A method, device and equipment for monitoring the changes of time, space and space of aquaculture, and a storage medium

Abstract

The application discloses a sea area change monitoring method, device and equipment and a storage medium, and relates to the technical field of sea area monitoring. The method comprises the following steps: performing super-resolution reconstruction based on target sea area boundary data at a set monitoring frequency to obtain remote sensing images of the target sea area; extracting scattered planar vector patches of culture facilities from the remote sensing images; determining culture range patches of the culture facilities according to the outer convex surface boundary geometry of the midpoint element patch in the scattered planar vector patch; analyzing NetCDF sea current multidimensional data to generate format grid sets including different direction flow rates and wind speed value vector fields according to the step length of the set monitoring frequency; determining a regression coefficient map according to the spatial relationship between the different direction flow rates and the wind speed, and monitoring the culture facilities by using the regression coefficient map. The application takes into account the influence of flow rates, flow directions and wind speeds on the position distribution of the culture facilities, and can efficiently and accurately monitor the temporal and spatial changes of the culture sea area by combining the analysis of the NetCDF data.

Description

Technical Field

[0001] This application relates to the field of marine monitoring technology, and in particular to a method, device, equipment and storage medium for monitoring the spatiotemporal changes of marine areas used for aquaculture. Background Technology

[0002] Aquaculture is a traditional marine development and utilization activity, and it represents the largest type of marine use in my country, playing a vital role in ensuring the livelihoods of fishermen and promoting the economic and social development of coastal areas. my country's marine aquaculture industry has developed rapidly and on a large scale, benefiting from a series of policies and measures to optimize the management of aquaculture use. The area of ​​suitable deep-water and offshore fishing grounds has been continuously expanded, and deep-sea aquaculture and marine ranching have developed rapidly. Raft and cage aquaculture are the main methods of open-type aquaculture, relying on natural seawater flow and ecological conditions. Due to human expansion of facilities or natural changes in sea conditions, the actual distribution area and location of aquaculture production facilities are prone to frequent changes. Therefore, to strengthen the supervision of aquaculture use, it is crucial to conduct routine monitoring to understand the distribution of fishery resources, promptly detect, stop, and handle illegal and irregular marine activities, supervise approved aquaculture use, promptly detect and correct illegal and irregular marine activities such as exceeding the approved scope of production or changing the mode of marine use, and conduct detailed identification and analysis of these behaviors. This is essential for the scientific management of marine resource utilization, the standardization of marine use behavior, and the maintenance of good marine management order. Summary of the Invention

[0003] The main objective of this application is to provide a method, device, equipment, and storage medium for monitoring the spatiotemporal changes of aquaculture sea areas, so as to efficiently and accurately monitor the spatiotemporal changes of aquaculture sea areas.

[0004] To achieve the above objectives, one aspect of this application proposes a method for monitoring the spatiotemporal changes of aquaculture sea areas, the method comprising the following steps: Based on the target sea area boundary data, super-resolution reconstruction is performed at a set monitoring frequency to obtain the remote sensing image of the target sea area. Extract scattered planar vector patterns of aquaculture facilities from remote sensing images of the target sea area; The centroid position of the scattered planar vector map patch is calculated using the centroid method and used as a point feature map patch. The geometry of the outer convex boundary of the point feature map patch is calculated to obtain the aquaculture range map patch of the aquaculture facility. Parse NetCDF ocean current multidimensional data and generate a raster set in a format that includes current velocities in different directions according to the step size of the set monitoring frequency; Analyze the NetCDF wind field multidimensional data, generate a wind speed value vector field grid according to the step size of the set monitoring frequency, and generate the wind speed based on the wind speed value vector field grid; Based on the spatial relationship between the flow velocity in different directions and the wind speed, a corresponding regression coefficient diagram is determined, and the regression coefficient diagram is used to monitor the aquaculture facility.

[0005] In some embodiments, the process of performing super-resolution reconstruction based on target sea area boundary data at a set monitoring frequency to obtain remote sensing images of the target sea area includes the following steps: Based on the target sea area boundary data, medium-resolution multispectral images are acquired at the set monitoring frequency. The medium-resolution multispectral images are stored as image datasets in batches according to the order of acquisition. Using existing control points, the image dataset is geometrically corrected and image registered to obtain a corrected TIFF format first monitoring image set for each period; wherein, each first monitoring image set for each period corresponds to each of the set monitoring frequencies; Super-resolution reconstruction is performed on the first monitoring image set for each period to generate a high-resolution multispectral image set; Stable registration points along the coastline and islands are selected to recorrect and register the high-resolution multispectral image set, and mosaicking is performed to form a complete second monitoring image set for each period of the target sea area as the remote sensing image of the target sea area.

[0006] In some embodiments, the extraction of scattered planar vector patterns of aquaculture facilities from remote sensing images of the target sea area includes the following steps: Training samples were created based on remote sensing images of the target sea area; Based on the pixel classification method, the training samples are configured and classified and a random forest model is selected. The threshold for the number of training samples for each category is limited, and the random forest model is trained by the relationship between the explanatory variables and the target dataset. Based on the random forest model and the training samples, raster classification is performed to generate initial aquaculture facility classification raster with different sawtooth shapes; For local areas affected by salt and pepper noise after classification, the target aquaculture facility classification grid after noise optimization is obtained by using GIS main filtering and smoothing the area boundaries. The target aquaculture facilities are classified into raster grids and converted into vector surfaces. The geometric shape of the patches is set to maintain a jagged, non-simplified shape. By constructing a multi-component batch removal of marine patches, a surface vector pattern of aquaculture facilities is formed. The surface vector pattern of aquaculture facilities includes circular vector patterns and grid-shaped vector patterns. The planar vector pattern of the aquaculture facility is regularized. The regularization process first involves splitting and breaking down multiple components, and then the extracted planar pattern is geometrically normalized to obtain the normalized scattered planar vector pattern. The geometric normalization of the extracted planar vector patterns includes: straightening the jagged outline of the grid-shaped vector pattern to a right angle, and fitting the jagged outline of the circular vector pattern to a circle.

[0007] In some embodiments, the step of calculating the centroid position of the scattered planar vector map patch using the centroid method as a point feature patch, and calculating the outer convex boundary geometry of the point feature patch to obtain the aquaculture range patch of the aquaculture facility includes the following steps: The centroid method is used to calculate the centroid position of the scattered planar vector vector patch in each period to obtain the point element patch in each period, and then generate the point element vector vector patch dataset; wherein, the scattered planar vector vector patch in each period corresponds to the remote sensing image of the target sea area at each set monitoring frequency. Spatial calculations are performed using the vector data of ownership boundaries of all aquaculture ranches in the target sea area and the point feature vector map dataset to redefine and obtain the aquaculture range map for each period. Specifically, for a point feature vector map dataset of a certain period, the original boundary space of all ranches in the target sea area is first expanded outward by a buffer range to generate a new areal range map. Then, the new single ranch ownership space vector range is overlaid and analyzed with the point feature vector map dataset. The selected point feature map is subjected to convex envelope calculation to regenerate a new minimum convex surface boundary geometry. At the same time, the ranch number information is recorded. Once the convex envelope calculation is completed for all ranches in the target sea area, the enclosed range map for each period is generated. Based on the enclosed range maps of all periods, an areal feature map dataset of aquaculture facility range is generated.

[0008] In some embodiments, parsing NetCDF ocean current multidimensional data and generating a raster set in a format including current velocities in different directions according to the set monitoring frequency step size includes the following steps: The NetCDF ocean current multidimensional data is parsed, and GeoTIFF format raster sets of eastward and northward current velocities are generated in batches according to the set monitoring frequency step size. The method further includes the following steps: The bilinear interpolation method is used to extract the directional flow velocity values ​​into the attribute fields of the eastward and northward flow velocities corresponding to the vector points based on the GeoTIFF format raster set, generating the first vector point patch dataset with eastward and northward flow velocity attribute values.

