Near-shore micro-plastic convergence ridge dynamic prediction method, device, equipment and medium

Abstract

The invention discloses a near-shore microplastic convergence ridge dynamic prediction method, device and equipment and a medium, and relates to the technical field of data processing, and the method comprises the steps: determining a mass point library; reading the main flow direction of each particle in the particle library; the mass points serve as nodes to be assembled to obtain two layers of heterogeneous graphs, and the shear rate and the diffusion weight obtained through main flow direction calculation are written into edge attributes of the upper layer of heterogeneous graphs and the lower layer of heterogeneous graphs; respectively splitting the two layers of heterogeneous graphs into a directed convergence adjacency matrix and an undirected diffusion adjacency matrix, and further extracting node characterization; selecting ridge line seeds based on node representation; calculating to obtain an optimal ridge line according to the ridge line seeds; and constructing a driving vector for the optimal ridge line of the historical multiple frames, and generating a predicted ridge line sequence of a preset future duration according to the driving vector and the geometric coordinates of the ridge points in the optimal ridge line. According to the method, each mass point simultaneously has continuous power and material attributes at any time, and the prediction accuracy of micro-plastic convergence can be improved based on mass center analysis with accurate alignment.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a method, device, equipment and medium for dynamically predicting nearshore microplastic convergence ridges. Background Art

[0002] Existing monitoring systems for microplastics and floating pollutants generally use a regional tidal-wave coupled model to provide an Eulerian velocity field. Optimal interpolation or ensemble Kalman assimilation of velocity and water level is then performed using real-time observations from HF radar, GPS drifting rafts, and buoy profiles. Virtual particles are then released into a three-dimensional grid and forward integrated along the model step size to obtain the temporal Lagrangian distribution of concentration. Particle concentrations are often inverted using wide-band satellite imagery or airborne multispectral data for optical proxies such as suspended sediment and chlorophyll, then mapped to the particle's location, integrating the dynamic and material fields within a unified spatiotemporal coordinate system. To characterize nearshore convergence fronts, existing methods often calculate instantaneous velocity divergence, strain rate, or the Euler-Leuvelen index at Eulerian grid points. These methods then combine multiscale Canny or Sobel gradient operations to extract convergence zones, and then employ morphological connectivity analysis to identify elongated ridges. Some studies have employed modified Floyd-Warshall or terrain profile search algorithms, constraining ridge extension in both the tangential and normal directions to ensure geometric continuity. At the same time, the large-scale background field of coarse-resolution drift simulation and the local details of high-resolution remote sensing are spliced ​​through bilinear or spline interpolation, so as to take into account both computational efficiency and nearshore spatial accuracy.

[0003] Although existing methods can assimilate model flow fields and multi-source observations into a unified coordinate system, velocity interpolation and Lagrangian particle integration in the Eulerian grid often belong to different time steps. Phase misalignment is prone to occur during high-frequency tidal oscillations, thereby weakening the resolution of nearshore narrow fronts. Convergence zone extraction steps mostly directly perform threshold segmentation on the divergence or gradient and rely on morphological connectivity to maintain ridge coherence. This approach, which mainly relies on purely geometric features, is difficult to maintain the sharpness of the front for a long time under conditions of alternating tide peaks and wind jumps. The ridges often drift too wide or are smoothed by background diffusion. Graph-based deep models usually set the adjacency relationship as a static undirected weight. Their message transmission does not distinguish between mainstream directional pull and lateral diffusion. The shear intensity and concentration increment are homogenized within a single convolution. As a result, the directionality of the convergence axis and the randomness of material diffusion cannot be simultaneously presented. The recurrent network used for short-term extrapolation focuses on the statistical correlation of time series and lacks explicit characterization of astronomical tide harmonics and sudden wind and wave events. Uncertainty assessment relies on a large number of Monte Carlo perturbations. Although field encrypted sampling can reduce errors, the information feedback cycle is long and cannot converge to meter-level spatial accuracy in a timely manner. Summary of the Invention

[0004] The main purpose of the embodiments of the present application is to propose a method, device, equipment and medium for dynamic prediction of nearshore microplastic convergence ridges to improve the accuracy of prediction of microplastic convergence.

[0005] To achieve the above objectives, one aspect of the embodiments of the present application proposes a method for dynamically predicting nearshore microplastic convergence ridges, the method comprising the following steps: Determine a plurality of particles evenly distributed in the tidal accessible area; Predicting an initial trajectory for each of the particle points and interpolating new particle points based on the initial trajectory to obtain a particle library; Reading the instantaneous flow velocity of each particle in the particle library at the Euler grid point, and then obtaining the main flow direction of each particle; Assembling the particles as nodes to obtain two layers of heterogeneous graphs, and writing the shear rate and diffusion weight calculated according to the main flow direction into the edge attributes of the upper layer heterogeneous graph and the edge attributes of the lower layer heterogeneous graph respectively; Splitting the two layers of heterogeneous graphs into directed convergent adjacency matrices and undirected diffuse adjacency matrices respectively; Obtaining node representations extracted according to the directed convergent adjacency matrix and the undirected diffusion adjacency matrix; Calculating a shear stability score based on the node representation and selecting the node with the shear stability score greater than a preset score threshold as a ridge seed; Calculate the optimal ridge line according to the ridge line seed; A driving vector is constructed for the optimal ridge line of multiple historical frames, and a predicted ridge line sequence of a preset future duration is generated according to the driving vector and the geometric coordinates of the ridge point in the optimal ridge line.

[0006] In some embodiments, predicting an initial trajectory for each particle includes the following steps: Calling the three-dimensional flow field of the tidal numerical model to perform forward integration on each of the particles to obtain the initial trajectory; The method of interpolating the new mass points according to the initial trajectory to obtain a mass point library includes the following steps: Cubic spline interpolation is performed on the model flow velocity and water level on the particle time axis, and the shore-based high-frequency radar flow velocity, airborne imaging spectrum concentration and autonomous submersible profile concentration are simultaneously superimposed as observation mapping; The observations are mapped to corresponding particles by forward and backward tracking, spatial proximity and time distance weighting, and after eliminating wind speed, effective wave height and turbidity anomalies, third-order B-spline interpolation is used along the particle history direction to obtain the particle library of spatiotemporal continuous dynamics and concentration.

[0007] In some embodiments, assembling the particles as nodes to obtain two layers of heterogeneous graphs, and writing the shear rate and diffusion weight calculated according to the main flow direction into the edge attributes of the upper layer heterogeneous graph and the edge attributes of the lower layer heterogeneous graph, respectively, includes the following steps: calculating the shear intensity along the main flow direction for the Lagrangian reference grid of each particle to extract a shear gradient band as a shear rate; Perform time consistency check in two consecutive frames to determine the convergence core node; recursively writing directed convergent edges according to the main flow direction, calculating diffusion weights according to the transverse concentration gradient and screening undirected diffusion edges; The two layers of heterogeneous graphs are assembled under the condition that the node indexes are consistent, and the shear rate and the diffusion weight are written into the edge attributes of the upper layer heterogeneous graph and the edge attributes of the lower layer heterogeneous graph respectively.

[0008] In some embodiments, extracting node representations based on the directed convergent adjacency matrix and the undirected diffusion adjacency matrix comprises the following steps: Read the node velocity modulus, local shear rate and particle concentration, and then splice them into physical characteristics; Performing shear rate weighted directed convolution on the directed convergent adjacency matrix and performing diffusion weight masked lateral convolution on the undirected diffusion adjacency matrix, and then obtaining a gating coefficient according to the wind wave potential energy index and the local relative vorticity through Sigmoid mapping; The embedded features of the corresponding channels are extracted respectively; The embedded features of the two channels are adaptively fused, added to the residual of the physical features, and then layer-normalized to obtain the node representation.

[0009] In some embodiments, the step of calculating the optimal ridge line according to the ridge line seed comprises the following steps: Iteratively generate a single-scale main ridge according to the ridge seed along the main flow direction under the condition that the shear strength is not less than a preset shear threshold and does not touch the coastline, and then generate multi-scale ridges in parallel according to the normal equidistant offset and under the condition that the shear strength is not less than the preset shear threshold and does not touch the coastline; Calculating the segment average shear, relative vorticity and wave field energy of the single-scale main ridge and each of the multi-scale ridge lines according to fixed arc length segments, and forming segment attributes; After truncation according to the set attribute threshold, the comprehensive dynamic score of the remaining ridges is calculated and the segmented ridges with a score greater than the preset dynamic score threshold are retained; Based on the segmented ridge line, the geometrically continuous optimal ridge line is output through first-order flow regression and third-order B-spline overall smoothing.

