi-dimensional compliance judgment mechanism based on sea use type and timeliness, combined with the semantic association of approval attributes and real-time monitoring data, solves the hidden problem of "approval compliance but actual specific target" in traditional supervision. The dynamic weighted model of the specific target confidence threshold integrates the credibility, spatiotemporal frequency and historical distribution characteristics of multi-source data, significantly reduces the interference of isolated false alarms, and enhances the recognition robustness of high-risk specific target areas. The spatial coupling analysis of historical specific target behavior patterns and ecologically sensitive areas gives the specific target type level classification a priori knowledge support, so that the early warning results have both legal basis and ecological risk prediction value.

[0140] In one embodiment, spatial analysis is performed on the coordinates of the suspected specific target area based on the legal sea use range data in the sea use project database to obtain an abnormal coordinate set of the permitted coverage area and a specific target coordinate set of the unlicensed area, including:

[0141] The coordinates of suspected specific target areas are derived from newly added or changed patches in the global dynamic patch dataset, extracted using the patch centroid coordinates or the vertex coordinates of the largest sub-patch. Legal sea use data is stored as a standardized vector layer, with each polygon associated with the approval document number, sea use type, spatial boundary coordinates, and attribute metadata. The polygon spatial boundary calculation utilizes the spatial containment criterion of the geospatial engine: a point-by-point spatial relationship between the suspected specific target coordinates and the legal sea use polygon layer is determined. This determination incorporates a topological tolerance mechanism, with a buffer distance set at twice the accuracy of the approved boundary coordinates (typically 5-10 meters) to eliminate boundary ambiguity caused by surveying errors or tidal dynamics. Ray penetration testing is performed on each coordinate point, counting the number of intersections with the polygon boundary. Odd-numbered intersections are considered interior points, while even-numbered intersections are considered exterior points. The results are classified as "candidate points in the permitted coverage area" (interior points) or "candidate points in the non-permitted area" (exterior points), and the confidence level of the penetration test is recorded (e.g., the confidence level for points near the boundary is reduced by 30%). After the preliminary screening coordinate set is generated, it is associated with the approval document number and sea use type attributes and stored as an event log table in the spatial database. The log table contains the unique identifier of the coordinate point, the determination timestamp, the original data source device number and the spatial inclusion status mark, providing a traceable intermediate data layer for subsequent refined filtering.

[0142] Spatial relationship filtering uses the spatial metadata of approved sea use blocks to perform multi-level validation on "candidate points in the permitted coverage area." First, a spatial join operation is used to link candidate points to the approval validity period field in the sea use project database, extracting the start and end time attributes for each point's corresponding approval block. Temporal validity filtering utilizes a sliding time window match: if a coordinate point's timestamp (derived from the monitoring data collection time) exceeds the approval deadline by more than three months, it is marked as an "expired permitted coordinate point." If it falls within the three-month window of the approval deadline, it is marked as a "point awaiting immediate review," triggering a manual review process. A secondary spatial refinement is performed on valid permitted coverage coordinate points: a dynamic buffer zone is extended outward from the approval block boundary. The buffer zone radius is set based on the sea use type (e.g., 5% of the approved area for reclamation projects and 10 meters for waterway projects). Coordinate points that exceed the buffer zone are screened out, marked as "suspected out-of-bounds points," and associated with an out-of-bounds distance attribute. The expired license coordinate set and the valid license coverage coordinate set are stored in independent spatial layers respectively. The layer attribute table records the approval expiration date, out-of-bounds distance, associated aerial photography responsible person and review status identification, and at the same time retains the foreign key association with the event log table of the preliminary screening coordinate set to ensure that the entire link data is traceable.

[0143] The permitted activity type attributes are compared to establish a semantic mapping relationship table between the approved sea use type and the monitored land feature type. The mapping table defines type compatibility rules (e.g., "dock construction" allows the subclasses "concrete structure" and "yard," but prohibits the subclasses "sand mining equipment") and area fluctuation thresholds (e.g., the area error for earthwork projects is ±15%). The comparison process is divided into two stages: type matching and area verification:

[0144] Type Matching: Extract the monitored feature types (from the feature classification results of the global dynamic dataset) at the coordinates covered by the valid permit and perform a rule-based mapping with the approved sea use type. If the monitored type is not in the compatible permit list (for example, the approved permit is "Dredging Project" but "Permanent Construction" is detected), it is marked as a "Type Conflict."

[0145] Area Verification: The actual area of ​​the monitored sea area (based on the polygonal area of ​​the patch's vector boundary) is calculated and compared with the maximum permitted area approved. If the percentage exceeds the permitted area by a floating threshold (e.g., approved 100 hectares, but 120 hectares are actually monitored), the area is marked as "Area Exceeded." If both a type conflict and an area exceeding the permitted area exist, the area is marked as "Compound Specific Target."