[0009] In some embodiments, parsing NetCDF wind field multidimensional data, generating a wind speed value vector field grid according to the set monitoring frequency step size, and generating the wind speed based on the wind speed value vector field grid include the following steps: The NetCDF wind field multidimensional data is parsed, and wind speed value vector field grids corresponding to each set monitoring frequency are generated in batches according to the step size of the set monitoring frequency. The wind speed is generated by interpolating the effective values ​​of adjacent pixels in the wind speed value vector field grid using a bilinear interpolation method. The method further includes the following steps: The wind speed value vector field raster, including the wind speed, is extracted in batches and added to the attribute table of the corresponding period and the same spatial location in the first vector point patch dataset to generate a second vector point patch dataset of wind speed values.

[0010] In some embodiments, determining the corresponding regression coefficient map based on the spatial relationship between the flow velocities in different directions and the wind speed, and using the regression coefficient map to monitor the aquaculture facility, includes the following steps: A geographic weighted regression model is used to model spatial relationships. A local regression equation is set for each spatial unit point, and a Gaussian function is selected to construct a spatial weight matrix. For the fitting problem of the geographic weighted regression model, the modified information criterion is used to seek the optimal bandwidth. Then, the model autocorrelation calculation and the coefficient value layer rendering of the explanatory variables are performed to obtain the regression coefficient map of the flow velocity, flow direction and wind speed, so as to monitor the spatiotemporal change information of the aquaculture facilities.

[0011] To achieve the above objectives, another aspect of this application proposes a marine temporal and spatial variation monitoring device for aquaculture, the device comprising: The image reconstruction unit is used to perform super-resolution reconstruction based on the target sea area boundary data at a set monitoring frequency to obtain remote sensing images of the target sea area. The image patch extraction unit is used to extract scattered planar vector images of aquaculture facilities from remote sensing images of the target sea area; The aquaculture range determination unit is used to calculate the centroid position of the scattered planar vector map patch using the centroid method as a point element map patch, calculate the geometry of the outer convex boundary of the point element map patch, and obtain the aquaculture range map patch of the aquaculture facility. The velocity and direction calculation unit is used to parse NetCDF ocean current multidimensional data and generate a format raster set including flow velocities in different directions according to the step size of the set monitoring frequency. The wind speed calculation unit is used to parse NetCDF wind field multidimensional data, generate wind speed value vector field grids according to the step size of the set monitoring frequency, and generate the wind speed based on the wind speed value vector field grids. The spatiotemporal monitoring unit is used to determine the corresponding regression coefficient diagram based on the spatial relationship between the flow velocity in different directions and the wind speed, and to monitor the aquaculture facility using the regression coefficient diagram.

[0012] To achieve the above objectives, another aspect of this application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the above-described method.

[0013] To achieve the above objectives, another aspect of the embodiments of this application proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.

[0014] To achieve the above objectives, another aspect of this application provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0015] The embodiments of this application include at least the following beneficial effects: This application provides a method, apparatus, equipment, and storage medium for monitoring the spatiotemporal changes of aquaculture sea areas. The scheme involves super-resolution reconstruction based on target sea area boundary data at a set monitoring frequency to obtain remote sensing images of the target sea area; calculating the centroid positions of scattered planar vector vector patches using the centroid method as point feature patches; calculating the geometry of the convex boundary of the point feature patches to obtain the aquaculture range patches of the aquaculture facilities; parsing NetCDF ocean current multidimensional data and generating a raster set including current velocities in different directions at a set monitoring frequency step size; parsing NetCDF wind field multidimensional data and generating a wind speed value vector field raster at a set monitoring frequency step size; generating wind speed based on the wind speed value vector field raster; determining the corresponding regression coefficient map based on the spatial relationship between current velocities in different directions and wind speed; and using the regression coefficient map to monitor the aquaculture facilities. This application takes into account the periodic and random effects of natural environmental factors such as ocean current velocity, direction, and wind speed on the location distribution of aquaculture facilities. It utilizes NetCDF data, which supports multidimensional array storage, efficient reading and writing, and flexibility, to conduct spatiotemporal correlation analysis with dynamic multidimensional marine environmental influencing factors, enabling efficient and accurate monitoring of spatiotemporal changes in aquaculture sea areas. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A flowchart illustrating a method for monitoring the spatiotemporal changes of marine environments used for aquaculture, provided in an embodiment of this application; Figure 2 A real-world example image of the target sea area provided in the embodiments of this application; Figure 3 The original remote sensing image of the target sea area provided in the embodiments of this application; Figure 4 This is a super-resolution processed remote sensing image of the target sea area provided in the embodiments of this application; Figure 5a The classification results of the random forest model provided in the embodiments of this application; Figure 5b The main filtering result provided in the embodiments of this application; Figure 5c The boundary cleanup results provided in the embodiments of this application; Figure 5d The rule-based results provided for the embodiments of this application; Figure 6 A schematic diagram of the structure of a marine temporal and spatial change monitoring device for aquaculture provided in this application embodiment; Figure 7 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit it. In the following description, when referring to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with those of this application; they are merely examples of apparatuses and methods consistent with some aspects of the embodiments of this application as detailed in the appended claims.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0020] Before providing a detailed description of the embodiments of this application, some related technologies involved in the embodiments of this application will be described first, as follows: Terminology Explanation: NetCDF (Network Common Data Format) is a file format used to store multidimensional scientific data (variables) such as temperature, humidity, air pressure, wind speed, and wind direction.

[0021] Image super-resolution reconstruction (SR) is a technique that restores low-resolution images (blurred, lacking detail) to high-resolution images, with the core being the replenishment of lost detail information.

[0022] UGRD: U-component of wind, in meteorology it specifically refers to the zonal component of wind. A positive value indicates a westerly wind, and a negative value indicates an easterly wind.

[0023] VGRD: V-component of wind, in meteorology specifically refers to the meridional component of wind; a positive value indicates a southerly wind, and a negative value indicates a northerly wind.

[0024] HTGL: Abbreviation for Specified height level above ground.

[0025] Because satellite remote sensing technology can provide large-area monitoring data of land and sea, the current operational needs for monitoring suspicious areas and patches of marine and island regions are typically met by using satellite remote sensing technologies such as the Gaofen and Ziyuan series, supplemented by field mapping and indoor visual interpretation techniques for identification and extraction. However, in areas with developed marine aquaculture, the need for more frequent monitoring, such as weekly or bi-weekly monitoring, is gradually increasing to ensure timely and accurate detection and prevention of illegal aquaculture activities. Currently, coastal areas typically produce and share satellite remote sensing orthophotos quarterly, supporting routine monitoring of marine and island regions. While current monitoring methods are somewhat time-sensitive, they typically require supplementary means such as drones and close-range patrols to obtain information on changes. Since marine aquaculture areas are mostly located in deep waters and far shores, covering large areas and scattered across the region, traditional close-range survey methods are slow, costly in terms of manpower and vessels, and currently insufficient to maintain and guarantee the needs of routine monitoring. Therefore, conducting spatiotemporal change monitoring and analysis of marine aquaculture areas, utilizing satellite remote sensing imagery for acquisition, interpretation, and statistical monitoring, requires consideration of both high-frequency temporal requirements and convenience, economy, and timeliness. It is urgent to combine new technologies to improve the efficiency of monitoring marine aquaculture areas.