[0010] In some embodiments, constructing a driving vector for the optimal ridge line of multiple historical frames comprises the following steps: The optimal ridge lines of multiple historical frames are sorted in ascending time order and resampled to a unified grid, thereby constructing the driving vector including the tidal phase, the shore direction or alongshore current velocity, the wind direction and wind speed pair, and the wave field energy; The step of generating a predicted ridge line sequence of a preset future duration according to the driving vector and the geometric coordinates of the ridge point in the optimal ridge line comprises the following steps: The driving vector and the geometric coordinates of the ridge point in the optimal ridge line are combined and input into a recursive network with a tidal harmonic memory gate, the main tide harmonic reference is explicitly introduced into the update gate and a migration gate is set for sudden changes in wind direction, and the ridge point displacement increment is output frame by frame to generate the predicted ridge line sequence of the preset future time length; Orthogonal perturbations are applied to the future wind field and tide amplitude and the extrapolation is repeated. The spatial confidence radius is calculated based on the sample standard deviation and the low-confidence ridge segments are marked.

[0011] In some embodiments, the method further comprises the following steps: Scanning the continuous ridge segments whose confidence radius exceeds a preset radius threshold in the predicted ridge line sequence; Retrieve the location and remaining endurance of the nearshore autonomous surface or underwater platform, select the execution platform based on a comprehensive cost function of distance, energy consumption and sea conditions, and plan the trajectory along the continuous ridge segment and the cross-section on both sides; issuing a sampling instruction to drive the execution platform to collect original flow velocity and concentration profiles along the track; Inserting the original flow velocity and the concentration profile into the corresponding mass point through a Lagrangian mapping interface and updating the mass point library; The step of reading the instantaneous flow velocity of each particle in the particle library at the Euler grid point to obtain the main flow direction of each particle is performed until the confidence radius of all ridge segments in the predicted ridge line sequence is lower than the preset radius threshold, and then outputting the final predicted ridge line sequence.

[0012] To achieve the above objectives, another aspect of the present application provides a device for dynamically predicting nearshore microplastic convergence ridges, the device comprising: A mass point determination unit is used to determine a plurality of mass points evenly distributed in the tidal accessible area; a particle library construction unit, configured to predict an initial trajectory for each particle and interpolate new particles according to the initial trajectory to obtain a particle library; a main flow direction reading unit, configured to read the instantaneous flow velocity of each particle in the particle library at an Euler grid point, and thereby obtain the main flow direction of each particle; a heterogeneous graph construction unit, configured to assemble the particles as nodes to obtain two layers of heterogeneous graphs, and write the shear rate and diffusion weight calculated according to the main flow direction into the edge attributes of the upper layer heterogeneous graph and the edge attributes of the lower layer heterogeneous graph respectively; A matrix splitting unit, used for splitting the two layers of heterogeneous graphs into directed convergent adjacency matrices and undirected diffusion adjacency matrices respectively; A node representation extraction unit, configured to extract node representations based on the directed convergent adjacency matrix and the undirected diffusion adjacency matrix; a ridge seed selection unit, configured to calculate a shear stability score based on the node representation and select the node having the shear stability score greater than a preset score threshold as a ridge seed; An optimal ridge line calculation unit, configured to calculate an optimal ridge line according to the ridge line seed; The ridge line prediction unit is used to construct a driving vector for the optimal ridge line of multiple historical frames, and generate a predicted ridge line sequence of a preset future time length according to the driving vector and the geometric coordinates of the ridge point in the optimal ridge line.

[0013] To achieve the above-mentioned purpose, another aspect of an embodiment of the present application provides an electronic device, which includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the above-mentioned method when executing the computer program.

[0014] To achieve the above-mentioned purpose, another aspect of an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and the computer program implements the above-mentioned method when executed by a processor.

[0015] The embodiments of the present application include at least the following beneficial effects: The present application provides a method, device, equipment and medium for dynamic prediction of nearshore microplastic convergence ridges. The present application scheme determines a plurality of particles evenly distributed in the area accessible by tides; predicts the initial trajectory of each particle and interpolates new particles based on the initial trajectory to obtain a particle library; reads the instantaneous flow velocity of each particle at the Euler grid point in the particle library, and then obtains the mainstream direction of each particle; assembles each particle as a node to obtain a two-layer heterogeneous graph, and writes the shear rate and diffusion weight calculated according to the mainstream direction into the edge attributes of the upper heterogeneous graph and the lower heterogeneous graph respectively. Edge attributes of heterogeneous graphs; splitting two layers of heterogeneous graphs into directed convergent adjacency matrices and undirected diffusion adjacency matrices respectively; obtaining node representations based on the directed convergent adjacency matrix and the undirected diffusion adjacency matrix; calculating the shear stability score based on the node representation and selecting nodes with shear stability scores greater than a preset score threshold as ridge seeds; calculating the optimal ridge based on the ridge seeds; constructing a driving vector for the optimal ridge of multiple historical frames, and generating a predicted ridge sequence of a preset future duration based on the driving vector and the geometric coordinates of the ridge point in the optimal ridge. This application scheme enables each particle to have continuous dynamic and material properties at any time, thereby enabling subsequent shear analysis and graph analysis to be carried out in a high-frequency tidal background without phase drift, thereby improving the alignment performance. The center of mass analysis based on accurate alignment can improve the prediction accuracy of microplastic aggregation. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0017] Figure 1 Comparison diagrams of some embodiments provided in the present application and the prior art; Figure 2 A schematic flow chart of a method for dynamically predicting nearshore microplastic convergence ridges provided in an embodiment of the present application; Figure 3 This is an example flow chart of a method for dynamically predicting nearshore microplastic convergence ridges based on a shear-diffusion dual flow graph neural network provided in an embodiment of the present application; Figure 4 A schematic diagram of the process for constructing a microplastic aggregation isomerization graph provided in an embodiment of the present application; Figure 5 A schematic diagram of a dual-channel flow-diffusion GNN encoding process provided in an embodiment of the present application; Figure 6 A schematic diagram of a flow chart of convergence ridge generation and optimization provided in an embodiment of the present application; Figure 7 A schematic diagram of the structure of a dynamic prediction device for nearshore microplastic convergence ridges provided in an embodiment of the present application; Figure 8 A schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. When the following description refers to the 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 the embodiments of the present application. They are merely examples of devices and methods consistent with some aspects of the embodiments of the present application as detailed in the appended claims.

[0019] It will be understood that the terms "first", "second", etc. used in this application may be used herein to describe various concepts, but unless otherwise specified, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another. For example, without departing from the scope of the embodiments of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the words "if" and "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0020] The terms "at least one", "plurality", "each", "any", etc. used in this application include "at least one", "two" or more, "plurality" or "each", "any" or "any one", "each" or "any one" as used herein.

[0021] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application pertains. The terms used herein are for the purpose of describing the embodiments of this application only and are not intended to limit this application.

[0022] Before describing the embodiments of the present application in detail, some of the related technologies involved in the embodiments of the present application are first described as follows: Dual-stream graph neural networks (GNNs) employ parallel message passing along two heterogeneous topologies: a convergent (shear) channel and a diffuse channel. The convergent layer captures the directional pull of the main flow, while the diffuse layer simulates the lateral spread. These two features are gated and fused to produce physically interpretable node embeddings.

[0023] Convergence ridges are elongated, narrow bands of microplastics at the ocean surface, formed along the axis of hydrodynamic convergence under the influence of tides, wind, and waves. These ridges, composed of a series of shear-stable, dynamically continuous, and geometrically connected nodes, are the core targets for localization and extrapolation.

[0024] Tidal harmonic memory gate: Introducing the main semi-diurnal / diurnal sinusoidal tidal frequency embedding term into the update gate of recurrent neural units (such as GRU) to enhance the memory ability of tidal periodic drift; and cooperating with the wind direction migration gate to quickly respond to sudden wind and wave events, realizing the dual-effect coupling of tidal and wind drive.

[0025] Uncertainty-driven sampling: Based on the confidence radius assessment of historical and extrapolated ridge lines, field platforms are automatically assigned to implement incremental profile sampling for low-confidence line segments that exceed the threshold; new observation data are assimilated through Lagrangian mapping, and predictions and confidence assessments are iteratively updated to form a closed-loop optimization.

[0026] At the data-driven level, convolutional neural networks have been used to segment floating debris strips from continuous optical images end-to-end. Graph-based approaches typically treat pattern grid nodes as vertices, establish undirected edges with fixed radius or K nearest neighbors, and then perform spectral convolution, GAT, or simplified random walks to extract joint flow field-concentration features. Most of these methods employ a single-layer homogeneous graph and represent adjacency relationships with static weights, focusing on reproducing the Eulerian gradient in the graph structure while not explicitly modeling the mechanistic differences between shear convergence and lateral diffusion.

[0027] In the short-term path prediction phase, recurrent neural networks, LSTMs, and even variational Seq2Seq structures are used to learn the nonlinear mapping between multi-source time series such as tide level, wind speed, and wave energy and ridge displacement. Uncertainty in future scenarios is typically accounted for by applying Monte-Carlo perturbations to the atmosphere and boundary conditions, forming multiple drift sets, and then estimating the spatial error ellipse using sample variance. For sea areas with large errors, existing operating platforms often utilize multi-objective planning based on ranging, energy consumption, and sea conditions to generate profiles or cross-sectional survey lines for AUVs or USVs. The value of incremental observations is evaluated using information entropy or the expected reduction in mean squared error, thus achieving a closed-loop prediction-sampling-re-prediction workflow.

