A real-time cooperative monitoring system and method for green tide prevention and control in near-shore sea areas

Abstract

The application provides a real-time cooperative monitoring system and method for green tide prevention and control in nearshore sea areas, relates to the technical field of marine ecological prevention and control, and performs fusion processing on multi-source monitoring data of the nearshore sea area to obtain a fusion data set; multi-dimensional features are extracted based on the fusion data set, a multi-dimensional cooperative perception vector is constructed, intelligent diagnosis is performed on the multi-dimensional cooperative perception vector, and a green tide situation diagnosis graph is obtained; coupling growth simulation is performed on the fusion data set to obtain the growth critical depth and occurrence probability distribution of algae, and then a growth risk prediction graph of the green tide is generated; double-track early warning decisions are made according to the growth risk prediction graph and the green tide situation diagnosis graph, comprehensive early warning signals and dynamic risk fields are obtained, then cooperative path planning and task scheduling are performed on the green tide prevention and control resources of the nearshore sea area, cooperative scheduling instructions are generated, and the application can construct a dynamic risk field to drive double-track early warning and cooperative scheduling decisions, so that the initiative of green tide prevention and control in the nearshore sea area can be improved.

Description

Technical Field

[0001] This application relates to the field of marine ecological control technology, and more specifically, to a real-time collaborative monitoring system and method for controlling green tides in nearshore waters. Background Technology

[0002] Marine ecosystems are key carriers of global carbon cycling and biodiversity maintenance. However, due to multiple pressures such as land-based pollution input, seawater eutrophication and global climate change, nearshore marine ecological disasters occur frequently. These disasters not only lead to water hypoxia and benthic ecosystem degradation, but also pose a serious threat to coastal tourism, mariculture and nuclear power energy security. At present, the prevention and control of green tides in nearshore waters mainly relies on a combination of satellite remote sensing monitoring and manual patrols.

[0003] In recent years, some studies have attempted to introduce hyperspectral monitoring by unmanned aerial vehicles (UAVs), automatic navigation by unmanned vessels, and water quality buoy sensor networks, initially constructing an integrated three-dimensional monitoring system encompassing air, land, sea, and air. Simultaneously, deep learning-based remote sensing image recognition technology and ecological dynamics-based algae growth models are gradually being applied to green tide biomass estimation and drift path prediction. However, relying solely on ecological models for long-term trend prediction is insufficient to support accurate decision-making and cannot characterize the intensity and direction of risk propagation in a spatially continuous domain. This results in resource allocation still relying on human experience to translate text-based early warning instructions into specific vessel navigation paths and task allocation schemes. Furthermore, the failure to dynamically adapt to real-time changes in the green tide situation and predicted risks hinders the precise deployment of risk and resource allocation, leading to wasted limited control resources. Therefore, how to construct a dynamic risk field to drive dual-track early warning and collaborative scheduling decisions to improve the proactive predictability of green tide control in nearshore waters remains a challenge for the industry. Summary of the Invention

[0004] This application provides a real-time collaborative monitoring system and method for green tide prevention and control in nearshore waters. It can construct a dynamic risk field to drive dual-track early warning and collaborative scheduling decision-making, thereby improving the proactive predictability of green tide prevention and control in nearshore waters.

[0005] In a first aspect, this application provides a real-time collaborative monitoring method for controlling green tides in nearshore waters, the real-time collaborative monitoring method comprising the following steps: Multi-source monitoring data from nearshore waters are collected, and the multi-source monitoring data is fused to obtain a spatiotemporally synchronized fused dataset. Based on the fused dataset, multidimensional features are extracted, a multidimensional collaborative perception vector is constructed, and then the multidimensional collaborative perception vector is used for green tide intelligent diagnosis to obtain a green tide situation diagnosis map. Coupled growth simulation is performed on the fused dataset to obtain the critical growth depth and occurrence probability distribution of algae in nearshore waters. A growth risk prediction map of green tide in nearshore waters is generated based on the critical growth depth and occurrence probability distribution. Based on the growth risk prediction map and the green tide situation diagnosis map, a dual-track early warning decision is made to obtain a comprehensive early warning signal and a dynamic risk field. The integrated early warning signal and the dynamic risk field are used to coordinate the path planning and task scheduling of green tide prevention and control resources in nearshore waters, and generate coordinated scheduling instructions.

[0006] In this embodiment, multi-source monitoring data of nearshore waters are collected based on a three-dimensional monitoring network, which includes: a satellite remote sensing platform, an unmanned aerial vehicle (UAV) patrol platform, a buoy sensor network, and a radar monitoring station.

[0007] In this embodiment, the process of fusing the multi-source monitoring data to obtain a spatiotemporally synchronized fused dataset specifically includes: The multi-source monitoring data is timestamped and spatially registered to obtain a spatiotemporally unified data stream; Missing data is imputed and outlier filtering is performed on the spatiotemporally unified data stream to obtain a spatiotemporally synchronized fused dataset.

[0008] In this embodiment, extracting multidimensional features based on the fused dataset and constructing a multidimensional collaborative perception vector specifically includes: The spectral reflectance bands, surface temperature field, and chlorophyll concentration field were extracted from the fused dataset. Wavelet packet transform is performed on the spectral reflection band to extract multi-scale spectral energy feature vectors; Simultaneously calculate the spatial gradient of the surface temperature field and the chlorophyll concentration field to obtain the green tide trend feature vector; The multi-scale spectral energy feature vector and the green tide trend feature vector are concatenated and normalized to obtain a multi-dimensional collaborative sensing vector.

[0009] In this embodiment, performing green tide intelligent diagnosis on the multi-dimensional collaborative sensing vector to obtain a green tide situation diagnosis map specifically includes: The multidimensional collaborative perception vector is input into a fusion neural network model based on bidirectional long short-term memory, and then local spatial pattern features are extracted through the convolutional layer of the fusion neural network model to obtain a primary feature map. The primary feature map is modeled for global dependencies and focused on key regions by the multi-head attention layer of the fusion neural network model to obtain a weighted feature map. The temporal evolution of the weighted feature map is captured by the bidirectional long short-term memory layer of the fused neural network model, and the biomass probability of each spatial unit is output. Then, a green tide situation diagnostic map is generated based on the biomass probability.