[0026] With the introduction of artificial intelligence (AI) technology, it has become possible to extract specific aquaculture facilities such as net cages and rafts using deep learning and image recognition. Related technologies include: automated identification, extraction, and accuracy verification of aquaculture areas using convolutional neural networks; extraction of raft and net cage aquaculture areas from remote sensing images generated by feature selection, spectral and textural feature fusion, and dimensionality reduction using the Big Law method; and identification and remote sensing monitoring of nearshore aquaculture areas based on the CMFPNet network. However, overall, AI technology is less widely used in monitoring aquaculture areas. To balance the high resolution of satellite remote sensing with the high timeliness requirements of marine monitoring, its ability to support routine, high-frequency monitoring needs further improvement. For example, when large-scale aquaculture facility patches exhibit texture features such as heterogeneous spectra, narrow outlines, scattered distribution, and extremely small areas, problems often arise such as unclear patch boundaries leading to identification errors and omissions, and time-consuming and laborious manual visual interpretation. Currently, the extraction of aquaculture facility patches relies excessively on network model algorithms, with little close integration to ensure the high timeliness required for routine monitoring and to optimize and improve the monitoring system. There is a lack of in-depth research on the spatial relationship between the texture features of aquaculture facility patches and surrounding marine hydrological data such as NetCDF, as well as analysis of the driving effects of hydrodynamic environmental factors such as multidimensional ocean current data and the artificial expansion of aquaculture areas on the spatiotemporal changes of aquaculture facility patches.

[0027] This application provides a method, apparatus, equipment, and storage medium for monitoring the spatiotemporal changes of aquaculture sea areas, relating to the field of marine monitoring technology. The method, apparatus, equipment, and storage medium provided in this application can be applied to a terminal, a server, or software running on a terminal or server. In some embodiments, the terminal can be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, or vehicle-mounted terminal, but is not limited to these. The server can be configured as an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The server can also be a node server in a blockchain network. The software can be an application implementing a method for monitoring the spatiotemporal changes of aquaculture sea areas, but is not limited to the above forms.

[0028] This application can be applied to numerous general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics devices, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0029] Reference Figure 1 This application provides a method for monitoring the spatiotemporal changes of aquaculture sea areas. This method may include, but is not limited to, steps S100 to S150, as detailed below: S100: Based on the target sea area boundary data, perform super-resolution reconstruction at a set monitoring frequency to obtain the remote sensing image of the target sea area; S110: Extract scattered planar vector plots of aquaculture facilities from remote sensing images of the target sea area; S120: The centroid position of the scattered planar vector map patch is calculated using the centroid method and used as a point feature map patch. The geometry of the outer convex boundary of the point feature map patch is calculated to obtain the aquaculture range map patch of the aquaculture facility. S130: Analyze NetCDF ocean current multidimensional data and generate a raster set in a format that includes current velocities in different directions according to the step size of the set monitoring frequency; S140: Analyze the NetCDF wind field multidimensional data, generate a wind speed value vector field grid according to the step size of the set monitoring frequency, and generate the wind speed according to the wind speed value vector field grid; S150: Determine the corresponding regression coefficient diagram based on the spatial relationship between the flow velocity in different directions and the wind speed, and use the regression coefficient diagram to monitor the aquaculture facility.

[0030] For example, the overall technical approach of this application embodiment includes: First, super-resolution reconstruction is performed based on the target sea area boundary data at a monitoring frequency of half a month. After coordinate transformation and geometric correction, irregular "circle-shaped" and "grid-shaped" aquaculture facility vector maps are automatically and collaboratively extracted using an RF model. Further, the centroid point feature maps of all aquaculture facilities and the minimum range of areal feature maps of the convex envelope of the marine ranch are obtained. Then, NetCDF ocean current multidimensional data is parsed, batch-converted and extracted according to a half-month step size, and the corresponding current velocity raster dataset is calculated using bilinear interpolation and Python function formula field operations. The concurrent flow velocity and direction values ​​at the locations of aquaculture facilities are used to generate a vector point map dataset P′ with eastward and northward flow velocity attribute values. Based on NetCDF wind field multidimensional data, the raster is converted in batches at half-month steps and bilinear interpolation is performed to calculate the concurrent wind speed values ​​at the corresponding aquaculture facility locations. The attributes of all 24 vector map patches are normalized and the coordinates are unified. By further selecting and calculating the GWR model variable parameters, the model autocorrelation operation and the coefficient value layer rendering of the explanatory variables are performed to obtain different regression coefficient maps such as flow velocity and wind speed, so as to realize the analysis and evaluation of the spatiotemporal changes of aquaculture facilities in different ranches.

[0031] Optionally, the step of performing super-resolution reconstruction based on the target sea area boundary data at a set monitoring frequency to obtain the remote sensing image of the target sea area includes the following steps: Based on the target sea area boundary data, medium-resolution multispectral images are acquired at the set monitoring frequency. The medium-resolution multispectral images are stored as image datasets in batches according to the order of acquisition. Using existing control points, the image dataset is geometrically corrected and image registered to obtain a corrected TIFF format first monitoring image set for each period; wherein, each first monitoring image set for each period corresponds to each of the set monitoring frequencies; Super-resolution reconstruction is performed on the first monitoring image set for each period to generate a high-resolution multispectral image set; Stable registration points along the coastline and islands are selected to recorrect and register the high-resolution multispectral image set, and mosaicking is performed to form a complete second monitoring image set for each period of the target sea area as the remote sensing image of the target sea area.

[0032] Optionally, the step of extracting scattered planar vector patterns of aquaculture facilities from remote sensing images of the target sea area includes the following steps: Training samples were created based on remote sensing images of the target sea area; Based on the pixel classification method, the training samples are configured and classified and a random forest model is selected. The threshold for the number of training samples for each category is limited, and the random forest model is trained by the relationship between the explanatory variables and the target dataset. Based on the random forest model and the training samples, raster classification is performed to generate initial aquaculture facility classification raster with different sawtooth shapes; For local areas affected by salt and pepper noise after classification, the target aquaculture facility classification grid after noise optimization is obtained by using GIS main filtering and smoothing the area boundaries. The target aquaculture facilities are classified into raster grids and converted into vector surfaces. The geometric shape of the patches is set to maintain a jagged, non-simplified shape. By constructing a multi-component batch removal of marine patches, a surface vector pattern of aquaculture facilities is formed. The surface vector pattern of aquaculture facilities includes circular vector patterns and grid-shaped vector patterns. The planar vector pattern of the aquaculture facility is regularized. The regularization process first involves splitting and breaking down multiple components, and then the extracted planar pattern is geometrically normalized to obtain the normalized scattered planar vector pattern. The geometric normalization of the extracted planar vector patterns includes: straightening the jagged outline of the grid-shaped vector pattern to a right angle, and fitting the jagged outline of the circular vector pattern to a circle.

[0033] Optionally, the step of calculating the centroid position of the scattered planar vector map patch using the centroid method as a point feature patch, and calculating the outer convex boundary geometry of the point feature patch to obtain the aquaculture range patch of the aquaculture facility includes the following steps: The centroid method is used to calculate the centroid position of the scattered planar vector vector patch in each period to obtain the point element patch in each period, and then generate the point element vector vector patch dataset; wherein, the scattered planar vector vector patch in each period corresponds to the remote sensing image of the target sea area at each set monitoring frequency. Spatial calculations are performed using the vector data of ownership boundaries of all aquaculture ranches in the target sea area and the point feature vector map dataset to redefine and obtain the aquaculture range map for each period. Specifically, for a point feature vector map dataset of a certain period, the original boundary space of all ranches in the target sea area is first expanded outward by a buffer range to generate a new areal range map. Then, the new single ranch ownership space vector range is overlaid and analyzed with the point feature vector map dataset. The selected point feature map is subjected to convex envelope calculation to regenerate a new minimum convex surface boundary geometry. At the same time, the ranch number information is recorded. Once the convex envelope calculation is completed for all ranches in the target sea area, the enclosed range map for each period is generated. Based on the enclosed range maps of all periods, an areal feature map dataset of aquaculture facility range is generated.