[0028] In response to the above-mentioned shortcomings, this application aims to construct a physical-data coupled representation that can simultaneously preserve the shear front sharpness and lateral diffusion morphology, and complete the real-time alignment and fusion of multi-source observations in the high-frequency tidal-wind wave background, so as to generate short-term drift prediction results of microplastic convergence ridges with clear physical stratification and quantitative uncertainty, and further provide spatiotemporal guidance for adaptive encrypted sampling. Figure 1 A comparison diagram of some embodiments of the present application and the prior art.

[0029] The embodiments of the present application provide a method, device, equipment and medium for the dynamic prediction of nearshore microplastic convergence ridges, which relates to the field of data processing technology. The method, device, equipment and medium for the dynamic prediction of nearshore microplastic convergence ridges provided by the embodiments of the present application can be applied to a terminal, can also be applied to a server, and can also be software running in a terminal or a server. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch and a car terminal, etc., but is not limited to this; the server side can be configured as an independent physical server, or as a server cluster or distributed system composed of multiple physical servers, and can also be configured as a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, 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 of a method, device, equipment and medium for the dynamic prediction of nearshore microplastic convergence ridges, etc., but is not limited to the above forms.

[0030] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, and the like. The present 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, and the like that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments in which tasks are performed by remote processing devices connected via a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media, including storage devices.

[0031] Reference Figure 2 The present application embodiment provides a method for dynamically predicting nearshore microplastic convergence ridges. The method may include but is not limited to steps S200 to S280, as follows: S200: determining a plurality of particles evenly distributed in an area accessible by tide; S210: predicting an initial trajectory for each particle and interpolating new particles according to the initial trajectory to obtain a particle library; S220: Reading the instantaneous flow velocity of each particle in the particle library at the Euler grid point, thereby obtaining the main flow direction of each particle; S230: Assembling the particles as nodes to obtain two layers of heterogeneous graphs, and writing the shear rate and diffusion weight calculated according to the main flow direction into the edge attributes of the upper layer heterogeneous graph and the edge attributes of the lower layer heterogeneous graph respectively; S240: Splitting the two layers of heterogeneous graphs into directed convergent adjacency matrices and undirected diffuse adjacency matrices respectively; S250: Obtaining node representations extracted according to the directed convergent adjacency matrix and the undirected diffusion adjacency matrix; S260: Calculating a shear stability score based on the node representation and selecting the node with the shear stability score greater than a preset score threshold as a ridge seed; S270: Calculate and obtain an optimal ridge line according to the ridge line seed; S280: Constructing a driving vector for the optimal ridge line of multiple historical frames, and generating a predicted ridge line sequence of a preset future duration according to the driving vector and the geometric coordinates of the ridge point in the optimal ridge line.

[0032] Optionally, predicting the initial trajectory of each particle includes the following steps: Calling the three-dimensional flow field of the tidal numerical model to perform forward integration on each of the particles to obtain the initial trajectory; The method of interpolating the new mass points according to the initial trajectory to obtain a mass point library includes the following steps: Cubic spline interpolation is performed on the model flow velocity and water level on the particle time axis, and the shore-based high-frequency radar flow velocity, airborne imaging spectrum concentration and autonomous submersible profile concentration are simultaneously superimposed as observation mapping; The observations are mapped to corresponding particles by forward and backward tracking, spatial proximity and time distance weighting, and after eliminating wind speed, effective wave height and turbidity anomalies, third-order B-spline interpolation is used along the particle history direction to obtain the particle library of spatiotemporal continuous dynamics and concentration.

[0033] Optionally, assembling the particles as nodes to obtain two layers of heterogeneous graphs, and writing the shear rate and diffusion weight calculated according to the main flow direction into the edge attributes of the upper layer heterogeneous graph and the edge attributes of the lower layer heterogeneous graph, respectively, comprises the following steps: calculating the shear intensity along the main flow direction for the Lagrangian reference grid of each particle to extract a shear gradient band as a shear rate; Perform time consistency check in two consecutive frames to determine the convergence core node; recursively writing directed convergent edges according to the main flow direction, calculating diffusion weights according to the transverse concentration gradient and screening undirected diffusion edges; The two layers of heterogeneous graphs are assembled under the condition that the node indexes are consistent, and the shear rate and the diffusion weight are written into the edge attributes of the upper layer heterogeneous graph and the edge attributes of the lower layer heterogeneous graph respectively.

[0034] Optionally, the step of extracting node representations based on the directed convergent adjacency matrix and the undirected diffusion adjacency matrix comprises the following steps: Read the node velocity modulus, local shear rate and particle concentration, and then splice them into physical characteristics; Performing shear rate weighted directed convolution on the directed convergent adjacency matrix and performing diffusion weight masked lateral convolution on the undirected diffusion adjacency matrix, and then obtaining a gating coefficient according to the wind wave potential energy index and the local relative vorticity through Sigmoid mapping; The embedded features of the corresponding channels are extracted respectively; The embedded features of the two channels are adaptively fused, added to the residual of the physical features, and then layer-normalized to obtain the node representation.

[0035] Optionally, calculating the optimal ridge line according to the ridge line seed comprises the following steps: Iteratively generate a single-scale main ridge according to the ridge seed along the main flow direction under the condition that the shear strength is not less than a preset shear threshold and does not touch the coastline, and then generate multi-scale ridges in parallel according to the normal equidistant offset and under the condition that the shear strength is not less than the preset shear threshold and does not touch the coastline; Calculating the segment average shear, relative vorticity and wave field energy of the single-scale main ridge and each of the multi-scale ridge lines according to fixed arc length segments, and forming segment attributes; After truncation according to the set attribute threshold, the comprehensive dynamic score of the remaining ridges is calculated and the segmented ridges with a score greater than the preset dynamic score threshold are retained; Based on the segmented ridge line, the geometrically continuous optimal ridge line is output through first-order flow regression and third-order B-spline overall smoothing.

[0036] Optionally, constructing a driving vector from the optimal ridge line of multiple historical frames comprises the following steps: The optimal ridge lines of multiple historical frames are sorted in ascending time order and resampled to a unified grid, thereby constructing the driving vector including the tidal phase, the shore direction or alongshore current velocity, the wind direction and wind speed pair, and the wave field energy; The step of generating a predicted ridge line sequence of a preset future duration according to the driving vector and the geometric coordinates of the ridge point in the optimal ridge line comprises the following steps: The driving vector and the geometric coordinates of the ridge point in the optimal ridge line are combined and input into a recursive network with a tidal harmonic memory gate, the main tide harmonic reference is explicitly introduced into the update gate and a migration gate is set for sudden changes in wind direction, and the ridge point displacement increment is output frame by frame to generate the predicted ridge line sequence of the preset future time length; Orthogonal perturbations are applied to the future wind field and tide amplitude and the extrapolation is repeated. The spatial confidence radius is calculated based on the sample standard deviation and the low-confidence ridge segments are marked.

[0037] Optionally, the method further comprises the following steps: Scanning the continuous ridge segments whose confidence radius exceeds a preset radius threshold in the predicted ridge line sequence; Retrieve the location and remaining endurance of the nearshore autonomous surface or underwater platform, select the execution platform based on a comprehensive cost function of distance, energy consumption and sea conditions, and plan the trajectory along the continuous ridge segment and the cross-section on both sides; issuing a sampling instruction to drive the execution platform to collect original flow velocity and concentration profiles along the track; Inserting the original flow velocity and the concentration profile into the corresponding mass point through a Lagrangian mapping interface and updating the mass point library; The step of reading the instantaneous flow velocity of each particle in the particle library at the Euler grid point to obtain the main flow direction of each particle is performed until the confidence radius of all ridge segments in the predicted ridge line sequence is lower than the preset radius threshold, and then outputting the final predicted ridge line sequence.

[0038] Next, some optional embodiments of the present application will be introduced and explained in detail with reference to specific application examples.

[0039] Application scenarios of this embodiment may include: Ocean multi-source data fusion and Lagrangian tracking platform: This platform integrates numerical tidal models, HF radar, airborne telemetry, and AUV observation data, providing visual Lagrangian particle placement, three-dimensional flow field interpolation, and trajectory backtracking capabilities to help users quickly generate high-precision dynamic-matter space-time fields.

[0040] Dual-flow graph neural network prediction engine: Designed specifically for heterogeneous graphs (convergence layer + diffusion layer), it has embedded shear convolution, lateral diffusion convolution, and wind wave-vorticity gating modules, which can be efficiently trained and online inferenced on GPUs to achieve real-time positioning of microplastic convergence ridges and short-term drift prediction.

[0041] Uncertainty-driven field sampling scheduling system: Integrates platform parameters such as available energy, remaining flight time, and sea conditions and waves, automatically plans AUV / ASV / USV sampling tracks and profiles based on confidence radius, supports task dispatching, real-time monitoring, and data feedback, and forms an uncertainty closed-loop iterative sampling system.