[0010] In this embodiment, coupled growth simulation is performed on the fused dataset to obtain the critical growth depth and occurrence probability distribution of algae in nearshore waters, specifically including: Light intensity and nutrient concentration are parsed from the fused dataset and simultaneously input into a multi-factor coupled growth mechanism model. The multi-factor coupled growth mechanism model calculates the algal compensating light intensity based on the light intensity and nutrient concentration. The light attenuation coefficient of seawater in nearshore waters is obtained, and the critical growth depth is calculated based on the algal compensation light intensity and the nearshore waters. Obtain the algal suspension depth in nearshore waters, and determine the occurrence probability distribution based on the critical growth depth and the algal suspension depth.

[0011] In this embodiment, generating a growth risk prediction map for nearshore green tides based on the critical growth depth and the occurrence probability distribution specifically includes: The gradient region of the risk of green tide in nearshore waters is obtained by mapping the probability distribution of occurrence. Using the critical growth depth as the guiding field, diffusion simulation is performed on the gradient region to obtain the risk propagation field; By performing similarity correction on the risk propagation field, a growth risk prediction map of green tide in nearshore waters is obtained.

[0012] In this embodiment, the dual-track early warning decision-making based on the growth risk prediction map and the green tide situation diagnosis map, to obtain a comprehensive early warning signal and a dynamic risk field, specifically includes: The situation warning coefficient is determined using the aforementioned green tide situation diagnostic map; The probability warning coefficient is determined using the growth risk prediction map. The situational warning coefficient and the probability warning coefficient are weighted and fused to obtain a comprehensive warning signal; Based on the comprehensive early warning signal and the topographic information of the nearshore sea area, a risk spatial diffusion simulation is performed to obtain a dynamic risk field.

[0013] In this embodiment, the collaborative path planning and task scheduling of green tide control resources in nearshore waters through the dynamic risk field, and the generation of collaborative scheduling instructions specifically include: Obtain location and status information of green tide control resources in nearshore waters; Based on the dynamic risk field, multi-resource collaborative path planning is performed on the location information and the status information to obtain the collaborative scheduling instruction.

[0014] Secondly, this application provides a real-time collaborative monitoring system for the prevention and control of green tides in nearshore waters, used to execute a real-time collaborative monitoring method for the prevention and control of green tides in nearshore waters, the real-time collaborative monitoring system comprising: The fusion sensing module is used to collect multi-source monitoring data from nearshore waters, and to fuse the multi-source monitoring data to obtain a spatiotemporally synchronized fusion dataset. The situation diagnosis module is used to extract multi-dimensional features based on the fused dataset, construct a multi-dimensional collaborative perception vector, and then perform green tide intelligent diagnosis on the multi-dimensional collaborative perception vector to obtain a green tide situation diagnosis map. The evolution prediction module is used to perform coupled growth simulation on the fused dataset to obtain the critical growth depth and occurrence probability distribution of algae in nearshore waters, and generate a growth risk prediction map of green tide in nearshore waters based on the critical growth depth and occurrence probability distribution. The early warning decision module is used to make a dual-track early warning decision based on the growth risk prediction map and the green tide situation diagnosis map to obtain a comprehensive early warning signal and a dynamic risk field. The collaborative scheduling module is used to perform collaborative path planning and task scheduling for green tide prevention and control resources in nearshore waters through the dynamic risk field, and generate collaborative scheduling instructions.

[0015] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects: Multi-source monitoring data from nearshore waters is collected and fused to obtain a spatiotemporally synchronized fused dataset. Multi-dimensional features are extracted from the fused dataset to construct a multi-dimensional collaborative sensing vector. This vector is then used for intelligent diagnosis of green tides, resulting in a green tide situation diagnosis map. Coupled growth simulation is performed on the fused dataset to obtain the critical growth depth and probability distribution of algae in nearshore waters. A green tide growth risk prediction map is generated based on the critical growth depth and probability distribution. A dual-track early warning decision-making process is implemented based on the growth risk prediction map and the green tide situation diagnosis map, resulting in a comprehensive early warning signal and a dynamic risk field. The comprehensive early warning signal and the dynamic risk field are used to perform collaborative path planning and task scheduling for green tide control resources in nearshore waters, generating collaborative scheduling instructions.

[0016] Therefore, this application demonstrates that a dynamic risk field can be constructed to drive dual-track early warning and collaborative scheduling decisions, thereby improving the proactive predictability of green tide control in nearshore waters. Firstly, by collecting multi-source monitoring data from nearshore waters and performing spatiotemporal fusion, a spatiotemporally synchronized fusion dataset is obtained, solving the underlying problems of information fragmentation and inconsistent spatiotemporal benchmarks among various platforms in the three-dimensional monitoring network. Based on the fusion dataset, multi-dimensional features are extracted and multi-dimensional collaborative sensing vectors are constructed. A green tide situation diagnostic map is generated through a green tide intelligent diagnostic model, enabling real-time, high-precision inversion of the current spatial distribution and biomass density of green tides. Secondly, by performing coupled growth simulation on the fusion dataset, the critical depth and probability distribution of algal growth are calculated, and a growth risk prediction map is generated, which is beneficial for identifying the potential and risk areas of green tides. The mechanism-driven forward-looking prediction of the domain supplements the missing time dimension of single status quo perception. Then, a dual-track early warning decision-making system is carried out based on the growth risk prediction map and the green tide situation diagnosis map. By integrating the current situation and predicted risks, a comprehensive early warning signal is obtained. Furthermore, a dynamic risk field with spatial continuity and gradient attributes is constructed, which can transform discrete and static early warning levels into a continuous and dynamic risk propagation map, thereby building a geometric mapping bridge from early warning information to spatial decision-making. Finally, the comprehensive early warning signal and dynamic risk field are used to coordinate the path planning and task scheduling of prevention and control resources, generating collaborative scheduling instructions. This is conducive to achieving adaptive matching between the spatial distribution of risks and the intensity of prevention and control resource deployment, thereby improving the input-output efficiency of limited prevention and control forces in a dynamic risk environment.