[0034] Optionally, the step of parsing NetCDF ocean current multidimensional data and generating a raster set including current velocities in different directions according to the set monitoring frequency step size includes the following steps: The NetCDF ocean current multidimensional data is parsed, and GeoTIFF format raster sets of eastward and northward current velocities are generated in batches according to the set monitoring frequency step size. The method further includes the following steps: The bilinear interpolation method is used to extract the directional flow velocity values ​​into the attribute fields of the eastward and northward flow velocities corresponding to the vector points based on the GeoTIFF format raster set, generating the first vector point patch dataset with eastward and northward flow velocity attribute values.

[0035] Optionally, the step of parsing the NetCDF wind field multidimensional data, generating a wind speed value vector field grid according to the set monitoring frequency step size, and generating the wind speed based on the wind speed value vector field grid includes the following steps: The NetCDF wind field multidimensional data is parsed, and wind speed value vector field grids corresponding to each set monitoring frequency are generated in batches according to the step size of the set monitoring frequency. The wind speed is generated by interpolating the effective values ​​of adjacent pixels in the wind speed value vector field grid using a bilinear interpolation method. The method further includes the following steps: The wind speed value vector field raster, including the wind speed, is extracted in batches and added to the attribute table of the corresponding period and the same spatial location in the first vector point patch dataset to generate a second vector point patch dataset of wind speed values.

[0036] Optionally, determining the corresponding regression coefficient map based on the spatial relationship between the flow velocity in different directions and the wind speed, and using the regression coefficient map to monitor the aquaculture facility, includes the following steps: A geographic weighted regression model is used to model spatial relationships. A local regression equation is set for each spatial unit point, and a Gaussian function is selected to construct a spatial weight matrix. For the fitting problem of the geographic weighted regression model, the modified information criterion is used to seek the optimal bandwidth. Then, the model autocorrelation calculation and the coefficient value layer rendering of the explanatory variables are performed to obtain the regression coefficient map of the flow velocity, flow direction and wind speed, so as to monitor the spatiotemporal change information of the aquaculture facilities.

[0037] The following sections will provide a detailed description and explanation of some optional embodiments of this application, using specific application examples.

[0038] Specifically, the embodiments of this application include the following technical solutions: (1) Collaborative extraction of “circle-shaped” and “grid-shaped” aquaculture facility patches from high-frequency remote sensing images reconstructed by super-resolution.

[0039] Specific steps: 1) Based on the target sea area boundary data, the publicly available medium-resolution multispectral images are downloaded at a monitoring frequency of half a month. The images are stored in batches according to the order of acquisition as image datasets. The image datasets are geometrically corrected and image registered using existing control points to obtain the corrected TIFF format monitoring image set T for each period, for a total of 24 periods in one year.

[0040] Among them, geometric correction and image registration methods: Taking a nearshore marine aquaculture area as an example, Sentinel-2 multispectral images were downloaded in batches every half month. Since the original images were in the WGS84 coordinate system and projected using UTM (Universal Transverse Mercator projection), the satellite image coordinate system was converted to the CGCS2000 coordinate system using a GIS coordinate projection conversion tool. The map projection adopted the 3° zone Gauss-Kruger projection, with the central meridian at 111°E. The image format was TIFF, and the image combination table format was SHPP. Eight sets of clearly visible, stable, and unchanged building structures were selected as registration points. Geometric correction of the images was performed using a GIS image georegistration tool, and the accuracy of the generated results was verified by combining ground control points. Optimization was continued until the registration error met the requirements for patch monitoring accuracy, ultimately yielding the corrected TIFF format monitoring image set for each period.

[0041] 2) Perform super-resolution reconstruction on each monitoring image set to generate a high-resolution multispectral image set. Select stable registration points along the coast and islands and reefs to recorrect and register the generated image set, and mosaic to form a complete monitoring image set R for each period of the target sea area.

[0042] Among them, the super-resolution reconstruction method uses deep learning models such as SR4RS and S-Farm to perform super-resolution processing on each monitoring multispectral remote sensing image, and reconstructs the blurred and detail-loss-lost medium- and low-resolution multispectral images into images with a resolution of 2 to 2.5 meters.

[0043] 3) Automatically extract vectors of irregular “circle-shaped” and “grid-shaped” aquaculture facilities based on the Random Forest (RF) model.

[0044] ① Training samples are created based on long-term feature images after super-resolution.

[0045] Based on the texture features and spatial distribution of aquaculture facilities in high-resolution multispectral images, vector samples in .shp format were collected. To identify different categories of aquaculture facilities and sea areas, classification attribute values ​​were set simultaneously. During sample creation, the number of acquisition units was increased as much as possible, selecting facilities with different resolutions and densities, sizes, and textures, while avoiding interference from similar targets and blurred pixel boundaries between different categories.

[0046] ② Train a pixel-based classification model.

[0047] Based on pixel classification, training samples were configured and classified, and a Randomized Field Probe (RF) model classifier was selected. A threshold for the number of samples in each category was limited. The RF model was further trained by analyzing the relationship between explanatory variables and the target dataset. Considering the dense and compact distribution of "circle-shaped" and "grid-shaped" aquaculture facilities, to improve the accuracy of the training model and ensure a sufficient number of generated rules, optimal configuration parameters were determined through repeated experiments in developed coastal aquaculture areas. For example, in this case, a certain aquaculture area was selected, and a total of 472 samples were collected (271 aquaculture facility patches and 201 seawater patches). The optimal parameters were configured as follows: the training parameters for the RF model regression analysis included a maximum of 60 trees, a maximum tree depth of 30, and a maximum number of samples of 1000.

[0048] ③ Based on the input training model and the raster dataset to be classified, raster classification is performed to generate raster classifications of aquaculture facilities with different sawtooth shapes. Furthermore, the RF model used in this method is compared with Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and Maximum Likelihood Classification (MLC) algorithms. The RF model can better extract the essential features of "circle-shaped" and "grid-shaped" aquaculture facilities, and performs well in classifying the boundaries of densely packed aquaculture facilities. It greatly preserves the regularity of the terrain outlines, has higher classification accuracy, and less noise.

[0049] ④ For local areas affected by salt-and-pepper noise after classification, a new classification raster for aquaculture facilities is obtained after noise optimization through GIS master filtering and smoothing of area boundaries. Specifically, the eight-adjacent-cell mode filtering method is used to filter out isolated pixels or noise points from the classified raster, and irregular class boundaries are smoothed and areas are aggregated by class.

[0050] ⑤ Convert the optimized classification raster into a vector surface, and then extract and regularize it.

[0051] Each period's classification raster is converted into a vector surface, and the geometric shape of the patches is set to maintain a jagged, non-simplified shape. By constructing a multi-component batch removal of marine patches, a surface vector vector patch Z of aquaculture facilities is formed.

[0052] For each period, the extracted "circle-shaped" and "grid-shaped" aquaculture facility surface vector vector patches Z are processed for regularization. First, the multi-components are split and scattered. Then, the extracted surface vector patches are further standardized in geometric shape. Specifically, the jagged outline of the "grid-shaped" facility patches is made into right angles, and the jagged outline of the "circle-shaped" facility patches is fitted into circles, resulting in the standardized and scattered surface vector vector patches Z′.

[0053] Furthermore, the method for right-angle reduction: Using GIS rule-based geoprocessing tools, the input "grid-shaped" aquaculture facility isomorphic patches are processed. Following the geometric node order of the isomorphic patches, and based on the tolerance of the angle between corners and level angles, or the relationship between cumulative angle offset and right angles, intermediate nodes are discarded, effectively simplifying the feature edges of the patches. Then, the patch outlines are right-angled. Vertices whose corner angles and right-angle deviations are within a specified tolerance range are adjusted to ensure right angles, while also ensuring correct handling of relationships between adjacent points. Specifically, in this practical case, the maximum boundary offset distance is set to 4 meters, and the rule-based processing precision is set to 0.25.