[0042] Reference Figure 3 This embodiment provides a method for dynamically predicting nearshore microplastic convergence ridges based on a shear-diffusion dual flow graph neural network, comprising the following steps S1 to S6: S1. Under the constraints of the shoreline digital elevation model and the intertidal zone boundary, particles are evenly distributed in the tidal-accessible area. The three-dimensional flow field of the tidal numerical model is called to perform forward integration on each particle to obtain the initial trajectory. Cubic spline interpolation is performed on the model flow velocity and water level on the particle time axis. The shore-based high-frequency radar flow velocity, airborne imaging spectral concentration and autonomous submersible profile concentration are simultaneously superimposed. The observations are mapped to the corresponding particles through forward and backward tracking, spatial proximity and time distance weighting. After eliminating wind speed, effective wave height and turbidity anomalies, third-order B-spline interpolation is performed along the particle history direction to obtain a spatiotemporally continuous dynamic-concentration particle library.

[0043] S2. Read the instantaneous flow velocity of the particle library on the Euler grid point, calculate the shear intensity along the main flow direction for each grid point to extract the shear gradient band (shear rate), and then perform a time consistency test within two consecutive frames to determine the convergence core node; further recursively write directed convergence edges based on the main flow direction, calculate the diffusion weight based on the transverse concentration gradient and filter the undirected diffusion edges, assemble a two-layer heterogeneous graph under the condition of consistent node index and write the shear rate and diffusion weight into the upper and lower layer edge attributes respectively. For example, Figure 4 Schematic diagram of the process for constructing a microplastic aggregation isomerogram.

[0044] S3. Split the heterogeneous graph into a directed convergent adjacency matrix and an undirected diffuse adjacency matrix. Read the node velocity modulus, local shear rate, and particle concentration and splice them as physical features. Perform shear rate-weighted directed convolution on the convergent layer and diffusion weight masked lateral convolution on the diffuse layer. Then, obtain the gating coefficient based on the wind wave potential energy index and the local relative vorticity through Sigmoid mapping. Adaptively fuse the two channel embeddings and add them to the input embedding residual and layer normalization to obtain the final node representation. For example, Figure 5 Schematic diagram of the process of dual-channel flow-diffusion GNN encoding.

[0045] S4. Calculate the shear stability score based on node characterization and select high-scoring nodes as ridge seeds. Iterate and generate single-scale main ridges along the main flow direction under the condition that the shear strength is not lower than the threshold and does not touch the shoreline. Then, generate multi-scale ridges in parallel by normal arithmetic offset and applying the same iterative rule. Calculate the average shear, relative vorticity and wave field energy of each ridge segment according to a fixed arc length. After truncating according to the set threshold, calculate the comprehensive dynamic score of the remaining ridges and retain the high-scoring ridges. Finally, output the geometrically continuous optimal ridge line through first-order flow direction regression and third-order B-spline overall smoothing. For example, Figure 6 A schematic diagram of the process of generating and optimizing convergent ridges.

[0046] S5. The optimal ridge lines of multiple historical frames are sorted in ascending time order and resampled to a unified grid. A driving vector containing tidal phase, shore / onshore current velocity, wind direction-speed pair, and wave field energy is constructed. After merging with the ridge point geometric coordinates, it is input into a recursive network with a tidal harmonic memory gate. The main tide harmonic benchmark is explicitly introduced in the update gate and a migration gate is set for sudden changes in wind direction. The ridge point displacement increment is output frame by frame to generate a predicted ridge line sequence with a preset future duration; orthogonal perturbations are applied to the future wind field and tidal amplitude and repeated extrapolation is performed. The spatial confidence radius is calculated based on the sample standard deviation and low-confidence ridge segments are marked.

[0047] S6. Scan the continuous ridge segments whose confidence radius exceeds the threshold in the predicted ridge line sequence, retrieve the position and remaining endurance of the nearshore autonomous surface or underwater platform, select the execution platform based on the comprehensive cost function of distance, energy consumption and sea conditions, and plan the trajectory along the ridge and the cross-sections on both sides. After issuing the sampling command, receive the original flow velocity and concentration profile in real time; insert the returned observations into the corresponding particles through the Lagrangian mapping interface of step S1 and update the particle library, then re-execute steps S3 to S5 to form a short closed loop until the confidence radius of all ridge segments drops to the target threshold and output the final prediction results.

[0048] More specifically, the above steps may include the following steps: Step S1 includes the following steps: S11. Establish a Lagrangian reference frame based on the shoreline DEM and the intertidal zone boundary at the initial time Evenly distributed within the tidal area Mass ; Then call the three-dimensional flow field of the tidal numerical model and integrate each particle to the future time , get the set of initial trajectories of each particle .

[0049] S12, numerical tidal current synchronization, on the particle time axis When the velocity and water level of the tidal model are interpolated by cubic spline, the velocity and water level of each particle are interpolated by cubic spline. The Lagrangian numerical dynamics pair that is completely consistent with the measured time series can be obtained. The timing error is limited to the single-step integration error This ensures that there is no time-scale drift when the numerical background field is subsequently fused with the measured velocity.

[0050] S13, shore-based radar velocity Lagrangian mapping, calling the real-time Euler velocity of shore-based HF radar , track each grid point forward and backward, and The displacement path before and after Compare and select the particle index with the smallest spatial distance. The corresponding particle entries are written to achieve fine shear correction in the nearshore shallow water area; if the particle lacks a corresponding radar grid point, the numerical model flow velocity is retained.

[0051] S14, airborne telemetry concentration Lagrangian interpolation, airborne imaging spectrum at time Obtained pixel center coordinates and its concentration Compensate for countercurrent drift: Invert the pixel according to the tidal field. Displacement vector within , calculate the position after drift and capture radius of the particle Compare; if the spatial proximity condition is met, the concentration is written to the particle using weighted average. The square exponential decay relationship between the concentration interpolation weight and the spatial deviation is given by the following formula: ; in, : point mass At the moment of telemetry fusion concentration; : point mass The spatial coordinates of : No. The original coordinates of the center of the airborne pixels; :The pixel is in Tidal current anti-drift displacement between : The observed concentration value of the pixel; : The number of pixels that meet the capture radius conditions; : Capture radius parameter to control the spatial influence range.

[0052] S15. Arrange the AUV profile observations in ascending order of time Project to the nearest particle; in order to avoid excessive influence of early observation on the current scene, the time difference is used The exponentially decaying time weight is then multiplied by the spatial Gaussian weight to perform a weighted average of the concentration. Through this decreasing mechanism, the contribution of observations from distant time periods and distant spaces to the current concentration is automatically weakened.

[0053] S16, wind wave-turbidity anomaly elimination and interpolation, continuous reading of wind speed at the on-site weather station and buoy at the particle level , effective wave height and turbidity If it appears or or The dynamic and concentration records of the corresponding particles are marked as abnormal and removed. For the spatial voids generated by the removal, a third-order B-spline is applied to perform local interpolation along the direction of the particle's historical trajectory to ensure that the dynamic grid database Concentration Grid Library There are no gaps in time and space, and the integrated arrangement of ocean dynamics and concentration fields is completed.

[0054] Step S2 includes the following steps: S21, shear gradient band extraction, at each dynamic grid point First, read the main flow direction and construct the neighborhood along the flow direction In view of the objective characteristics of microplastic convergence ridges showing sharp speed mutations, the shear strength is calculated using the following formula: , and then Percentile thresholds set by scenario Sorting is performed to retain the grid point set with the steepest gradient and continuous strip distribution. The set narrows the search space and lays the geometric foundation for the subsequent locking of the convergence axis: ; in, : Grid Shear strength; : Grid The instantaneous flow velocity vector; and The unit vector along the main flow direction; and The displacement vector of :and A set of adjacent grid points with the same flow direction; : Percentile percentage of thin band extraction.

[0055] S22, the velocity mutation in the shear zone is often affected by short-period wind and wave noise, which is prone to instantaneous pseudo peaks. In order to improve the physical credibility of the core node, the consistency test is performed on the same shear zone grid in two consecutive frames of the dynamic field. Only when the grid point meets the requirements of both frames, the grid point is When Temporal consistency screening simultaneously considers particle flux conservation and convergence front stability to filter out transient misjudgments.

[0056] S23, downstream directed convergence edge, for each convergence core node , read its main flow direction unit vector And iteratively search for the first-order neighbor along this direction When the velocity of the adjacent point is The angle satisfies When the node Write directed convergent edges between The process then recursively advances until it is blocked by the shoreline or terminates in the low-speed buffer zone. The process outputs a set of chain structures that strictly follow the hydrodynamic direction, clearly defining the main channels through which microplastic particles are pulled within the narrow band.

[0057] S24, lateral diffusion weight calculation, switch to the concentration grid library, for each node and its lateral neighbors Perform concentration gradient detection. Considering the lateral spread effect caused by wind and wave disturbances, the following formula is used to couple and quantify the node concentration difference and lateral distance to obtain the diffusion weight: The weights not only reflect the increasing trend of concentration, but also embed the spatial attenuation, and perform a normalized assessment of the diffusion contribution at different particle size scales: ; in, : Node pair The diffusion weight of :node particle concentration; : lateral displacement vector between nodes; : spatial attenuation scale; : numerical stability constant; and The set of horizontally adjacent grid points; : Diffusion edge threshold.