[0017] In summary, the technical solution adopted in this application can construct a dynamic risk field to drive dual-track early warning and collaborative scheduling decisions, thereby improving the proactive predictability of green tide prevention and control in nearshore waters. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart of a real-time collaborative monitoring method for controlling green tides in nearshore waters, provided in this application. Figure 2 This is an exemplary flowchart for determining multidimensional collaborative sensing vectors provided in this application; Figure 3 This is an exemplary flowchart for determining a growth risk prediction map provided in this application; Figure 4This is a module structure diagram of a real-time collaborative monitoring system for the prevention and control of green tides in nearshore waters, provided in this application. Detailed Implementation

[0020] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0021] This application provides a real-time collaborative monitoring system and method for green tide control in nearshore waters. The core of this system involves collecting multi-source monitoring data from nearshore waters, fusing the data to obtain a spatiotemporally synchronized fused dataset, extracting multi-dimensional features from the fused dataset to construct a multi-dimensional collaborative sensing vector, and then performing intelligent diagnosis of the green tide using this vector to obtain a green tide situation diagnosis map. Coupled growth simulation is performed on the fused dataset to obtain the critical growth depth and probability distribution of algae in nearshore waters, and a green tide growth risk prediction map is generated based on the critical growth depth and probability distribution. A dual-track early warning decision is made based on the growth risk prediction map and the green tide situation diagnosis map to obtain a comprehensive early warning signal and a dynamic risk field. The comprehensive early warning signal and the dynamic risk field are used to perform collaborative path planning and task scheduling for green tide control resources in nearshore waters, generating collaborative scheduling instructions.

[0022] Example 1: To better understand the above technical solution, the following will provide a detailed description of the technical solution in conjunction with the accompanying drawings and specific implementation methods. (Refer to...) Figure 1 As shown in the figure, this is a flowchart of a real-time collaborative monitoring method for controlling green tides in nearshore waters according to this embodiment of the present application. The real-time collaborative monitoring method includes the following steps: In step S1, multi-source monitoring data of nearshore waters are collected, and the multi-source monitoring data are fused to obtain a spatiotemporally synchronized fused dataset.

[0023] In this specific implementation, a multi-source monitoring network is used to collect nearshore marine data. This network includes a satellite remote sensing platform, a drone patrol platform, a buoy sensor network, and radar monitoring stations. The satellite remote sensing platform is an open-source platform for acquiring multispectral remote sensing image data of nearshore marine areas. This multispectral image data includes surface reflectance data in the blue, green, red, near-infrared, and short-wave infrared bands. The drone patrol platform is a drone system equipped with a hyperspectral imager, a visible light camera, and an infrared thermal imager, used to collect hyperspectral data. Imagery and visible light video data; the buoy sensor network refers to an array of monitoring buoys deployed in nearshore waters, including temperature sensors, salinity sensors, chlorophyll fluorescence sensors, dissolved oxygen sensors, and nutrient analyzers, used to continuously collect time-series water quality and hydrological data, including surface water temperature, salinity, chlorophyll concentration, dissolved oxygen concentration, nitrate concentration, and phosphate concentration; the radar monitoring station refers to a high-frequency ground wave radar system used to acquire raster data of sea surface current fields in nearshore waters, including sea surface current field, wave field, and wind speed and direction data, wherein the sea surface current field includes the direction and velocity vector information of surface currents. It should be noted that the multi-source monitoring data in this application is a collection of raw observation data with different physical meanings and data formats, including: multispectral remote sensing image data, hyperspectral imagery and visible light video data, water quality and hydrological time-series data, and sea surface current field raster data.

[0024] In this embodiment, the process of fusing the multi-source monitoring data to obtain a spatiotemporally synchronized fused dataset specifically includes: The multi-source monitoring data is timestamped and spatially registered to obtain a spatiotemporally unified data stream; Missing data is imputed and outlier filtering is performed on the spatiotemporally unified data stream to obtain a spatiotemporally synchronized fused dataset.

[0025] In specific implementation, firstly, the acquisition time of multi-source monitoring data is unified to Coordinated Universal Time (UTC) through the Global Positioning System (GPS) timing service to obtain a data stream with a unified time reference. Then, the data with different spatial resolutions is resampled to a grid of the same size using a bilinear interpolation algorithm, and the resampled data is used as a spatiotemporally unified data stream. Next, the Kriging interpolation algorithm is used to spatially interpolate and fill in missing pixels in the spatiotemporally unified data stream caused by cloud cover or sensor failure, resulting in a continuous, hole-free filled data stream. Then, the median filtering algorithm is used to smooth high-frequency noise and isolated outliers in the filled data stream to obtain a filtered data stream. The multi-source observations of each grid cell in the filtered data stream are encapsulated and stored according to a preset data structure, and the encapsulated and stored dataset is used as a spatiotemporally synchronized fused dataset.

[0026] It should be noted that the fusion dataset in this application refers to a multi-dimensional array structure formed by organizing multi-source heterogeneous monitoring data in a grid-like manner under a unified spatiotemporal benchmark. Each grid cell corresponds to a spatial location, each time slice corresponds to an observation time, and each feature dimension corresponds to a monitoring indicator.

[0027] In step S2, multi-dimensional features are extracted based on the fused dataset to construct a multi-dimensional collaborative perception vector, and then the multi-dimensional collaborative perception vector is used for green tide intelligent diagnosis to obtain a green tide situation diagnosis map.

[0028] Preferably, in this embodiment, reference Figure 2 As shown, this diagram is an exemplary flowchart for determining a multidimensional collaborative sensing vector according to the present application. In this embodiment, the extraction of multidimensional features based on the fused dataset and the construction of the multidimensional collaborative sensing vector can be achieved through the following steps: First, in step S21, the spectral reflectance band, surface temperature field, and chlorophyll concentration field are extracted from the fused dataset; Secondly, in step S22, wavelet packet transform is performed on the spectral reflection band to extract multi-scale spectral energy feature vectors; Then, in step S23, the spatial gradient of the surface temperature field and the chlorophyll concentration field is calculated simultaneously to obtain the green tide trend feature vector. Finally, in step S24, the multi-scale spectral energy feature vector and the green tide trend feature vector are spliced ​​and normalized to obtain a multi-dimensional collaborative sensing vector.