[0054] Furthermore, the method for fitting circles: For the "circular" facility patches whose polygonal outlines, automatically extracted using deep learning or machine learning, are formed by arbitrary polylines creating a jagged pattern and tending towards circular features, the compactness field of the area patch is calculated using a GIS field calculator. A threshold is set (a value of 1 indicates a perfect circle, selecting shapes with a compactness value close to 1) for filtering, identification, and regularization of the circular structure. This makes the circular boundaries smoother and more regular, resulting in "circular" facility patches with regularized circular features. Specifically, in this practical case, a threshold of 0.7 is set to achieve the best results in fitting a circle.

[0055] Compactness Calculation formula: ; Where S represents the area of ​​the patch, C represents the perimeter of the patch, and π represents the constant pi.

[0056] 4) Obtain the centroid feature polygons of all aquaculture facilities and the minimum area feature polygons of the convex envelope of the marine ranch, and calculate them using GIS tools: The centroid method is used to calculate the centroid positions of scattered feature patches to obtain point feature patches P. A point feature vector patch dataset P{P1, P2, P3, ..., P} is generated periodically. i}

[0057] Spatial calculations were performed using vector data of ownership boundaries of all aquaculture ranches within the target jurisdiction's sea area and a dataset P of vector feature patches for all periods to accurately redefine and obtain the current aquaculture area patches for each period; specifically: for a certain period's vector feature patch dataset P i First, a new areal area map is generated by extending the original boundaries of all ranch ownership rights within the target sea area outward by a buffer zone of 2 km. Then, the new individual ranch R... i The ownership space vector range and point feature vector map dataset are overlaid for analysis. The selected point feature maps are then subjected to convex envelope calculation to regenerate a new minimum convex surface boundary geometry. Simultaneously, ranch number information is recorded. Once the convex envelope calculation is completed for all ranches within the target jurisdiction's sea area, the i-th phase outer enclosure range map W is generated. i Using the external convex envelope calculation method, a dataset of areal feature patches W{ W1, W2, W3, ..., W} is generated by iterating through the data in each phase. i}

[0058] The implementation process of the centroid method is as follows: Latitude and longitude coordinates of point feature patches are generated in batches using geometric calculations. First, the point feature P... i Create two new fields, lon and lat, in the layer's attribute table to store the coordinates of the points. The data type is text. Calculate the geometric centroids of the lon and lat fields using the same projected coordinate system as the data, and generate a point feature vector map dataset P with coordinate values ​​in the attribute table.

[0059] The implementation process of the convex envelope calculation method is as follows: First, a unified grouping field is established and assigned an identifier value. After merging all polygons with the same identifier value to generate multi-component polygons, the minimum convex boundary geometry of the input polygon is obtained using the minimum convex boundary geometry generation method, generating the closed aquatic polygon W of the aquaculture facility. Specifically: First, a unified grouping field with a unified name is established for the scattered polygons inside the minimum convex boundary to be generated, and the grouping field is assigned the values ​​{1,2,3,...} as identifier values ​​according to the construction order; then, polygons with the same identifier value in the grouping field are merged to form multi-component polygons. During the convex envelope calculation, the minimum convex boundary geometry is independently constructed based on the same identifier value of the grouping field, generating the closed aquatic polygon W of the entire aquaculture facility range; this process is repeated periodically to generate the facility range aquatic vector map polygon dataset W{ W1, W2, W3,..., W... i}

[0060] (2) Calculate the synchronous current velocity and direction at the corresponding aquaculture facility locations based on NetCDF ocean current data.

[0061] 1) Parse NetCDF ocean current multidimensional data and generate GeoTIFF format raster sets of eastward current velocity (water_u) and northward current velocity (water_v) in batches at half-month steps.

[0062] By constructing a multidimensional data processing model, the NetCDF format ocean current data with 24 time steps (StdTime=24) is processed into a long-term series with data from half a month as a single step and a cumulative duration of 1 year. The variables of eastward current velocity (water_u) and northward current velocity (water_v) are selected to generate AFR format dual-function raster layers, resulting in 24 time steps (equal time intervals) for both water_u and water_v. Then, by maintaining a unified spatial reference and using the sea area under the jurisdiction of the monitoring city for cropping, the data is output in batches in GeoTIFF format.

[0063] It should be noted that AFR format stands for ArcGIS Function Raster Layer. water_u: Eastward Water Velocity (unit: m / s, positive value is eastward); water_v: Northward Water Velocity (unit: m / s, positive value for northward flow).

[0064] Using a multidimensional data processing model, ocean current data (water_u, water_v) in NetCDF format, containing 24 time steps across the entire year, was processed as follows: First, the NetCDF raster layer creation tool was opened using GIS. The target .nc format annual ocean current file was selected, and the eastward current velocity (water_u) was chosen as the variable. Longitude (lon) was automatically set as the X dimension, and latitude (lat) as the Y dimension. An iterator was created and iterated cyclically with a half-month step size (value 1), where the loop starts at 1 and ends at 24. Then, AFR format function raster layers were generated for each time step according to the iteration loop index. The raster was copied, the projection was redefined, and the raster was output. This process was repeated cyclically to batch output GeoTIFF format water_u raster {water_u1, water_u2, ..., water_u...} with a half-month step size. 24}

[0065] Similarly, for extracting the northward current velocity (water_v) raster, open the NetCDF raster layer creation tool in GIS, select the target .nc format annual ocean current file, and select the northward current velocity (water_v) as the variable; automatically set longitude (lon) as the X dimension and latitude (lat) as the Y dimension; create an iterator and iterate cyclically with a half-month step size (value 1), where the loop start value is 1 and the end value is 24; then extract and generate an AFR format function raster layer according to the corresponding time step based on the iteration loop index value. By copying the raster tool, redefining the projection, and outputting the raster, iterate cyclically, and batch output GeoTIFF format water_v rasters {water_v1, water_v2, ..., water_v...} with a half-month step size. 24}

[0066] 2) The bilinear interpolation method is used to extract the directional flow velocity values ​​to the corresponding val_u (eastward flow velocity) and val_v (northward flow velocity) attribute fields of the vector points, generating a vector point patch dataset P′ with eastward and northward flow velocity attribute values. For the 24-period vector point patch dataset P obtained by the centroid method, the fields val_u (eastward flow velocity) and val_v (northward flow velocity) are added to its data structure table using GIS tools. These fields are of floating-point data type and numeric format. Using geoprocessing tools, the first-period vector point patch data is used as input point features, and the corresponding first-period GeoTIFF format rasters water_u1 and water_v1 are also used as input raster data. The output field names correspond to the val_u and val_v fields of the point features, respectively. Bilinear interpolation is used to calculate cell values ​​based on the effective values ​​of adjacent cells and assign them to the corresponding val_u and val_v fields. Interpolation operations automatically ignore cells with NoData values. This process is iterated until the multi-value raster data containing eastward and northward flow velocities from all 24 periods are extracted in batches and added to the attribute data of point patches in the corresponding period and at the same spatial location, generating a vector point patch dataset P′ with eastward and northward flow velocity attribute values.

[0067] Furthermore, the interpolation method aims to perform first-order interpolation on the pixel positions in the horizontal and vertical coordinate directions, and calculate the pixel value in the output image using the four effective gray values ​​(or color values) in the left, right, up, and down directions of the pixel whose mapping position is closest to the pixel point in the output image. The formula is as follows: ; In the formula, the horizontal and vertical coordinates represent the positions of image pixels. Assuming a pixel T′ in the output image has coordinates (x′, y′), its position T in the original image has coordinates (x, y); L 11L 12 L 21 L 22 Let g(m,n), g(m+1,n), g(m,n+1), and g(m+1,n+1) be the four nearest pixels to point T, with coordinates (m,n), (m+1,n), (m,n+1), and (m+1,n+1), respectively. Then g(m,n), g(m+1,n), g(m,n+1), and g(m+1,n+1) represent the grayscale values ​​(or color values) of the corresponding pixels. Let a = xm and b = yn, where m and n are integers and a and b are decimals.