[0058] S25, diffusion edge screening and write right, the previous step generated Perform full-field quantile normalization and select The node pairs establish undirected diffusion edges ,Will Directly write edge attributes to ensure that only diffusion paths that contribute significantly to lateral spread are retained, reducing the interference of low-intensity noise on subsequent GNN propagation.

[0059] S26, assemble two layers of heterogeneous topology, keep the node index completely consistent, and merge all convergent core nodes and their directed convergent edges into the upper layer graph , merge all diffusion nodes and their undirected diffusion edges into the lower graph Node sharing ensures that the two layers of information can be aligned in the same spatial reference system, while avoiding duplicate storage, forming a two-layer heterogeneous graph with a compact structure and clear physical meaning. .

[0060] S27, on the upper side Write shear rate As weight, on the lower edge Write diffusion weight . Then generate the node-level cross-layer mapping table The table provides a synchronous index for the dual-flow graph neural network, enabling a one-to-one correspondence between shear flow and diffusion flow at each node in the subsequent gated fusion phase. This completes the construction of a shear-dominated heterogeneous graph, providing a physically interpretable graph structure prior for accurate short-term prediction of the dynamic drift trajectory of microplastic convergence ridges.

[0061] Wherein, step S3 includes the following steps: S31, topology entry splitting, for heterogeneous graphs , resolve the upper layer directed convergent edge into an adjacency matrix ; At the same time, the lower layer undirected diffusion edge is parsed into an adjacency matrix By splitting the topological inlet, we can avoid initial confusion between the convergence chain and the diffusion network, while also providing clear edge direction constraints for the subsequent flow-related convolution kernel design. This ensures that during periods of high-frequency tidal oscillations, convergence information is not diluted by diffusion smoothing, preserving the geometric sharpness of the narrowband front.

[0062] S32, node physical feature injection, targeting any node in the sea area , read the instantaneous flow rate model , local shear rate and particle concentration , concatenated into the original feature vector . Perform linear mapping and add sinusoidal position encoding to convert to dimension The starting embedding .

[0063] The convergence-diffusion dual channel has both dynamic gradient and material gradient semantics when entering the network, which fits the tidal contraction + lateral diffusion coupling scenario.

[0064] S33, convergence flow message transmission, along Perform a directed convolution and apply a shear rate to the upstream node information Weighted aggregation, and below the threshold The weak gradient edge of is exponentially decayed. This keeps the front sharp during the tidal surge phase and prevents weak gradient noise from widening the convergence axis. Output embedding The focus is on describing the narrow-band aggregation trend of particles being pulled by the mainstream, which is critical for locating the main ridge of microplastics.

[0065] S34, diffusion flow message transmission, independent along Run lateral diffusion convolution, and the aggregation function takes the diffusion weight Soft masking is performed and random walk regularization is superimposed within the kernel function to simulate the random diffusion caused by micro-scale agitation of wind and waves. The influence of concentration increment in the lateral neighborhood is amplified to make the network more sensitive to the spread within the water mass and output embedding It mainly describes the lateral material diffusion morphology.

[0066] S35, wind wave potential energy – vorticity gate calculation, for nodes , firstly, the effective wave height of the buoy Calculating the Wind and Wave Potential Energy Index , and then calculate the curl of the local velocity field to obtain the relative vorticity The gating factor is obtained by linearly combining the two and applying Sigmoid mapping. , which is used to measure the wave shear strength at the current moment. , making the model more trustworthy in the shear channel; on the contrary, when the sea surface is stable and the vorticity is weakened, Lower, the model automatically tends to the diffusion channel: ; in, : gating coefficient; : Sigmoid mapping function; : Trainable weights and biases; : wind and wave potential energy index; : relative vorticity mode; : Convergent flow embedding; : diffusion flow embedding; : Fusion embedding.

[0067] S36. Based on S35, when the wind and wave potential energy is high and the vorticity is obvious, near , the convergent flow weight is amplified, and the model focuses on keeping the front sharp; when the sea state is calm, near , the diffusion flow weight increases, and the model focuses on the details of horizontal spread. Through adaptive strategies, fusion embedding It also contains two types of mechanism characteristics, reducing the ridge positioning error caused by single-stream bias.

[0068] S37, in order to avoid the numerical scale drift caused by the dual-stream fusion, With starting embedding Add the residuals and perform layer norm to get the final node representation The representation maintains strong physical interpretability and gradient stability, and can seamlessly represent the input into the multi-scale ridge growth process, thereby achieving short-term and accurate prediction of nearshore microplastic convergence ridges in a high-frequency tide-wind-wave coupled environment.

[0069] Wherein, step S4 comprises the following steps: S41. Fusion Embedding The weights of the shear and diffusion mechanisms have been integrated, but the actual location of microplastics accumulation still needs further screening. This step first calculates the shear stability score of all nodes. . Instantaneous shear weight at the node The main factor is used, and Gaussian decay is applied to the embedding difference between two adjacent frames to measure the continuity + strength dual standard. A high score indicates that the convergence is both sharp and continuous, which is consistent with the physical characteristics of the narrow convergence ridge source. After traversing the entire field, only The nodes are used as seeds to ensure that the starting point of subsequent ridge growth has a stable physical support and is not disturbed by short-lived wind and wave peaks: ; in, :node The current clipping weights come directly from the convergence channel of the two-stream GNN; 、 : The fusion embedding vector of the node in two consecutive frames is used to characterize temporal stability; : Controls the attenuation scale of the cross-frame difference influence range; : Seed threshold, set to full field Percentile.

[0070] S42, single-scale main ridge extension, using trusted seeds, expand them into a continuous ridge line. This step is done at each seed node Read the local main flow direction , then at a fixed step size To avoid the ridge line from entering the shelf shoal or shear attenuation zone, the iterative process must also meet the shear strength of not less than The two conditions of not touching the shoreline are combined with the logic and geometric constraints in the following formula. For each legal advance, the new node coordinates are recorded. The local flow direction is updated until any of the conditions fails. The main ridge generated in this way is naturally in the same direction as the hydrodynamic transport conveyor belt and can cover the maximum length of the main aggregation axis of microplastics.

[0071] Iteration conditions: ; in, :Iteration The coordinates of the spine nodes; : Single-step advancement distance along the flow direction, adaptively set based on grid resolution; :node The main flow direction unit vector is : Node shear strength, used to determine whether the front is still sharp; : Shear attenuation threshold, below which the front is considered to have disappeared; : Coastline discriminant function, value is When it hits the shore.

[0072] S43, multi-scale parallel ridge growth, microplastics often not only accumulate on a single axis, but also form parallel ridges of varying widths on both sides. To capture this multi-scale spatial hierarchy, this step takes the main ridge as the geometric center and Upper arithmetic increasing distance Copy the starting point and extend it using the same rules as S42. The copy offset relationship and scale parameter are expressed in the following formula. By scanning at multiple scales, from meters to hundreds of meters, both fine textures (facilitating estimation of the upper limit of aggregate concentration) and coarse-scale channels (determining overall drift velocity) can be identified simultaneously, providing a richer pool of candidates for subsequent kinetic optimization.

[0073] ; in, : Parallel ridges In step Node coordinates of : normal unit vector orthogonal to the main flow direction; : minimum normal deviation, usually equal to a single grid distance; : The scale increment step is set based on the experience of the scene diffusion width; : scale index, Represents the main spine itself.

[0074] S44, ridge segment dynamic attribute annotation, in order to select the most reliable main ridge among many candidate ridges, it is necessary to perform dynamic scoring on each local segment of each ridge. Segment and cut , relative vorticity and wave energy Calculate the average of each paragraph to form a paragraph attribute vector , as shown in the following formula. The vector maps the three major dynamic factors to the same scale, which can quantitatively reveal whether the shear and wave energy are sufficient to overcome the vorticity tearing, and is the core basis for subsequent optimization judgment: ; in, : Arc length The local shear strength at , reflects the strength of the convergence front; : Arc length The relative vorticity at , which measures the local rotational tearing tendency; : Arc length The wave field energy indicates the potential energy supply of wind and waves to maintain the front; : Arc length coordinates of the paragraph endpoint; : Segment arc length interval, usually 3–5 grid intervals.

[0075] S45, according to the previous step , scan from the ridge starting point to the downstream: if a certain segment Exceeding the stability threshold , indicating that the vortex tearing effect is strong and cuts off immediately; if the wave energy Below maintenance threshold , is also cut off. This prevents the ridge line from accidentally extending into the structural fragile area and maintains the physical stability of the predicted axis.

[0076] S46. For the ridge lines that remain after truncation, the shear strength and wave energy of all segments are reconciled with the vorticity through exponential decay, and the comprehensive score is calculated. ; According to the full field quantile threshold Only high score ridges are retained. By superimposing strong shear and high wave energy and applying an exponential penalty to high vorticity, true ridges and false ridges can be effectively distinguished, thereby excluding error sources from subsequent extrapolation.