[0029] In practice, firstly, each grid cell in the fused dataset is traversed, and the reflectance values ​​of that grid cell in the blue, green, red, and near-infrared bands are extracted. The extracted reflectance values ​​of the four bands are arranged in band order to obtain the spectral reflectance band of that grid cell. Simultaneously, the surface water temperature and chlorophyll concentration of that grid cell are extracted. The discrete buoy observations are extended into a continuous grid using an inverse distance weighted interpolation algorithm. The interpolated continuous grid is used as the surface temperature field and chlorophyll concentration field, respectively. Secondly, wavelet packet decomposition is performed on the spectral reflectance band of each grid cell. The energy values ​​of each frequency band obtained after decomposition are calculated, and the calculated energy values ​​of each frequency band are arranged into a vector shape in frequency band order. The formula is used to calculate the first-order partial derivatives of the surface temperature field in the horizontal and vertical directions using the Sobel operator. The magnitude of the composite vector of the gradient components in the two directions is used as the temperature gradient amplitude. The same calculation process is used to obtain the chlorophyll concentration gradient amplitude. The temperature gradient amplitude and the chlorophyll concentration gradient amplitude are normalized and then concatenated into a two-dimensional vector. This two-dimensional vector is then used as the green tide trend feature vector. Finally, the multi-scale spectral energy feature vector and the green tide trend feature vector of each grid cell are concatenated end to end. The concatenated vector is normalized and then used as the multi-dimensional collaborative sensing vector.

[0030] It should be noted that the spectral reflectance bands in this application refer to the numerical sequence of the ability of ground objects to reflect solar radiation within different wavelength ranges. The blue light band, green light band, red light band, and near-infrared band are the four most commonly used spectral bands in optical remote sensing, corresponding to typical ground object features such as chlorophyll absorption, vegetation reflectance, soil brightness, and water body boundaries, respectively. Inverse distance weighted interpolation is a deterministic interpolation algorithm based on spatial similarity, suitable for extending discrete point observation data into a continuous distribution field. Wavelet packet decomposition is a time-frequency analysis method for finely dividing the full frequency band of a signal, capable of simultaneously extracting low-frequency approximate information and high-frequency detail information. The Sobel operator is a discrete differential operator that calculates the gradient magnitude of a grayscale image through a two-dimensional convolution operation between the image and horizontal and vertical convolution kernels. The multidimensional collaborative sensing vector refers to the numerical vector representing the marine environmental state and the response potential of green tide organisms.

[0031] In this embodiment, performing green tide intelligent diagnosis on the multi-dimensional collaborative sensing vector to obtain a green tide situation diagnosis map specifically includes: The multidimensional collaborative perception vector is input into a fusion neural network model based on bidirectional long short-term memory, and then local spatial pattern features are extracted through the convolutional layer of the fusion neural network model to obtain a primary feature map. The primary feature map is modeled for global dependencies and focused on key regions by the multi-head attention layer of the fusion neural network model to obtain a weighted feature map. The temporal evolution of the weighted feature map is captured by the bidirectional long short-term memory layer of the fused neural network model, and the biomass probability of each spatial unit is output. Then, a green tide situation diagnostic map is generated based on the biomass probability.

[0032] In specific implementation, firstly, the multidimensional collaborative perception vectors of all grid cells within the study area are organized into a three-dimensional tensor according to spatial arrangement. This three-dimensional tensor is used as the input feature tensor of a fusion neural network model based on bidirectional long short-term memory. The input feature tensor is then subjected to two-dimensional convolution operation through the first convolutional layer of the fusion neural network model. Local weighted sums are calculated by sliding the convolution kernels in the spatial dimension. The feature map output from the first convolutional layer is used as a shallow spatial feature map. The shallow spatial feature map is then subjected to a second convolution operation through the second convolutional layer of the fusion neural network model. A larger number of convolution kernels are used for higher-level feature abstraction. The feature map output from the second convolutional layer is then used as a primary feature map. Then, the primary feature map is flattened in the spatial dimension, and the resulting serialized features are used as the input sequence of the attention layer. The input sequence is self-attention calculated by a multi-head attention layer that integrates the neural network model. Each sequence position is assigned a correlation weight with all other sequence positions. The weighted features output by multiple attention heads are concatenated and fused. The resulting serialized features are then reshaped back into a three-dimensional tensor form, and the reshaped feature map is used as the weighted feature map. Finally, feature sequences of the same spatial location at different times are extracted from the weighted feature maps of multiple consecutive observation times. The feature sequences of each spatial location are used as the input sequences of the bidirectional long short-term memory layer. The bidirectional long short-term memory layer of the fusion neural network model is used to propagate the input sequences forward and backward. The forward hidden state sequences and the backward hidden state sequences are concatenated at each time step. The temporal fusion features obtained after concatenation are used as the deep temporal features of each spatial location. The deep temporal features are linearly transformed by the fully connected layer of the fusion neural network model. The scalar values ​​obtained after the linear transformation are compressed to the interval between zero and one through the activation function. The compressed values ​​are used as the biomass probability of the spatial unit at the current time step. The biomass probability of each spatial unit is then filled into a two-dimensional grid according to the original spatial arrangement order. The filled two-dimensional grid image is then used as the green tide situation diagnostic map.

[0033] It should be noted that the multi-head attention layer in this application captures global dependencies in different semantic subspaces through multiple parallel attention computation branches, and fuses the outputs of each branch to form a feature enhancement structure, which is used to strengthen the feature representation of key regions and suppress the feature response of irrelevant regions; the bidirectional long short-term memory layer is a recurrent neural network structure composed of a forward long short-term memory network and a backward long short-term memory network; the biomass probability is the probability that there is green tide algae aggregation or that it is in a state of rapid proliferation in the current spatial unit; the green tide situation diagnosis map is a spatial distribution map describing the probability of green tide occurrence in various spatial units in the nearshore sea area, and is the core output product for real-time situational awareness and visualization decision-making of green tide disasters.