[0068] 3) Based on Python function formula field calculation, the flow velocity value (val_speed) and flow direction value (val_direction) are obtained in batches. The output result is still a vector point patch dataset P′ with eastward and northward flow velocity attribute values.

[0069] ① Calculate the flow rate value (val_speed).

[0070] Using GIS tools, a field val_speed (flow velocity value) is added to the vector point patch dataset P′. Its data type is floating point, and the number format is numeric. After creation, in the same vector point patch table structure, a field calculation tool is used in conjunction with Python language to edit the mathematical calculation formula of the flow velocity value using the sqrt function on the known val_u and val_v field values, and calculate all flow velocity values ​​val_speed in the vector point patch dataset P′.

[0071] Mathematical formula for calculating flow velocity: ; Convert the calculation formula to Python: ; It should be noted that the exclamation mark in the above expression is a formatted representation of the programming language and is not garbled text.

[0072] Where val_v represents the eastward flow velocity in m / s, with positive values ​​indicating eastward movement; val_u represents the northward flow velocity in m / s, with positive values ​​indicating northward movement; and val_speed represents the magnitude of the flow velocity in m / s.

[0073] ② Calculate the flow direction value (val_direction).

[0074] Using GIS tools, a field `val_direction` (flow velocity direction value) is added to the vector point patch dataset P′. The data type is long integer, and the number format is numeric. After creation, within the same vector point patch table structure, a field calculation tool, supplemented by Python, is used to create a direction value calculation formula. The `atan2` and `degrees` functions are used to edit the mathematical calculation model for the flow velocity direction based on the known `val_u` and `val_v` field values. The direction angle is calculated in the patch attribute table, and a geographic coordinate system transformation is performed. This process yields all direction values ​​`val_direction` in the vector point patch dataset P′.

[0075] Direction value calculation formula: ; Convert the calculation formula to Python: ; It should be noted that the exclamation mark in the above expression is a formatted representation of the programming language and is not garbled text.

[0076] Where val_v represents the eastward flow velocity in m / s, with positive values ​​indicating eastward movement; val_u represents the northward flow velocity in m / s, with positive values ​​indicating northward movement; and val_direction represents the flow direction, measured from 0° clockwise from true north in degrees.

[0077] (3) Calculate the concurrent wind speed at the corresponding aquaculture facility location based on NetCDF wind field data.

[0078] ① Parse NetCDF wind field multidimensional data and generate vector field raster of wind speed values ​​(wind_speed) in batches according to a half-month step size.

[0079] First, download NetCDF format wind speed data containing the near-surface 10-meter height wind layer (HTGL) from routine meteorological observations. By constructing a multidimensional data processing model, the NetCDF format wind speed data with 24 time steps (StdTime=24) is processed into a long-term series lasting one year, with each half-month's data as a single step. The zonal component (UGRD) and meridional component (VGRD) of the wind are extracted to generate AFR format dual-function raster layers, resulting in 24 time steps (equal time intervals) each for wind_u (zonal component of wind speed) and wind_v (meridional component of wind speed). Then, GIS raster functions are used... The vector field function conversion tool in the data structure selects wind_u and wind_v as the input raster for the first step, sets the input data type to "UV", and sets the output data type to "magnitude-direction" to ensure that the output results include both wind speed (magnitude) and wind direction (direction) bands. Further calculations generate the wind speed vector field raster for the first step. Then, by maintaining a unified spatial reference, the current raster is cropped using the monitored city's jurisdictional sea area to obtain the vector field raster for the first time step. This process is iterated repeatedly, batch-outputting 24 periods of wind speed vector field raster {wind1, wind2, ..., wind...} at a half-month step size. 24}

[0080] ② The bilinear interpolation method is used to calculate the vector point patch dataset P″ with wind_speed (wind speed value) based on the effective values ​​of adjacent pixels.

[0081] For the 24-period vector point patch dataset P′ after obtaining the flow velocity and direction, a field `wind_speed` (wind speed value) is added to its data structure table using GIS tools. The data type is floating-point, and the numeric format is numerical. Using geoprocessing tools, the first-period vector point patch data is used as input point features, and the corresponding first-period wind speed vector field raster `wind1` is used as input raster data. The output field name corresponds to the `wind_speed` field of the point feature patch. Bilinear interpolation is used to calculate the cell value based on the effective values ​​of adjacent cells and assign it to the corresponding `wind_speed` field. The interpolation operation automatically ignores cells with NoData values. All input raster attributes are appended to the output point features. This process is iterated until the multi-value raster containing wind speed values ​​from the 24 periods is extracted in batches to the attribute table of the point patches in the corresponding period and at the same spatial location, generating a vector point patch dataset P″ with `wind_speed` (wind speed value).

[0082] (4) Perform attribute normalization and coordinate unification on all 24 newly generated vector point patch datasets P″. The normalized attributes include: num (ranch number), lon (longitude), lat (latitude), val_u (eastward flow velocity), val_v (northward flow velocity), val_speed (flow velocity value), val_direction (flow velocity direction value), wind_speed (wind speed value), and time (time of each period, date format).

[0083] (5) By further selecting and calculating the GWR model variable parameters, performing model autocorrelation calculations and rendering the coefficient values ​​of the explanatory variables, different regression coefficient maps such as flow velocity and wind speed are obtained to analyze and evaluate the spatiotemporal changes of different ranch aquaculture facilities. Specifically: To implement GIS spatial autocorrelation calculation, the first step is to prepare input element and variable data. Then, a geographically weighted regression (GWR) model is used to model spatial relationships, taking into account the influence of geographic location on variable relationships. A local regression equation is set for each spatial unit point, and a Gaussian function is selected to construct a spatial weight matrix. For the GWR fitting problem, the corrected akaike information criterion (AICc) is used to seek the optimal bandwidth, so as to effectively capture the spatial non-stationary relationships between variables and reflect the changes in the relationship between dependent and explanatory variables at different spatial locations.

[0084] First, prepare the data parameters. Select the original isometric map of the ranch sea ownership boundary as the input element. Use attribute connection to extract the corresponding parameter values ​​of the dependent variable (displacement and value) and the explanatory variables (average flow velocity, average wind speed, average number of aquaculture facilities, and average area exceeding the ranch sea ownership boundary) to the corresponding ranch sea ownership boundary map.

[0085] The dependent variable is selected as displacement and value. The specific calculation method is as follows: based on the previously obtained pasture outer envelope planar vector pattern dataset W{ W1, W2, W3, ..., W i Using two periods of latitude and longitude coordinates, batch numerical calculations are performed through PYTHON. The displacement value of the centroid point of the outer envelope of each pasture in the two periods is calculated one by one. The cumulative sum of the displacement values ​​of all two consecutive periods is recorded in the "Displacement Sum" field of the input feature isometric patch. The explanatory variables include four items: average current velocity, average wind speed, average number of aquaculture facilities, and average area exceeding the sea boundary of the ranch. Specifically, the average current velocity is calculated using batch numerical calculations in Python, by analyzing the areal features W{W1, W2, W3, ..., W} of each ranch. i The average flow velocity is calculated from the vector point patch dataset P″ within the geometric boundary, and then averaged again for each pasture over 24 periods. The average wind speed and the average number of aquaculture facilities are calculated in the same way as the average flow velocity. The average area exceeding the sea rights boundary of the pasture is calculated by performing an overlay and erase spatial operation on the sea rights boundary and the outer geometric range of each pasture to obtain the area value, and then averaging it for each pasture over 24 periods. Afterwards, batch attribute assignments are performed based on the pasture number field in the input features to form explanatory variable parameters.