[0077] S47, ridge geometry smoothing and output: Finally, a two-step geometry optimization is performed to preserve the ridges: Local hydrodynamic regression: Use first-order linear fitting to fine-tune the position of the point along the flow direction to remove jagged edges caused by integer grids; Third-order B-spline smoothing: reconstruct the entire ridge curve, suppress short-wave pseudo-shapes but maintain the overall direction. It satisfies the triple requirements of geometric continuity, shear sharpness, and dynamic stability at the same time, and provides the initial ridge line with the smallest error and the highest credibility when performing short-term drift extrapolation.

[0078] Wherein, step S5 comprises the following steps: S51, historical ridge series arrangement, in the data window Collect high confidence convergence ridge centerline per frame , sorted in ascending order by observation time and resampled to unity The minimum grid ensures that adjacent frames meet tidal position coherence. The alignment operation provides the recursive model with a complete and gap-free ridge evolution context, avoiding memory gate desynchronization due to unequal time intervals.

[0079] S52, driving vector construction, for any ridge point in the sequence , add the tide phase at the same time , shoreward current velocity , alongshore current velocity , wind direction – wind speed and wave field energy , spliced ​​into a fixed-dimensional driving vector And concatenate with the geometric coordinates to form the input vector The driving information is propagated along the nodes, enabling the recursive unit to adaptively switch the dynamic weights between tidal surge and wind jump.

[0080] S53, tidal harmonic memory gate recursive propagation, Feed into the recurrent unit with tidal harmonic memory gate: the basic gating structure follows GRU, but the main tidal harmonic reference is explicitly introduced in the update gate and ; At the same time, set an additional migration gate for wind direction mutation events , the gated update relationship is as follows: ; in, : current step input vector; 、 : The current and previous hidden states; : Update gate and reset gate; : tidal harmonic embedding matrix; : Astronomical semi-diurnal and heliosine tide frequencies; : Wind direction difference between the current frame and the previous frame; : Migration gate, sensitive to wind direction jumps; : wind mutation migration vector; 、 : Sigmoid and hyperbolic tangent activation; : Hadamard element-by-element multiplication.

[0081] Through the tidal harmonic term, the recursive unit can circulate memory along the tidal phase in the stable period; when When the sudden rise Open, hidden state fast injection migration vector , capturing sudden drift acceleration.

[0082] S54, frame-by-frame displacement extrapolation, from the latest observation frame Start by setting the step length Iteratively run the recursive unit and output the ridge point displacement increment ; Then superimpose it on the previous frame coordinates to generate the future ( The displacement increments are directly derived from the hidden state map, ensuring that tidal periodicity and wind-induced nonlinear drifts are synchronized to the geometry level.

[0083] S55. To quantify the uncertainty of extrapolation, the future wind farm and tidal amplitude Imposition Group orthogonal perturbation, repeat S54 to obtain Group ridge samples. Calculate the sample standard deviation for the same point coordinates And mapped to the spatial confidence radius : ; in, : No. The coordinates of the sample ridge points after the perturbation; : The mean coordinates of the ridge point samples; : collection size; :correspond %Confidence level of the coefficient.

[0084] S56. If there is a node on a certain ridge line that satisfies , the segment is marked as low confidence and the trigger threshold is recorded , write metadata corresponding to the main controlling factor of disturbance (wind or tide) to facilitate rapid positioning during incremental sampling by the on-site mobile platform.

[0085] S57, integrate the predicted ridge sequence and node confidence radius label to generate the time series ridge package ; Write the forecast database according to the time index and publish it to the navigation, maritime monitoring and assimilation modules in real time through the message bus to realize the dynamic push and sharing of drift paths and potential risk areas.

[0086] Wherein, step S6 comprises the following steps: S61. Scan the ridge package with confidence radius output in the previous period point by point, and calculate the average uncertainty of each continuous over-threshold line segment according to the following formula: , and extract its center coordinates Corresponding time window : ; in, : No. Continuous transthreshold segments at time The average confidence radius of : The number of ridge points contained in the line segment; : The index set of the line segment ridge points; : Ridge point The single point confidence radius of .

[0087] when When , the line segment is recorded as low confidence and enters the encrypted candidate pool to ensure that subsequent tasks only focus on the real error source.

[0088] S62, for each low-confidence line segment, real-time retrieval of the location set of nearshore AUV, ASV and USV platforms and remaining battery life ; Calculate the arrival time from the platform to the center of the line segment ,like and , then the platform Considered as an executable unit, preferred The smallest one locks the execution carrier. The screening is immediately combined with the tidal backflow window to avoid the high time consumption of platform backflow.

[0089] S63, adopt a comprehensive cost function based on distance, energy consumption and sea conditions Planned track: ; in, :platform Execution Segment The overall cost of the task; : The surface distance from the platform to the center of the line segment; : platform cruising speed; : The energy required to complete the planned profile; : remaining available energy of the platform; : effective wave height of segment sea condition; : The upper limit of platform safety wave height; : Adjustable weight.

[0090] By minimizing , automatically obtain the survey lines, sampling depth and frequency along the ridge and the cross profiles on both sides, form an incremental observation list, and ensure maximum information gain sampling of uncertain areas under limited energy consumption conditions.

[0091] S64, send the observation list together with the wave height restriction threshold to the selected platform control terminal; the platform automatically collects the concentration and velocity profile according to the planned track, and returns the original data packet through high-frequency narrowband communication. During the execution of the mission, the sea condition monitoring module monitors the sea condition in real time. and Perform threshold comparison. If the limit is exceeded, adjust the measurement line or terminate immediately to ensure platform safety.

[0092] S65. After receiving the returned data, call the Lagrangian coordinate mapping interface of step S1 to insert the newly collected concentration profile into the corresponding particle node, replacing the original low-confidence observation; synchronously update the concentration grid library and the driving vector set to provide the latest status for rapid assimilation.

[0093] S66, trigger short loop mode: re-execute S3 dual stream encoding, S4 ridge generation and S5 time sequence extrapolation, and output a new confidence radius assessment. If there are no line segments, the system returns to S62 to re-screen the platform to form a closed-loop iteration; if all line segments meet the requirements, the system enters the termination logic.

[0094] S67, when the confidence radius of the entire ridge segment is , the system automatically sends a termination command and records the remaining flight time of each platform and remaining energy , archived to the resource scheduling database, providing a priori information for the next round of encryption. This completes a small-batch, uncertainty-driven on-site encryption, achieving continuous reduction of uncertainty in the location of microplastic convergence ridges with minimal field work costs.

[0095] This implementation first introduces a full Lagrangian coordinate system in the data preprocessing layer, using this as the core to achieve strict spatiotemporal alignment of multi-source observations. Constrained by the shoreline DEM and intertidal zone boundaries, this implementation evenly distributes particles within the tidal reach. A three-dimensional tidal numerical model is then used for forward integration, and high-order interpolation of model velocity and water level is achieved using cubic splines on the particle's own time axis. Within this unified reference system, Eulerian velocities from shore-based HF radar are mapped to particles through forward and backward tracking. Concentration information is then weighted based on spatial proximity and temporal distance indices using airborne imaging spectral pixels and autonomous submersible profiling observations. If wind speed, significant wave height, or turbidity trigger an anomaly threshold, the corresponding particle is recorded as missing and a third-order B-spline local interpolation is performed along its historical trajectory. This process ensures that each particle possesses continuous dynamic and material properties at all times, enabling subsequent shear analysis and graph convolution to be performed in a high-frequency tidal context without phase drift. This precise alignment capability is unattainable using the traditional Euler interpolation-particle integration separation framework.

[0096] In the feature modeling and representation learning layer, this embodiment constructs a two-layer heterogeneous graph consisting of an upper-layer directed convergence chain and a lower-layer undirected diffusion network. The convergence layer recursively generates chain edges based on the mainstream gradient and incorporates the shear rate into the edge weights. The diffusion layer quantizes the edge weights using lateral concentration differences and spatial attenuation, preserving the main channels for lateral diffusion. Node indices are kept consistent across both layers, enabling parallel convolution of shear and diffusion information followed by node-level gating fusion. Within the graph neural network, shear-rate-weighted directed convolution is used in the convergence channel to enhance the particle pulling characteristics along the mainstream direction. In the diffusion channel, a soft masking of diffusion weights and a superposition of random walk regularization are applied to capture the randomness of lateral diffusion induced by wind and waves. In the fusion stage, dynamic gating coefficients are generated using a sigmoid mapping of the wind and wave potential energy index and the local relative vorticity. The contributions of the two channels are adaptively adjusted, and the final output is normalized by a residual connection with the input features. The above-mentioned layered mapping, dual-channel convolution and physical quantity-driven gating mechanism avoids the problem of homogenizing shearing and diffusion in existing static undirected graphs, while ensuring that the sharpness of the convergence is not diluted by lateral smoothing during high-frequency tides.