[0034] In step S3, coupled growth simulation is performed on the fused dataset to obtain the critical growth depth and occurrence probability distribution of algae in nearshore waters. Based on the critical growth depth and occurrence probability distribution, a growth risk prediction map of green tide in nearshore waters is generated.

[0035] In this embodiment, coupled growth simulation is performed on the fused dataset to obtain the critical growth depth and occurrence probability distribution of algae in nearshore waters, specifically including: Light intensity and nutrient concentration are parsed from the fused dataset and simultaneously input into a multi-factor coupled growth mechanism model. The multi-factor coupled growth mechanism model calculates the algal compensating light intensity based on the light intensity and nutrient concentration. The light attenuation coefficient of seawater in nearshore waters is obtained, and the critical growth depth is calculated based on the algal compensation light intensity and the nearshore waters. Obtain the algal suspension depth in nearshore waters, and determine the occurrence probability distribution based on the critical growth depth and the algal suspension depth.

[0036] In specific implementation, firstly, each grid cell in the fused dataset is traversed, and the cumulative daily surface photosynthetically active radiation value of that grid cell is extracted, which is used as the light intensity. Simultaneously, the nitrate concentration and phosphate concentration of that grid cell are extracted, and the numerical values ​​of the nitrate concentration and phosphate concentration are compared. The smaller of the two concentration values ​​is used as the limiting nutrient concentration. The light intensity and limiting nutrient concentration are simultaneously input into a multi-factor coupled growth mechanism model. Preferably, the multi-factor coupled growth mechanism model can use the Mitchell-Ries-Menton equation to describe the nutrient limitation effect and the Steer equation to describe the light inhibition effect. Then, by solving the critical condition that the product of the nutrient limitation coefficient and the light limitation coefficient equals the algal respiration consumption coefficient, the obtained light intensity value can be used as the algal compensation light intensity. Then, the concentrations of suspended particulate matter, colored dissolved organic matter, and chlorophyll are extracted from the fused dataset. The concentrations of suspended particulate matter, colored dissolved organic matter, and chlorophyll are substituted into the empirical relationship of inherent optical parameters of seawater to calculate the diffuse attenuation coefficient of each grid cell. The diffuse attenuation coefficient can be used as the seawater light attenuation coefficient. According to Beer-Lambert's law, a functional relationship is established between the sea surface light intensity, the seawater light attenuation coefficient, and the underwater light intensity attenuation with depth. The algal compensation light intensity is substituted into the left side of the functional relationship as the target light intensity value. The water depth value corresponding to the light intensity exactly attenuating to the algal compensation light intensity is solved inversely. This water depth value can be used as the critical growth depth of the grid cell. Finally, the algae suspension depth in the nearshore waters is obtained, which is determined by statistical analysis of historical observation data. The ratio of the critical growth depth to the algae suspension depth of each grid cell is calculated, and this ratio is substituted into a preset probability mapping function. The probability mapping function is a monotonically increasing function with a value range between zero and one, and the function output value is used as the occurrence probability value of that grid cell. The occurrence probability values ​​of all grid cells are filled into a two-dimensional grid in spatial order, and the resulting two-dimensional grid is used as the occurrence probability distribution.

[0037] It should be noted that, in this application, light intensity refers to the photosynthetically active radiation energy passing vertically through a unit area per unit time, which is the core environmental factor driving algal photosynthesis; nutrient concentration refers to the content of dissolved inorganic nutrients such as nitrates and phosphates in seawater, which is a key resource element limiting the growth rate of algae; the Mitchell-Ries-Menton equation is a classic mathematical model describing the relationship between substrate concentration and reaction rate in enzyme-catalyzed reaction kinetics, used to characterize the degree of limitation of nutrient concentration on algal growth; the Steer equation is an empirical model describing the effect of light intensity on the rate of photosynthesis, capable of simultaneously characterizing light-limiting and photoinhibition effects; algal compensating light intensity refers to the critical light intensity at which the organic matter produced by algal photosynthesis exactly offsets the organic matter consumed by respiration. Below this intensity, algae cannot achieve net growth; the seawater light attenuation coefficient refers to the proportion of light intensity attenuation per unit distance when light propagates in seawater, used to describe the overall effect of seawater, dissolved organic matter, phytoplankton, and suspended particulate matter on light absorption and scattering; the critical growth depth refers to a depth from the sea surface downwards where the average light intensity is exactly equal to the algal compensating light intensity, and when the actual suspension depth of algae is less than the critical growth depth, algae can achieve net biomass growth; algal suspension depth refers to the average vertical distribution depth of floating Ulva prolifera algae in the water column; the occurrence probability distribution refers to the spatial distribution map of the probability of green tide outbreaks based on environmental driving factors and algal physiological and ecological parameters, characterizing the degree to which different sea areas have the conditions for rapid green tide biomass growth in the future.

[0038] Preferably, in this embodiment, reference Figure 3 As shown, this figure is an exemplary flowchart for determining a growth risk prediction map according to the present application. In this embodiment, generating a growth risk prediction map of nearshore green tides based on the critical growth depth and the occurrence probability distribution can be achieved by the following steps: First, in step S31, the gradient region of the risk of green tide occurring in nearshore waters is obtained based on the probability distribution of occurrence. Then, in step S32, the gradient region is subjected to diffusion simulation using the critical growth depth as the guiding field to obtain the risk propagation field; Finally, in step S33, the risk propagation field is corrected for similarity to obtain a growth risk prediction map of green tide in nearshore waters.