[0146] The anomaly coordinate set is output as a spatial hotspot layer. The layer's attribute table records the specific target type code (e.g., T1 for type conflict, A2 for 20-30% area overrun), the overrun value, the associated approval document number, and the original monitoring image ID. This layer is indexed with the approval details table in the sea use project database, enabling one-click access to design drawings and responsible party information from approval files as supporting evidence for specific target identification.

[0147] Unlicensed area clustering involves a joint analysis of spatial density and time series for "unlicensed area candidate points" and "expired license coordinate points." Spatial clustering utilizes a modified DBSCAN model, with parameters tailored to the spatial distribution characteristics of specific target behaviors in the maritime area. The neighborhood radius is adaptively calculated based on the average distribution density of historically specific target hotspots (typically 200-1000 meters), and the minimum point count threshold is dynamically adjusted based on the credibility of the monitoring data source (satellite data sources require ≥5 points, and drone data sources require ≥3 points). A time weighting factor is introduced into the clustering process: coordinate points occurring within three consecutive monitoring periods are given a 2x weight, while isolated single points are given a weight of 0.5.

[0148] The clustering results generate the spatial boundaries of specific target clusters. This boundary generation uses the α-shape algorithm to adapt to complex coastline morphology and avoid errors caused by regular convex hulls. A legality review is performed on each specific target cluster: spatial overlay is performed with the pre-approved blocks in the sea use project database. If the overlap exceeds 50%, the cluster is marked as a "cluster pending approval" and the warning level is lowered. The final set of specific target coordinates in the unlicensed area is stored as a thematic layer for specific target clusters. The attribute table records the number of coordinate points within the cluster, the spatiotemporal activity index (calculated based on frequency and duration of occurrence), the distance to adjacent sensitive areas (such as ecological red lines), and historical penalties for similar specific targets. The spatiotemporal indexing mechanism of the spatial database enables second-level association queries.

[0149] This embodiment combines the topological tolerance mechanism with the dynamic buffer zone to accurately determine the spatial ownership of coordinate points and sea use boundaries, effectively eliminate surveying and mapping errors and tidal dynamic interference, and improve the recognition accuracy of specific target behaviors within and outside the legal range. Dynamic adaptive clustering parameters are combined with historical specific target hot zone analysis to enhance the environmental adaptability of specific target cluster identification in unlicensed areas, and simultaneously suppress isolated noise point interference and misjudgment of new specific target patterns. Full-link data lineage management connects original monitoring, approval attributes and processing intermediate results, constructing a traceable spatiotemporal evidence chain for specific target events, ensuring the judicial credibility of early warning reports and the operability of aerial photography.

[0150] Reference Figure 2 As shown, the present invention also provides a multi-source sea area data intelligent supervision system, which is applied to any of the multi-source sea area data intelligent supervision methods mentioned above, comprising:

[0151] The acquisition module is used to obtain the sea area geospatial data and historical sea area supervision data of the target sea area, and to carry out monitoring construction to obtain the island monitoring grid;

[0152] The analysis module is used to obtain multi-source monitoring data of the target sea area based on the island monitoring grid, and perform grid data recognition to obtain a dynamic data set of the entire area;

[0153] The association module is used to extract the coordinates of the suspected specific target area in the global patch dynamic data set and calculate the specific target confidence of the coordinates;

[0154] The processing module is used to compare the coordinates of the suspected specific target area, and conduct behavior assessment based on the confidence of the specific target in the coordinates to obtain a sea area specific target early warning report.

[0155] The present invention provides a multi-source sea area data intelligent supervision system, which realizes the dynamic perception of the changes in the entire sea area and islands by constructing an island monitoring grid and integrating multi-source monitoring data, overcomes the lag problem of traditional single remote sensing or manual inspection, and significantly improves the early detection capability of behavior. Confidence calculation is performed on suspected specific target areas to avoid the misjudgment and omission problems of static threshold judgment, making the identification of specific target behavior more scientific and reliable. By combining the sea use project database with real-time monitoring data for range comparison, it can effectively distinguish between legal construction and specific target behavior, reduce regulatory misjudgments, and improve the accuracy of aerial photography decisions. By comprehensively analyzing the specific target confidence and behavior assessment results, an accurate sea area specific target early warning report is generated, providing data-driven decision-making basis for regulatory departments, improving regulatory response speed and governance efficiency.

[0156] Reference Figure 3 As shown, the present invention also provides a multi-source sea area data intelligent monitoring device, comprising:

[0157] Memory, used to store programs;

[0158] A processor is used to execute a program to implement each step of any one of the above-mentioned multi-source sea area data intelligent supervision methods.