[0097] At the time series deduction and field closed-loop layer, this embodiment uses a harmonic memory gate GRU explicitly embedded with astronomical semidiurnal tide and solar tidal frequencies to recursively propagate historical ridgeline sequences and driving vectors, outputting ridge point displacement increments in real time. When encountering sudden changes in wind direction, an additional migration gate rapidly injects wind jump information into the hidden state, thereby simultaneously describing both periodic and transient dynamics. Future scenario uncertainty is accounted for by applying orthogonal perturbations to the wind field and tidal amplitude to obtain a ridgeline set. The sample standard deviation is mapped to a node confidence radius and used to locate low-confidence ridge segments requiring encryption. Subsequent task scheduling uses a comprehensive cost function based on distance, energy consumption, and sea conditions to plan trajectories along ridges and normal cross-sections for AUVs, ASVs, or USVs. Newly acquired profile data is instantly written back to the particle database via a Lagrangian mapping interface, triggering a short re-encoding and re-forecast cycle, forming an adaptive closed-loop with minute-level updates. This multi-layered collaborative process significantly reduces error convergence time using the same observation resources and ensures that predictions remain consistent with the latest tidal and wind wave conditions.

[0098] Compared to existing solutions that primarily rely on Eulerian grid assimilation and single-layer static graph convolution, the full Lagrangian alignment mechanism proposed in this embodiment enables the model flow field and multi-source observations to obtain a unified time reference at each particle, fundamentally eliminating the phase mismatch problem during the high-frequency tidal oscillation phase. This allows for the resolution of the convergence front sharpness to be maintained at the sub-grid scale. A two-layer heterogeneous graph constructed based on this unified reference frame coexists with a shear-driven directed convergence chain and an undirected diffusion network driven by lateral concentration differences. Through node index sharing and physical quantity weighting, shear pull and diffusion propagation can be separated and modeled within the network before being gated and fused, avoiding the frontal blunting caused by traditional single-layer undirected graph homogenization.

[0099] In the time series deduction phase, a memory gate GRU embedded with astronomical tide harmonics, combined with a wind direction migration gate, can simultaneously maintain periodic tide memory and sensitive response to sudden wind and wave events, providing a more stable drift vector output compared to conventional LSTM or purely data-driven approaches. The on-site closed loop guided by the confidence radius further maps the forecast uncertainty directly into the platform sampling trajectory planning conditions, enabling minute-by-minute data re-injection and model re-forecasting. This adaptive encryption strategy for error sources overcomes the time constraints of traditional batch processing models, which require waiting for full tide cycles for updates. Overall, this embodiment has differentiated advantages in structural aspects such as spatiotemporal alignment accuracy, shear-diffusion mechanism decoupling, and real-time closed-loop updates. It can continuously output physically consistent convergent ridge drift results with quantified uncertainty under complex nearshore sea conditions.

[0100] An optional specific embodiment: From July 2023 to June 2024, the L Province Marine Ecological and Environmental Monitoring Center conducted a special mission to dynamically track and urgently address microplastic convergence ridges in the nearshore waters of F Bay. Over the past two years, this area has been impacted by the migration of discharge outlets from upstream plastic processing parks and monsoon shifts, resulting in repeated incidents of floating microplastics landing at fishing ports. Traditional Euler grid forecasts based on offshore buoys and single-station HF radar often have errors of hundreds of meters in frontal position, failing to meet the precise requirements of the emergency salvage window. This mission was fully supported by the technology platform presented in this invention, from multi-source data access to adaptive encrypted sampling and closed-loop operation, aiming to reduce the convergence ridge positioning error to less than 50 meters.

[0101] During the data collection phase, the technical team deployed a dual-station WERA HF radar array spanning 30 km, outputting surface current velocities with a 15-minute resolution. Three AUV profiling lines were deployed simultaneously, providing daily feedback on current velocities and turbidity at depths of 0–20 m. Fixed-wing drones equipped with 400–1000 nm hyperspectral cameras flew along the coastal zone during early morning low-cloud periods, acquiring 18 surface images with a resolution of 1.2 m. The numerical background employed the CMA 2022 version of the tide-wave coupled model, with a grid resolution of 150 m and a time step of 300 s. After coordinate unification and time-scale correction, all data were input into the Lagrangian preprocessing module of the proposed method. This ultimately released approximately 450,000 particles within the tidal-accessible area, creating a particle library with 48-hour continuous dynamic-concentration properties.

[0102] A shear-diffusion dual-layer heterogeneous graph was constructed online with an update frequency of 5 minutes. The upper layer, directed convergent chains, were recursively grown using the main directional gradient, with an average chain length of 1.8 km. The lower layer, an undirected diffusion network, selected edges based on lateral concentration differences, with an edge density controlled at 6.2%, ensuring GPU memory usage was less than 32 GB. The dual-flow graph neural network was pre-trained for 18 rounds on historical observations from 2021 to 2023 before being put into online inference. Each inference cycle took 42 seconds and continuously produced a sequence of convergent ridge extrapolations with a 1-hour interval and a maximum length of 48 hours, with a 95% confidence radius for each node.

[0103] In the first month after deployment, the system identified 57 high-confidence major ridges, nine of which were marked as low-confidence sections. Based on the confidence radius, the platform automatically planned 14 along-ridge and cross-profile tracks for two USVs. The USVs traveled 137 km and transmitted 112 profiles. This data was mapped back to the particle database via Starlink in real time before returning to port, facilitating five short-cycle re-forecasts. Ultimately, the average error between the measured and predicted distance of the convergent ridge centerline was 34 meters.

[0104] To verify performance, the monitoring center simultaneously ran two control schemes: a static Euler mesh with a single-layer GCN approach, and an LCS analysis method based on a flow field divergence threshold. Table 1 shows a comparative analysis of the three methods during 12 spring tides from April to June 2024.

[0105] Table 1

[0106] From the convergent ridge positioning evaluation results of the method of this embodiment and the control scheme in Table 1, it can be seen that this embodiment is significantly better than the two control schemes in both positioning accuracy and ridge line detection rate, while maintaining real-time level in inference time, providing timely and reliable decision-making basis for emergency control during rapid tidal reversal.

[0107] Reference Figure 7 The present application also provides a device for dynamically predicting nearshore microplastic convergence ridges, which can implement the above-mentioned method for dynamically predicting nearshore microplastic convergence ridges. The device includes: A mass point determination unit is used to determine a plurality of mass points evenly distributed in the tidal accessible area; a particle library construction unit, configured to predict an initial trajectory for each particle and interpolate new particles according to the initial trajectory to obtain a particle library; a main flow direction reading unit, configured to read the instantaneous flow velocity of each particle in the particle library at an Euler grid point, and thereby obtain the main flow direction of each particle; a heterogeneous graph construction unit, configured to assemble the particles as nodes to obtain two layers of heterogeneous graphs, and write the shear rate and diffusion weight calculated according to the main flow direction into the edge attributes of the upper layer heterogeneous graph and the edge attributes of the lower layer heterogeneous graph respectively; A matrix splitting unit, used for splitting the two layers of heterogeneous graphs into directed convergent adjacency matrices and undirected diffusion adjacency matrices respectively; A node representation extraction unit, configured to extract node representations based on the directed convergent adjacency matrix and the undirected diffusion adjacency matrix; a ridge seed selection unit, configured to calculate a shear stability score based on the node representation and select the node having the shear stability score greater than a preset score threshold as a ridge seed; An optimal ridge line calculation unit, configured to calculate an optimal ridge line according to the ridge line seed; The ridge line prediction unit is used to construct a driving vector for the optimal ridge line of multiple historical frames, and generate a predicted ridge line sequence of a preset future time length according to the driving vector and the geometric coordinates of the ridge point in the optimal ridge line.

[0108] It can be understood that the contents of the above method embodiments are all applicable to the present device embodiments, the functions specifically 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.

[0109] The present application also provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method of the present application. The electronic device can be any smart terminal, such as a tablet computer or an in-vehicle computer.

[0110] It can be understood that the contents of the above method embodiments are all applicable to the embodiments of the present device, the functions specifically implemented by the embodiments of the present device are the same as those of the method of the present application, and the beneficial effects achieved are also the same as those achieved by the method of the present application.

[0111] See also Figure 8 , Figure 8 The hardware structure of an electronic device according to another embodiment is shown. The electronic device includes: The processor 801 can be implemented as a general-purpose CPU (Central Processing Unit), a microprocessor, an 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 the present application. The memory 802 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 802 can store an operating system and other application programs. 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 802 and is called by the processor 801 to execute the methods of the embodiments of this application. Input / output interface 803, used to implement information input and output; Communication interface 804, used to implement communication interaction between this device and other devices, which can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WiFi, Bluetooth, etc.); Bus 805 , which transmits information between various components of the device (e.g., processor 801 , memory 802 , input / output interface 803 , and communication interface 804 ); The processor 801 , the memory 802 , the input / output interface 803 and the communication interface 804 are connected to each other in communication within the device via a bus 805 .

[0112] An embodiment of the present application further provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the method of the present application is implemented.

[0113] It can be understood that the contents of the above method embodiments are all applicable to the present storage medium embodiment, the functions specifically implemented by the present storage medium embodiment 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.