[0039] In specific implementation, firstly, all grid cells in the probability distribution that are greater than a preset probability threshold are marked as high-risk seed cells. Taking each high-risk seed cell as the starting point for growth, the occurrence probability value of adjacent grid cells is determined sequentially by an eight-neighborhood region growth algorithm to determine whether it is greater than the connectivity threshold, thereby identifying the gradient region of the risk of green tide in nearshore waters. The probability threshold can be set according to the actual climate. For example, based on actual experience, the probability threshold for spring can be set to 0.5, and the probability threshold for winter can be set to 0.8. Then, sea surface velocity vector fields from multiple consecutive observation times are extracted from the fused dataset, and these sea surface velocity vector fields are used as the driving field for advection transport. Using the occurrence probability value of each grid cell in the gradient region as the initial concentration distribution and the growth critical depth value as the proliferation rate coefficient, a convective diffusion numerical model including advection, diffusion, and source terms is constructed. The finite difference method is used to solve the convective diffusion numerical model in a spatiotemporal discretization manner, and the spatial distribution of risk concentration at each time within a future preset time period is iteratively deduced step by step in the time dimension. The maximum value of risk concentration at each time within the future preset time period is synthesized pixel by pixel, and the synthesized raster layer is used as the risk propagation field. Finally, data on the actual distribution range and drift path of green tide disasters during the same period in historical years were obtained, and this data was used as a correction template. The spatial morphological similarity between the risk propagation field and the correction template was calculated. An affine transformation algorithm was used to perform spatial translation, rotation, and scaling correction on the risk propagation field, which can maximize the match between the spatial distribution of the corrected risk propagation field and the characteristic distribution of the correction template. The corrected risk propagation field was then normalized, and the resulting raster layer was used as a prediction map of the growth risk of green tides in nearshore waters.

[0040] It should be noted that the region growing algorithm in this application is an image segmentation algorithm that starts from a seed pixel and gradually merges pixels in the neighborhood that meet the growth conditions into the same region according to a preset similarity criterion; the gradient region refers to a spatially connected high-risk spatial continuous block, representing the core source area where green tide disasters are most likely to occur or spread rapidly; the sea surface current velocity vector field refers to the two-dimensional vector distribution of the horizontal direction and velocity of surface currents obtained by radar monitoring stations, used to simulate the advection transport of green tide drift paths; the convection diffusion numerical model is a partial differential equation describing the process of material migration and turbulent mixing under the action of the flow field, where the advection term represents the directional transport effect of the ocean current on the green tide algae, and the diffusion term represents the tidal turbulence. The source term characterizes the contribution of green tide algae to the in-situ proliferation of green tide algae under suitable environments; the finite difference method is a computational means of converting the partial differential equation into a system of algebraic equations for numerical solution by discretizing the continuous solution domain into a finite number of grid nodes and approximating the partial derivatives with the difference quotient; the risk propagation field refers to the spatial continuous distribution of green tide disaster risk concentration in the future period generated by numerical simulation, which integrates environmental suitability, resource availability, and dynamic transport conditions; affine transformation is a linear geometric transformation that preserves the parallel relationship of straight lines in a two-dimensional image, including basic transformation forms such as translation, rotation, scaling, and shearing; the growth risk prediction map is a forward-looking prediction map characterizing the probability of green tide disasters occurring in various spatial units of nearshore waters in the future period.

[0041] In step S4, a dual-track early warning decision is made based on the growth risk prediction map and the green tide situation diagnosis map to obtain a comprehensive early warning signal and a dynamic risk field.

[0042] In this embodiment, the dual-track early warning decision-making based on the growth risk prediction map and the green tide situation diagnosis map, to obtain a comprehensive early warning signal and a dynamic risk field, specifically includes: The situation warning coefficient is determined using the aforementioned green tide situation diagnostic map; The probability warning coefficient is determined using the growth risk prediction map. The situational warning coefficient and the probability warning coefficient are weighted and fused to obtain a comprehensive warning signal; Based on the comprehensive early warning signal and the topographic information of the nearshore sea area, a risk spatial diffusion simulation is performed to obtain a dynamic risk field.

[0043] In specific implementation, firstly, each grid cell in the green tide situation diagnosis map is traversed, and the biomass probability of that grid cell is read. The average of all biomass probabilities is used as the situation warning coefficient. Secondly, each grid cell in the growth risk prediction map is traversed, and the normalized risk value of that grid cell is read. This normalized risk value is used as the probability warning coefficient. Then, a first weight value is assigned to the situation warning coefficient, and a second weight value is assigned to the probability warning coefficient. The sum of the first and second weight values ​​is one. The first weight value can be set according to the actual season. For example, based on practical experience, the first weight value for spring is set to 0.6, and the first weight value for winter is set to 0.4. The product of the situation warning coefficient and the first weight value is calculated, and the product of the probability warning coefficient and the second weight value is calculated simultaneously. The two product results are summed, and the summed value is used as the comprehensive warning signal. Finally, topographic information of the nearshore waters is acquired, including the vector boundary of the coastline, the distribution of tidal channels, water depth grid data, and the coordinates of offshore structures. The geographical center point is used as the seed point of the risk source, and the comprehensive early warning signal is used as the risk intensity of the seed point. A cellular automata model is used to simulate the spatial diffusion of risk. The evolution rule of the cellular automata is set as follows: the risk intensity is mainly diffused in the direction of ocean current vector, the topographic obstacle is used as the insurmountable barrier boundary, and the water depth gradient is used as the weight coefficient of the diffusion rate. The risk intensity value of each cell after a preset diffusion time is calculated through iterative evolution. The risk intensity values ​​of all grid cells after the iteration convergence are filled into a two-dimensional grid in spatial order. The two-dimensional grid obtained after filling is used as the dynamic risk field.

[0044] It should be noted that the situational warning coefficient in this application represents the real-time severity of the green tide disaster at the current moment; the probability warning coefficient represents the potential outbreak probability of the green tide disaster in the future period; the comprehensive warning signal is a dual-track warning decision result describing the real-time situation and predicted risk; the cellular automaton is a dynamic model driven by local evolution rules on a discrete spatiotemporal grid, where the state update of each cell depends only on its own and its neighboring cells' current state, used to simulate the continuous diffusion process of risk in space; the dynamic risk field refers to a spatial distribution map of risk intensity with continuous gradient attributes generated by a spatial diffusion model based on the spatial distribution of the comprehensive warning signal and taking into account ocean current transport capacity, topographic barrier effect, and water depth gradient influence. The value of each grid cell represents the urgency of the location needing to be prioritized at the current decision moment.