[0086] Then, using GIS tools, the parameters and methods for GWR model calculation were set. Input elements, dependent variables, and explanatory variables were set sequentially. The data type used for modeling was set to Gaussian function model, and the bandwidth method was set to AICc (Corrected Akaike Information Criterion). Spatial autocorrelation calculation was then performed to obtain auxiliary tables with different values ​​of AICc and R2 (R-Square, goodness-of-fit test). The generated GWR layer was automatically visualized and rendered based on the standardized residuals. Then, the layers were rendered based on the coefficient values ​​of each independent variable obtained from the calculation to obtain regression coefficient maps of current velocity, wind speed, average point number, and average area. The fitting effect of regression models in different marine pastoral areas was reviewed. For different coefficient maps, the positive or negative effects were analyzed. Combining historical storm surges, historical typhoon time series paths, and human activities, the driving factors of spatiotemporal changes in aquaculture facilities were comprehensively evaluated.

[0087] Furthermore, the formula for the GWR model is: ; In the formula, Indicates the dependent variable to be predicted; For the intercept term, for the first term... A constant term for each sample point; as independent variable At sample points The regression coefficients at a given location reflect spatial heterogeneity; , ... Represent the independent variable , ... At sample points Observations at that location Represent the independent variable At sample points Observations at; This is the random error term.

[0088] GWR uses the Gaussian kernel function for calculation during parameter calibration, and its formula is as follows: ; In the formula, represents the distance between sample point i and sample point j; μ represents the non-negative decay coefficient of the functional relationship between spatial weights and distance, called bandwidth, which controls the local influence range. The smaller μ is, the stronger the weights... The decay rate increases with distance, and vice versa. The decay rate decreases with increasing distance.

[0089] As bandwidth increases, the GWR model will perform ordinary least squares regression. When the bandwidth is 0, overfitting will occur. Therefore, an adaptive bandwidth needs to be introduced, and the optimal bandwidth is determined by finding the corresponding value. The Corrected Akaike Information Criterion (AICc) is used to select the optimal bandwidth for the weight function in the GWR analysis. The formula is as follows: ; In the formula, This is the maximum likelihood estimate of the variance of the random error term. The matrix of GWR The trace is a function of bandwidth μ. In the same sample data, the bandwidth corresponding to the model weight function that minimizes the AICc value is the optimal bandwidth.

[0090] For example, embodiments of this application may also provide relevant application cases.

[0091] like Figure 2 As shown, Figure 2 Here is a real-world example image of the target sea area. Figure 3 The original remote sensing image of the target sea area, Figure 4 The target sea area is a remote sensing image after super-resolution processing. Figures 5a to 5d This is an example image showing the results of classification based on the Random Forest (RF) model (taking a "grid" pattern as an example). Figure 5a This is the classification result from the random forest model. Figure 5b This is the result of the main filter. Figure 5c This is the result of boundary clearing. Figure 5d It is a result of rule-based processing.

[0092] The beneficial effects of the embodiments of this application include at least the following: I. Starting with addressing the timeliness of monitoring, we fully utilize the advantages of satellites such as the Sentinel series, including short replay cycles, long time series continuity, wide coverage, rapid data distribution, and free access. We also combine image super-resolution technology to address the limitations of spatial resolution, restore and acquire high-resolution images, and ensure high-precision imagery through continuous multi-temporal phases to effectively meet the high-frequency monitoring needs of weekly or bi-weekly monitoring.

[0093] Second, the Random Forest (RF) model is used to automatically extract vectors of irregular "circle-shaped" and "grid-shaped" aquaculture facilities. It faces challenges such as the different spectrums of marine aquaculture facilities, narrow outlines, scattered distribution and extremely small area. It has a better classification effect on the boundary of the "grid-shaped" aquaculture facilities that are clustered and compact, with higher classification accuracy and less noise, and greatly preserves the regularity of the outline of the land features.

[0094] Third, taking into account the periodic and random influences of natural environmental factors such as ocean currents and wind speeds on the location distribution of aquaculture facilities, a spatiotemporal geographic weighted regression model is created using NetCDF data that supports multidimensional array storage, efficient reading and writing, and flexibility, in conjunction with spatiotemporal information of map patches intelligently extracted from marine facilities. This model is then used to conduct a correlation analysis with the spatiotemporal information of dynamic multidimensional marine environmental influencing factors, and to quantify the impact of significant factors in the variables on the spatiotemporal displacement of aquaculture marine facility map patches.

[0095] Reference Figure 6 This application also provides a device for monitoring the spatiotemporal changes of aquaculture sea areas, which can realize the above-mentioned method for monitoring the spatiotemporal changes of aquaculture sea areas. The device includes: The image reconstruction unit is used to perform super-resolution reconstruction based on the target sea area boundary data at a set monitoring frequency to obtain remote sensing images of the target sea area. The image patch extraction unit is used to extract scattered planar vector images of aquaculture facilities from remote sensing images of the target sea area; The aquaculture range determination unit is used to calculate the centroid position of the scattered planar vector map patch using the centroid method as a point element map patch, calculate the geometry of the outer convex boundary of the point element map patch, and obtain the aquaculture range map patch of the aquaculture facility. The velocity and direction calculation unit is used to parse NetCDF ocean current multidimensional data and generate a format raster set including flow velocities in different directions according to the step size of the set monitoring frequency. The wind speed calculation unit is used to parse NetCDF wind field multidimensional data, generate wind speed value vector field grids according to the step size of the set monitoring frequency, and generate the wind speed based on the wind speed value vector field grids. The spatiotemporal monitoring unit is used to determine the corresponding regression coefficient diagram based on the spatial relationship between the flow velocity in different directions and the wind speed, and to monitor the aquaculture facility using the regression coefficient diagram.

[0096] It is understood that the content of the above method embodiments is applicable to the present device embodiments. The specific functions implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0097] This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the method of this application. This electronic device can be any smart terminal, including tablet computers, in-vehicle computers, etc.

[0098] It is understood that the content of the above method embodiments is applicable to the device embodiments. The specific functions implemented by the device embodiments are the same as those of the methods of this application, and the beneficial effects achieved are the same as those achieved by the methods of this application.

[0099] Figure 7 The hardware structure of an electronic device according to another embodiment is illustrated. The electronic device includes: The processor 101 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application. The memory 102 can be implemented as a read-only memory (ROM), static storage device, dynamic storage device, or random access memory (RAM). The memory 102 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 102 and is called and executed by the processor 101. Input / output interface 103 is used to implement information input and output; The communication interface 104 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.). Bus 105 transmits information between various components of the device (e.g., processor 101, memory 102, input / output interface 103, and communication interface 104); The processor 101, memory 102, input / output interface 103 and communication interface 104 are connected to each other within the device via bus 105.

[0100] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method of this application.

[0101] It is understood that the content of the above method embodiments is applicable to this storage medium embodiment. The specific functions implemented in this storage medium embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0102] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0103] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0104] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0105] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0106] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0107] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0108] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0109] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0110] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0111] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0112] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0113] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0114] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.