[0159] In this embodiment, the processor and memory may be connected via a bus or other means. The memory may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as read-only memory, flash memory, hard disk, or solid-state drive. The processor may be a general-purpose processor, such as a central processing unit, a digital signal processor, an application-specific integrated circuit, or one or more integrated circuits configured to implement the embodiments of the present invention.

[0160] The present invention also provides a storage medium storing computer instructions, wherein the computer instructions are used to enable a computer to execute any of the above methods.

[0161] It should be noted that, those skilled in the art will clearly understand that, for the sake of convenience and brevity of description, the specific working processes of the above-described system and each module can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0162] The above description is only a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention description and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A multi-source marine data intelligent supervision method, characterized in that: include: Obtain maritime geospatial data and historical maritime regulatory data for the target sea area, and construct an island monitoring grid; Acquire multi-source monitoring data of the islands in the target sea area based on the island monitoring grid, perform grid data recognition, and obtain a dynamic data set of global patches; Extracting the coordinates of the suspected specific target area of ​​the global spot dynamic data set and calculating the specific target confidence of the coordinates; Performing range comparison on the coordinates of the suspected specific target area, and performing behavior assessment based on the specific target confidence of the coordinates, to obtain a sea area specific target early warning report; The coordinates of the suspected specific target area are compared, and the behavior assessment is performed in combination with the specific target confidence of the coordinates to obtain a sea area specific target early warning report, including: The legal sea use spatial range analysis is performed on the coordinates of the suspected specific target area according to the preset sea use project database to obtain the abnormal coordinate set of the permitted coverage area and the specific target coordinate set of the unlicensed area; Based on the sea use permit information in the sea use project database, regional sea use type matching is performed on the abnormal coordinate set of the permit coverage area to obtain a sea use compliance determination result; Performing regional coordinate screening on the unlicensed area specific target coordinate set according to a preset specific target confidence threshold to obtain a high-confidence specific target coordinate set; performing specific target behavior pattern matching on the high-confidence specific target coordinate set according to the historical sea area supervision data to obtain a specific target behavior type level; Performing a specific target assessment on the sea use compliance determination result, the high-confidence specific target coordinate set, and the specific target behavior type level to obtain the sea area specific target early warning report; The extracting the coordinates of the suspected specific target area of ​​the global spot dynamic data set and calculating the coordinate specific target confidence includes: According to the preset marine ecological protection rules, the protected area space overlay analysis is performed on the dynamic dataset of the global map spots to obtain the initial specific target area set; Performing regional change detection on the initial specific target area set based on the global patch dynamic data set, and identifying abnormal areas to obtain the outline of the suspected specific target area; Performing spatial coordinate conversion on the outline of the suspected specific target area according to the sea area geographic spatial data to obtain the coordinates of the suspected specific target area; Calculating the probability of a specific target in the area of ​​the suspected specific target based on the historical sea area supervision data to obtain a preliminary confidence level; The specific target level of the preliminary confidence is calculated according to the marine ecological protection rules to obtain the specific target confidence of the coordinates.

2. The multi-source marine data intelligent supervision method according to claim 1 is characterized in that: The acquisition of the target sea area's geospatial data and historical sea area supervision data, and the construction of monitoring to obtain an island monitoring grid, includes: Acquiring sea area boundary data and geographic information based on the target sea area, and performing spatial construction to obtain the sea area geographic spatial data; Obtaining marine aerial photography records and manual inspection reports of the target sea area, conducting sea use behavior analysis, and obtaining the historical sea area supervision data; Spatially superimposing the sea area geographic spatial data according to the historical sea area supervision data, and performing initial grid setting in combination with a preset grid division threshold to obtain an initial monitoring grid unit; The initial monitoring grid unit is subjected to graph iterative optimization according to the resource deployment constraint conditions of the target sea area to obtain the island monitoring grid.

3. The multi-source marine data intelligent supervision method according to claim 1 is characterized in that: The multi-source monitoring data of the target sea area is obtained based on the island monitoring grid, and grid data recognition is performed to obtain a global dynamic data set, including: Planning multi-source monitoring points for the island monitoring grid to obtain topological locations of monitoring nodes; Acquire multispectral remote sensing images, drone aerial photography data, and nearshore monitoring data of the target sea area according to the topological positions of the monitoring nodes, and perform multi-source alignment and integration to obtain the multi-source monitoring data; Performing ground object classification processing on the multi-source monitoring data based on a preset ground object classification rule library to obtain a gridded ground object label distribution map; Performing regional edge detection and segmentation on the gridded feature label distribution map to obtain independent patch boundary vector data; Global matching is performed on the independent patch boundary vector data according to the historical patch database to obtain the global patch dynamic data set.