[0045] In step S5, the integrated early warning signal and the dynamic risk field are used to conduct collaborative path planning and task scheduling for green tide prevention and control resources in nearshore waters, and collaborative scheduling instructions are generated.

[0046] In this embodiment, the collaborative path planning and task scheduling of green tide control resources in nearshore waters through the dynamic risk field, and the generation of collaborative scheduling instructions specifically include: Obtain location and status information of green tide control resources in nearshore waters; Based on the dynamic risk field, multi-resource collaborative path planning is performed on the location information and the status information to obtain the collaborative scheduling instruction.

[0047] In practical implementation, firstly, the green tide control resources in nearshore waters include salvage vessels, patrol drones, and interception nets. The real-time latitude and longitude coordinates, ground speed, ground heading, and current cargo weight of the salvage vessel can be received through the Automatic Identification System (AIS) as the location and status information of the salvage vessel. The takeoff point coordinates, remaining battery percentage, and mission type code of the patrol drone can be received through the drone ground control station as the location and status information of the drone. The latitude and longitude coordinates of the interception net deployment points and the current opening and closing status can be received through the fixed interception net control terminal as the location and status information of the interception net. Then, the dynamic risk field is used as the spatial travel cost, and the higher the risk intensity, the higher the priority of the grid cell that needs to be covered and dealt with. Taking the cell with the highest risk intensity as the optimization objective and minimizing the total flight distance or total energy consumption as the secondary optimization objective, a multi-objective and multi-resource collaborative path planning model is constructed, which includes the waypoint sequence of salvage ships, the boundary of the UAV patrol area, and the decision variables of the opening and closing time of the interception net. The model is solved by ant colony algorithm. In the state transition rule, the gradient information of the dynamic risk field is used as a heuristic factor to guide the artificial ants to move towards the grid cells with high risk intensity and fast gradient rise. After multiple iterations, the optimal cruise and salvage pathpoint sequence of each salvage ship, the optimal collaborative patrol area boundary polygon of each UAV, and the optimal opening and closing time sequence of each interception net are obtained. An instruction message containing resource identity, task execution type, task execution location coordinates, task execution timestamp, and waypoint sequence coordinates is generated, and this instruction message is used as a collaborative scheduling instruction.

[0048] It should be noted that the Automatic Identification System (AIS) is a shore-based ship traffic management system that continuously and automatically broadcasts static ship information and dynamic navigation parameters; the UAV ground control station is a comprehensive ground command platform for planning, monitoring, and data link communication of UAV flight missions; the multi-objective, multi-resource collaborative path planning model is an operations research optimization model that considers multiple resource types, multiple conflicting optimization objectives, and dynamic risk spatial distribution constraints; the ant colony algorithm is a heuristic swarm intelligence optimization algorithm that simulates the positive feedback mechanism of pheromones in ants foraging, and is suitable for combinatorial optimization path planning problems; the heuristic factor is a weighted coefficient used to guide the search direction in the calculation of state transition probability; and the collaborative scheduling instruction is a standardized digital instruction set that directs different types and locations of green tide prevention and control resources to perform collaborative operations, which can be automatically parsed and executed by the execution terminal through a wireless communication network.

[0049] In summary, the technical solution adopted in this application can construct a dynamic risk field to drive dual-track early warning and collaborative scheduling decisions, thereby improving the proactive predictability of green tide prevention and control in nearshore waters.

[0050] Example 2: This application provides a real-time collaborative monitoring system for the prevention and control of green tides in nearshore waters, referring to... Figure 4 As shown, this figure is a modular structure diagram of a real-time collaborative monitoring system for green tide control in nearshore waters according to the present application. The real-time collaborative monitoring system includes: The fusion sensing module 100 is used to collect multi-source monitoring data of nearshore sea areas, and to perform fusion processing on the multi-source monitoring data to obtain a spatiotemporally synchronized fusion dataset. The situation diagnosis module 200 is used to extract multi-dimensional features based on the fused dataset, construct a multi-dimensional collaborative perception vector, and then perform green tide intelligent diagnosis on the multi-dimensional collaborative perception vector to obtain a green tide situation diagnosis map. The evolution prediction module 300 is used to perform coupled growth simulation on the fused dataset to obtain the critical growth depth and occurrence probability distribution of algae in nearshore waters, and generate a growth risk prediction map of green tide in nearshore waters based on the critical growth depth and occurrence probability distribution. The early warning decision module 400 is used to make a dual-track early warning decision based on the growth risk prediction map and the green tide situation diagnosis map to obtain a comprehensive early warning signal and a dynamic risk field. The collaborative scheduling module 500 is used to perform collaborative path planning and task scheduling for green tide prevention and control resources in nearshore waters through the dynamic risk field, and generate collaborative scheduling instructions.

[0051] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0052] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, including read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-Erasable Programmable Read-Only Memory (EEPROM), compactdisc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium capable of carrying or storing data.

[0053] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

Claims

1. A real-time collaborative monitoring method for controlling green tides in nearshore waters, characterized in that, The real-time collaborative monitoring method includes the following steps: Multi-source monitoring data from nearshore waters are collected, and the multi-source monitoring data are fused to obtain a spatiotemporally synchronized fused dataset. Based on the fused dataset, multidimensional features are extracted, a multidimensional collaborative perception vector is constructed, and then the multidimensional collaborative perception vector is used for green tide intelligent diagnosis to obtain a green tide situation diagnosis map. Coupled growth simulation is performed on the fused dataset to obtain the critical growth depth and occurrence probability distribution of algae in nearshore waters. A growth risk prediction map of green tide in nearshore waters is generated based on the critical growth depth and occurrence probability distribution. Based on the growth risk prediction map and the green tide situation diagnosis map, a dual-track early warning decision is made to obtain a comprehensive early warning signal and a dynamic risk field. The integrated early warning signal and the dynamic risk field are used to coordinate the path planning and task scheduling of green tide prevention and control resources in nearshore waters, and generate coordinated scheduling instructions.