[0114] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely arranged relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0115] The embodiments described in the embodiments of this application are intended to more clearly illustrate 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. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0116] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0117] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, i.e., they may be located in one place or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of this embodiment.

[0118] Those skilled in the art will appreciate that all or some of the steps in the methods, systems, and functional modules / units in the devices disclosed above may be implemented as software, firmware, hardware, or appropriate combinations thereof.

[0119] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0120] It should be understood that in this application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following items" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0121] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the above-mentioned units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0122] The units described above as separate components may or may not be physically separate, and 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 these units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0123] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0124] If the integrated unit is implemented in the form of 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 the present application, 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, which is stored in a storage medium and includes multiple instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: various media that can store programs, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0125] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but are not intended to limit the scope of the present invention. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and essence of the present invention should be within the scope of the present invention.

Claims

1. A method for dynamic prediction of nearshore microplastic convergence ridges, characterized in that: The method comprises the following steps: Determine a plurality of particles evenly distributed in the tidal accessible area; Predicting an initial trajectory for each of the particle points and interpolating new particle points based on the initial trajectory to obtain a particle library; Reading the instantaneous flow velocity of each particle in the particle library at the Euler grid point, and then obtaining the main flow direction of each particle; Assembling the particles as nodes to obtain two layers of heterogeneous graphs, and writing the shear rate and diffusion weight calculated according to the main flow direction into the edge attributes of the upper layer heterogeneous graph and the edge attributes of the lower layer heterogeneous graph respectively; Splitting the two layers of heterogeneous graphs into directed convergent adjacency matrices and undirected diffuse adjacency matrices respectively; Obtaining node representations extracted according to the directed convergent adjacency matrix and the undirected diffusion adjacency matrix; Calculating a shear stability score based on the node representation and selecting the node with the shear stability score greater than a preset score threshold as a ridge seed; Calculate the optimal ridge line according to the ridge line seed; A driving vector is constructed for the optimal ridge line of multiple historical frames, and a predicted ridge line sequence of a preset future duration is generated according to the driving vector and the geometric coordinates of the ridge point in the optimal ridge line.

2. The method for dynamic prediction of nearshore microplastic convergence ridges according to claim 1, characterized in that: The step of predicting the initial trajectory of each particle comprises the following steps: Calling the three-dimensional flow field of the tidal numerical model to perform forward integration on each of the particles to obtain the initial trajectory; The method of interpolating the new mass points according to the initial trajectory to obtain a mass point library includes the following steps: Cubic spline interpolation is performed on the model flow velocity and water level on the particle time axis, and the shore-based high-frequency radar flow velocity, airborne imaging spectrum concentration and autonomous submersible profile concentration are simultaneously superimposed as observation mapping; The observations are mapped to corresponding particles by forward and backward tracking, spatial proximity and time distance weighting, and after eliminating wind speed, effective wave height and turbidity anomalies, third-order B-spline interpolation is used along the particle history direction to obtain the particle library of spatiotemporal continuous dynamics and concentration.

3. The method for dynamic prediction of nearshore microplastic convergence ridges according to claim 1, characterized in that: The method of assembling the particles as nodes to obtain two layers of heterogeneous graphs, and writing the shear rate and diffusion weight calculated according to the main flow direction into the edge attributes of the upper heterogeneous graph and the lower heterogeneous graph, respectively, includes the following steps: calculating the shear intensity along the main flow direction for the Lagrangian reference grid of each particle to extract a shear gradient band as a shear rate; Perform time consistency check in two consecutive frames to determine the convergence core node; recursively writing directed convergent edges according to the main flow direction, calculating diffusion weights according to the transverse concentration gradient and screening undirected diffusion edges; The two layers of heterogeneous graphs are assembled under the condition that the node indexes are consistent, and the shear rate and the diffusion weight are written into the edge attributes of the upper layer heterogeneous graph and the edge attributes of the lower layer heterogeneous graph respectively.

4. The method for dynamic prediction of nearshore microplastic convergence ridges according to claim 1, characterized in that: The step of extracting node representations based on the directed convergent adjacency matrix and the undirected diffusion adjacency matrix comprises the following steps: Read the node velocity modulus, local shear rate and particle concentration, and then splice them into physical characteristics; Performing shear rate weighted directed convolution on the directed convergent adjacency matrix and performing diffusion weight masked lateral convolution on the undirected diffusion adjacency matrix, and then obtaining a gating coefficient according to the wind wave potential energy index and the local relative vorticity through Sigmoid mapping; The embedded features of the corresponding channels are extracted respectively; The embedded features of the two channels are adaptively fused, added to the residual of the physical features, and then layer-normalized to obtain the node representation.

5. The method for dynamic prediction of nearshore microplastic convergence ridges according to claim 1, characterized in that: The step of calculating the optimal ridge line according to the ridge line seed comprises the following steps: Iteratively generate a single-scale main ridge according to the ridge seed along the main flow direction under the condition that the shear strength is not less than a preset shear threshold and does not touch the coastline, and then generate multi-scale ridges in parallel according to the normal equidistant offset and under the condition that the shear strength is not less than the preset shear threshold and does not touch the coastline; Calculating the segment average shear, relative vorticity and wave field energy of the single-scale main ridge and each of the multi-scale ridge lines according to fixed arc length segments, and forming segment attributes; After truncation according to the set attribute threshold, the comprehensive dynamic score of the remaining ridges is calculated and the segmented ridges with a score greater than the preset dynamic score threshold are retained; Based on the segmented ridge line, the geometrically continuous optimal ridge line is output through first-order flow regression and third-order B-spline overall smoothing.

6. The method for dynamic prediction of nearshore microplastic convergence ridges according to claim 1, characterized in that: The method of constructing a driving vector for the optimal ridge line of multiple historical frames includes the following steps: The optimal ridge lines of multiple historical frames are sorted in ascending time order and resampled to a unified grid, thereby constructing the driving vector including the tidal phase, the shore direction or alongshore current velocity, the wind direction and wind speed pair, and the wave field energy; The step of generating a predicted ridge line sequence of a preset future duration according to the driving vector and the geometric coordinates of the ridge point in the optimal ridge line comprises the following steps: The driving vector and the geometric coordinates of the ridge point in the optimal ridge line are combined and input into a recursive network with a tidal harmonic memory gate, the main tide harmonic reference is explicitly introduced into the update gate and a migration gate is set for sudden changes in wind direction, and the ridge point displacement increment is output frame by frame to generate the predicted ridge line sequence of the preset future time length; Orthogonal perturbations are applied to the future wind field and tide amplitude and the extrapolation is repeated. The spatial confidence radius is calculated based on the sample standard deviation and the low-confidence ridge segments are marked.

7. A method for dynamic prediction of nearshore microplastic convergence ridges according to any one of claims 1 to 6, characterized in that: The method further comprises the following steps: Scanning the continuous ridge segments whose confidence radius exceeds a preset radius threshold in the predicted ridge line sequence; Retrieve the location and remaining endurance of the nearshore autonomous surface or underwater platform, select the execution platform based on a comprehensive cost function of distance, energy consumption and sea conditions, and plan the trajectory along the continuous ridge segment and the cross-section on both sides; issuing a sampling instruction to drive the execution platform to collect original flow velocity and concentration profiles along the track; Inserting the original flow velocity and the concentration profile into the corresponding mass point through a Lagrangian mapping interface and updating the mass point library; The step of reading the instantaneous flow velocity of each particle in the particle library at the Euler grid point to obtain the main flow direction of each particle is performed until the confidence radius of all ridge segments in the predicted ridge line sequence is lower than the preset radius threshold, and then outputting the final predicted ridge line sequence.

8. A dynamic prediction device for nearshore microplastics convergence ridges, characterized in that: The device comprises: A mass point determination unit is used to determine a plurality of mass points evenly distributed in the tidal accessible area; a particle library construction unit, configured to predict an initial trajectory for each particle and interpolate new particles according to the initial trajectory to obtain a particle library; a main flow direction reading unit, configured to read the instantaneous flow velocity of each particle in the particle library at an Euler grid point, and thereby obtain the main flow direction of each particle; a heterogeneous graph construction unit, configured to assemble the particles as nodes to obtain two layers of heterogeneous graphs, and write the shear rate and diffusion weight calculated according to the main flow direction into the edge attributes of the upper layer heterogeneous graph and the edge attributes of the lower layer heterogeneous graph respectively; A matrix splitting unit, used for splitting the two layers of heterogeneous graphs into directed convergent adjacency matrices and undirected diffusion adjacency matrices respectively; A node representation extraction unit, configured to extract node representations based on the directed convergent adjacency matrix and the undirected diffusion adjacency matrix; a ridge seed selection unit, configured to calculate a shear stability score based on the node representation and select the node having the shear stability score greater than a preset score threshold as a ridge seed; An optimal ridge line calculation unit, configured to calculate an optimal ridge line according to the ridge line seed; The ridge line prediction unit is used to construct a driving vector for the optimal ridge line of multiple historical frames, and generate a predicted ridge line sequence of a preset future time length based on the driving vector and the geometric coordinates of the ridge point in the optimal ridge line.

9. An electronic device, characterized in that: The electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images