4. The multi-source sea area data intelligent supervision method according to claim 3 is characterized in that: The multi-spectral remote sensing image, drone aerial data and nearshore monitoring data of the target sea area are obtained according to the topological position of the monitoring node, and multi-source alignment and integration are performed to obtain the multi-source monitoring data, including: remote sensing image acquisition is performed according to the topological position of the monitoring node, and atmospheric and terrain errors are eliminated to obtain the multispectral remote sensing image; Performing UAV aerial photography planning based on the topological positions of the monitoring nodes to obtain the UAV aerial photography data; Deploy nearshore monitoring nodes according to the topological positions of the monitoring nodes, collect wave monitoring data, tide level monitoring data, and water quality monitoring data, and integrate the nearshore data to obtain the nearshore monitoring data; Performing aerial triangulation and position registration on the UAV aerial data based on the topological positions of the monitoring nodes to obtain an aerial image dataset; Multi-source registration is performed on the multispectral remote sensing image, the aerial image dataset, and the nearshore monitoring data to obtain the multi-source monitoring data.

5. The multi-source sea area data intelligent supervision method according to claim 1 is characterized in that: The legal sea use spatial range analysis is performed on the coordinates of the suspected specific target area according to the preset sea use project database to obtain the abnormal coordinate set of the permitted coverage area and the specific target coordinate set of the unlicensed area, including: Perform polygonal spatial boundary calculation on the coordinates of the suspected specific target area based on the legal sea use scope data in the sea use project database to obtain a preliminary screening coordinate set; Performing spatial relationship and time validity filtering on the preliminary screening coordinate set to obtain an expired permission coordinate set and a valid permission coverage coordinate set; Performing a comparison of the permitted activity type attributes on the valid permitted coverage coordinate set to obtain the permitted coverage area abnormal coordinate set; The preliminary screening coordinate set and the expired license coordinate set are clustered into non-licensed areas according to the legal sea use range data to obtain the non-licensed area specific target coordinate set.

6. A multi-source marine data intelligent monitoring system, characterized by: The multi-source maritime data intelligent supervision method as described in any one of claims 1 to 5 above comprises: The acquisition module is used to obtain the sea area geospatial data and historical sea area supervision data of the target sea area, and to perform monitoring construction to obtain the island monitoring grid; An analysis module, configured to obtain multi-source monitoring data of the target sea area based on the island monitoring grid, and perform grid data recognition to obtain a global dynamic data set; An association module, the association module is used to extract the coordinates of the suspected specific target area of ​​the global spot dynamic data set and calculate the specific target confidence of the coordinates; A processing module, the processing module is used to perform range comparison on the coordinates of the suspected specific target area, and perform behavior assessment based on the specific target confidence of the coordinates to obtain a sea area specific target early warning report; The coordinates of the suspected specific target area are compared, and the behavior assessment is performed in combination with the specific target confidence of the coordinates to obtain a sea area specific target early warning report, including: The legal sea use spatial range analysis is performed on the coordinates of the suspected specific target area according to the preset sea use project database to obtain the abnormal coordinate set of the permitted coverage area and the specific target coordinate set of the unlicensed area; Based on the sea use permit information in the sea use project database, regional sea use type matching is performed on the abnormal coordinate set of the permit coverage area to obtain a sea use compliance determination result; Performing regional coordinate screening on the unlicensed area specific target coordinate set according to a preset specific target confidence threshold to obtain a high-confidence specific target coordinate set; performing specific target behavior pattern matching on the high-confidence specific target coordinate set according to the historical sea area supervision data to obtain a specific target behavior type level; Performing a specific target assessment on the sea use compliance determination result, the high-confidence specific target coordinate set, and the specific target behavior type level to obtain the sea area specific target early warning report; The extracting the coordinates of the suspected specific target area of ​​the global spot dynamic data set and calculating the coordinate specific target confidence includes: According to the preset marine ecological protection rules, the protected area space overlay analysis is performed on the dynamic dataset of the global map spots to obtain the initial specific target area set; Performing regional change detection on the initial specific target area set based on the global patch dynamic data set, and identifying abnormal areas to obtain the outline of the suspected specific target area; Performing spatial coordinate conversion on the outline of the suspected specific target area according to the sea area geographic spatial data to obtain the coordinates of the suspected specific target area; Calculating the probability of a specific target in the area of ​​the suspected specific target based on the historical sea area supervision data to obtain a preliminary confidence level; The specific target level of the preliminary confidence is calculated according to the marine ecological protection rules to obtain the specific target confidence of the coordinates.

7. A multi-source sea area data intelligent monitoring device, characterized in that: include: Memory, used to store programs; A processor is used to execute the program to implement the various steps of a multi-source sea area data intelligent supervision method as described in any one of claims 1 to 5.

8. A storage medium, characterized in that: Computer instructions are stored, and the computer instructions are used to make a computer execute the method according to any one of claims 1 to 5.

Images