2. The real-time collaborative monitoring method for controlling green tides in nearshore waters as described in claim 1, characterized in that, Multi-source monitoring data of nearshore waters are collected based on a three-dimensional monitoring network, which includes: a satellite remote sensing platform, an unmanned aerial vehicle (UAV) patrol platform, a buoy sensor network, and a radar monitoring station.

3. The real-time collaborative monitoring method for controlling green tides in nearshore waters as described in claim 1, characterized in that, The multi-source monitoring data is fused to obtain a spatiotemporally synchronized fused dataset, specifically including: The multi-source monitoring data is timestamped and spatially registered to obtain a spatiotemporally unified data stream; Missing data is imputed and outlier filtering is performed on the spatiotemporally unified data stream to obtain a spatiotemporally synchronized fused dataset.

4. The real-time collaborative monitoring method for controlling green tides in nearshore waters as described in claim 1, characterized in that, Extracting multidimensional features from the fused dataset and constructing a multidimensional collaborative perception vector specifically includes: The spectral reflectance bands, surface temperature field, and chlorophyll concentration field were extracted from the fused dataset. Wavelet packet transform is performed on the spectral reflection band to extract multi-scale spectral energy feature vectors; Simultaneously calculate the spatial gradient of the surface temperature field and the chlorophyll concentration field to obtain the green tide trend feature vector; The multi-scale spectral energy feature vector and the green tide trend feature vector are concatenated and normalized to obtain a multi-dimensional collaborative sensing vector.

5. The real-time collaborative monitoring method for controlling green tides in nearshore waters as described in claim 1, characterized in that, The green tide intelligent diagnosis of the multi-dimensional collaborative sensing vector to obtain the green tide situation diagnosis map specifically includes: The multidimensional collaborative perception vector is input into a fusion neural network model based on bidirectional long short-term memory, and then local spatial pattern features are extracted through the convolutional layer of the fusion neural network model to obtain a primary feature map. The primary feature map is modeled for global dependencies and focused on key regions by the multi-head attention layer of the fusion neural network model to obtain a weighted feature map. The temporal evolution of the weighted feature map is captured by the bidirectional long short-term memory layer of the fused neural network model, and the biomass probability of each spatial unit is output. Then, a green tide situation diagnostic map is generated based on the biomass probability.

6. The real-time collaborative monitoring method for controlling green tides in nearshore waters as described in claim 1, characterized in that, Coupled growth simulations were performed on the fused dataset to obtain the critical growth depth and occurrence probability distribution of algae in nearshore waters, specifically including: Light intensity and nutrient concentration are parsed from the fused dataset and simultaneously input into a multi-factor coupled growth mechanism model. The multi-factor coupled growth mechanism model calculates the algal compensating light intensity based on the light intensity and nutrient concentration. The light attenuation coefficient of seawater in nearshore waters is obtained, and the critical growth depth is calculated based on the algal compensation light intensity and the nearshore waters. Obtain the algal suspension depth in nearshore waters, and determine the occurrence probability distribution based on the critical growth depth and the algal suspension depth.

7. A real-time collaborative monitoring method for controlling green tides in nearshore waters as described in claim 1, characterized in that, The generation of a nearshore green tide growth risk prediction map based on the critical growth depth and the occurrence probability distribution specifically includes: The gradient region of the risk of green tide in nearshore waters is obtained by mapping the probability distribution of occurrence. Using the critical growth depth as the guiding field, diffusion simulation is performed on the gradient region to obtain the risk propagation field; By performing similarity correction on the risk propagation field, a growth risk prediction map of green tide in nearshore waters is obtained.

8. A real-time collaborative monitoring method for controlling green tides in nearshore waters as described in claim 1, characterized in that, Based on the growth risk prediction map and the green tide situation diagnosis map, a dual-track early warning decision is made to obtain a comprehensive early warning signal and a dynamic risk field, specifically including: The situation warning coefficient is determined using the aforementioned green tide situation diagnostic map; The probability warning coefficient is determined using the growth risk prediction map. The situational warning coefficient and the probability warning coefficient are weighted and fused to obtain a comprehensive warning signal; Based on the comprehensive early warning signal and the topographic information of the nearshore sea area, a risk spatial diffusion simulation is performed to obtain a dynamic risk field.

9. A real-time collaborative monitoring method for controlling green tides in nearshore waters as described in claim 1, characterized in that, The dynamic risk field is used to perform collaborative path planning and task scheduling for green tide control resources in nearshore waters, generating collaborative scheduling instructions, specifically including: Obtain location and status information of green tide control resources in nearshore waters; Based on the dynamic risk field, multi-resource collaborative path planning is performed on the location information and the status information to obtain the collaborative scheduling instruction.

10. A real-time collaborative monitoring system for controlling green tides in nearshore waters, used to execute a real-time collaborative monitoring method for controlling green tides in nearshore waters as described in any one of claims 1 to 9, characterized in that, The real-time collaborative monitoring system includes: The fusion sensing module is used to collect multi-source monitoring data from nearshore waters, and to fuse the multi-source monitoring data to obtain a spatiotemporally synchronized fusion dataset. The situation diagnosis module is used to extract multi-dimensional features based on the fused dataset, construct a multi-dimensional collaborative perception vector, and then perform green tide intelligent diagnosis on the multi-dimensional collaborative perception vector to obtain a green tide situation diagnosis map. The evolution prediction module is used to perform coupled growth simulation on the fused dataset to obtain the critical growth depth and occurrence probability distribution of algae in nearshore waters, and generate a growth risk prediction map of green tide in nearshore waters based on the critical growth depth and occurrence probability distribution. The early warning decision module is used to make a dual-track early warning decision based on the growth risk prediction map and the green tide situation diagnosis map to obtain a comprehensive early warning signal and a dynamic risk field. The collaborative scheduling module is used to perform collaborative path planning and task scheduling for green tide prevention and control resources in nearshore waters through the dynamic risk field, and generate collaborative scheduling instructions.

Images