Claims

1. A method for monitoring the spatiotemporal changes of marine environments used for aquaculture, characterized in that, The method includes the following steps: Based on the target sea area boundary data, super-resolution reconstruction is performed at a set monitoring frequency to obtain the remote sensing image of the target sea area. Extract scattered planar vector patterns of aquaculture facilities from remote sensing images of the target sea area; The centroid position of the scattered planar vector map patch is calculated using the centroid method and used as a point feature map patch. The geometry of the outer convex boundary of the point feature map patch is calculated to obtain the aquaculture range map patch of the aquaculture facility. Parse NetCDF ocean current multidimensional data and generate a raster set in a format that includes current velocities in different directions according to the step size of the set monitoring frequency; Analyze the NetCDF wind field multidimensional data, generate a wind speed value vector field grid according to the step size of the set monitoring frequency, and generate the wind speed based on the wind speed value vector field grid; Based on the spatial relationship between the flow velocity in different directions and the wind speed, a corresponding regression coefficient diagram is determined, and the regression coefficient diagram is used to monitor the aquaculture facility. The process of calculating the centroid position of the scattered planar vector map features using the centroid method, and then calculating the geometry of the outer convex boundary of the point feature map features to obtain the aquaculture range map features of the aquaculture facility includes the following steps: The centroid method is used to calculate the centroid position of the scattered planar vector vector patch in each period to obtain the point element patch in each period, and then generate the point element vector vector patch dataset; wherein, the scattered planar vector vector patch in each period corresponds to the remote sensing image of the target sea area at each set monitoring frequency. Spatial calculations are performed using the vector data of ownership boundaries of all aquaculture ranches in the target sea area and the point feature vector map dataset to redefine and obtain the aquaculture range map for each period. Specifically, for a point feature vector map dataset of a certain period, the original boundary space of all ranches in the target sea area is first expanded outward by a buffer range to generate a new areal range map. Then, the new single ranch ownership space vector range is overlaid with the point feature vector map dataset for analysis. The selected point feature map is subjected to convex envelope calculation to regenerate a new minimum convex surface boundary geometry. At the same time, the ranch number information is recorded. After the convex envelope calculation is completed for all ranches in the target sea area, the enclosed range map for each period is generated. Based on the enclosed range maps of all periods, an areal feature map dataset of aquaculture facility range is generated.

2. The method for monitoring the spatiotemporal changes of marine environments used for aquaculture according to claim 1, characterized in that, The process of performing super-resolution reconstruction based on the target sea area boundary data at a set monitoring frequency to obtain remote sensing images of the target sea area includes the following steps: Based on the target sea area boundary data, medium-resolution multispectral images are acquired at the set monitoring frequency. The medium-resolution multispectral images are stored as image datasets in batches according to the order of acquisition. Using existing control points, the image dataset is geometrically corrected and image registered to obtain a corrected TIFF format first monitoring image set for each period; wherein, each first monitoring image set for each period corresponds to each of the set monitoring frequencies; Super-resolution reconstruction is performed on the first monitoring image set for each period to generate a high-resolution multispectral image set; Stable registration points along the coastline and islands are selected to recorrect and register the high-resolution multispectral image set, and mosaicking is performed to form a complete second monitoring image set for each period of the target sea area as the remote sensing image of the target sea area.

3. The method for monitoring the spatiotemporal changes of marine environments used for aquaculture according to claim 1, characterized in that, The extraction of scattered planar vector patterns of aquaculture facilities from remote sensing images of the target sea area includes the following steps: Training samples were created based on remote sensing images of the target sea area; Based on the pixel classification method, the training samples are configured and classified and a random forest model is selected. The threshold for the number of training samples for each category is limited, and the random forest model is trained by the relationship between the explanatory variables and the target dataset. Based on the random forest model and the training samples, raster classification is performed to generate initial aquaculture facility classification raster with different sawtooth shapes; For local areas affected by salt and pepper noise after classification, the target aquaculture facility classification grid after noise optimization is obtained by using GIS main filtering and smoothing the area boundaries. The target aquaculture facilities are classified into raster grids and converted into vector surfaces. The geometric shape of the patches is set to maintain a jagged, non-simplified shape. By constructing a multi-component batch removal of marine patches, a surface vector pattern of aquaculture facilities is formed. The surface vector pattern of aquaculture facilities includes circular vector patterns and grid-shaped vector patterns. The planar vector pattern of the aquaculture facility is regularized. The regularization process first involves splitting and breaking down multiple components, and then the extracted planar pattern is geometrically normalized to obtain the normalized scattered planar vector pattern. The geometric normalization of the extracted planar vector patterns includes: straightening the jagged outline of the grid-shaped vector pattern to a right angle, and fitting the jagged outline of the circular vector pattern to a circle.

4. The method for monitoring the spatiotemporal changes of marine environments used for aquaculture according to claim 1, characterized in that, The parsed NetCDF ocean current multidimensional data is used to generate a raster set in a format that includes current velocities in different directions, according to the set monitoring frequency step size. This includes the following steps: The NetCDF ocean current multidimensional data is parsed, and GeoTIFF format raster sets of eastward and northward current velocities are generated in batches according to the set monitoring frequency step size. The method further includes the following steps: The bilinear interpolation method is used to extract the directional flow velocity values ​​into the attribute fields of the eastward and northward flow velocities corresponding to the vector points based on the GeoTIFF format raster set, generating the first vector point patch dataset with eastward and northward flow velocity attribute values.

5. The method for monitoring the spatiotemporal changes of marine environments used for aquaculture according to claim 4, characterized in that, The process of parsing NetCDF wind field multidimensional data, generating wind speed value vector field grids according to the set monitoring frequency step size, and generating the wind speed based on the wind speed value vector field grids includes the following steps: The NetCDF wind field multidimensional data is parsed, and wind speed value vector field grids corresponding to each set monitoring frequency are generated in batches according to the step size of the set monitoring frequency. The wind speed is generated by interpolating the effective values ​​of adjacent pixels in the wind speed value vector field grid using a bilinear interpolation method. The method further includes the following steps: The wind speed value vector field raster, including the wind speed, is extracted in batches and added to the attribute table of the corresponding period and the same spatial location in the first vector point patch dataset to generate a second vector point patch dataset of wind speed values.

6. A method for monitoring the spatiotemporal changes of marine environments used for aquaculture according to any one of claims 1 to 5, characterized in that, The process of determining the corresponding regression coefficient map based on the spatial relationship between the flow velocity in different directions and the wind speed, and using the regression coefficient map to monitor the aquaculture facility, includes the following steps: A geographic weighted regression model is used to model spatial relationships. A local regression equation is set for each spatial unit point, and a Gaussian function is selected to construct a spatial weight matrix. For the fitting problem of the geographic weighted regression model, the modified information criterion is used to seek the optimal bandwidth. Then, the model autocorrelation calculation and the coefficient value layer rendering of the explanatory variables are performed to obtain the regression coefficient map of the flow velocity, flow direction and wind speed, so as to monitor the spatiotemporal change information of the aquaculture facilities.

7. A marine temporal and spatial variation monitoring device for aquaculture, characterized in that, The device is used to implement the method for monitoring the spatiotemporal changes of aquaculture sea areas as described in claim 1, and the device includes: The image reconstruction unit is used to perform super-resolution reconstruction based on the target sea area boundary data at a set monitoring frequency to obtain remote sensing images of the target sea area. The image patch extraction unit is used to extract scattered planar vector images of aquaculture facilities from remote sensing images of the target sea area; The aquaculture range determination unit is used to calculate the centroid position of the scattered planar vector map patch using the centroid method as a point element map patch, calculate the geometry of the outer convex boundary of the point element map patch, and obtain the aquaculture range map patch of the aquaculture facility. The velocity and direction calculation unit is used to parse NetCDF ocean current multidimensional data and generate a format raster set including flow velocities in different directions according to the step size of the set monitoring frequency. The wind speed calculation unit is used to parse NetCDF wind field multidimensional data, generate wind speed value vector field grids according to the step size of the set monitoring frequency, and generate the wind speed based on the wind speed value vector field grids. The spatiotemporal monitoring unit is used to determine the corresponding regression coefficient diagram based on the spatial relationship between the flow velocity in different directions and the wind speed, and to monitor the aquaculture facility using the regression coefficient diagram.

8. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method as described in any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 6.

Images