Ocean hydrological dynamic monitoring method and system based on cloud edge collaboration

Abstract

The application discloses a marine hydrological dynamic monitoring method and system based on cloud-edge collaboration, relates to the technical field of marine environment monitoring, and comprises the following steps: acquiring temperature-salinity flow time series data, constructing a three-dimensional temperature-salinity matrix, and extracting a space-time gradient vector; encapsulating and broadcasting the space-time gradient vector together with wave energy storage surplus to form a collaborative detection cluster; analyzing a message to divide an abnormal water mass area, extracting a vortex boundary topology point set according to a storage surplus distribution node, and transmitting the vortex boundary topology point set to a main network; fusing an ocean current vector field to deduce a discrete drift coordinate sequence, constructing a triangular net to sweep and generate a time series space flow form envelope of a hydrological anomaly; performing intersection testing on the envelope and a global coordinate library, extracting time evolution depth parameters corresponding to intersection nodes, and issuing an asynchronous wake-up time slot table and a variable-frequency sampling instruction to the detection nodes according to the time evolution depth parameters. The system is used for implementing the above method.

Description

Technical Field

[0001] This invention relates to the field of marine environmental monitoring technology, specifically to a method and system for dynamic monitoring of marine hydrology based on cloud-edge collaboration. Background Technology

[0002] Marine hydrological dynamic monitoring plays a crucial role in marine ecological protection, marine resource development, and disaster early warning. With the continuous expansion of marine observation networks, underwater sensor nodes deployed in various sea areas continuously collect time-series parameters such as temperature, salinity, and current velocity, generating massive amounts of multi-source hydrological environmental data. To achieve high-precision dynamic monitoring of the vast ocean, it is necessary to aggregate and analyze this bottom-level hydrological data in real time, thereby identifying the evolutionary trajectories of mesoscale eddies or anomalous water masses in the ocean.

[0003] However, in actual monitoring, the acoustic communication bandwidth in the underwater environment is extremely narrow, and the communication channel resources between surface nodes and satellites are also severely limited. Existing methods mostly employ purely centralized processing or static edge processing modes. Centralized processing continuously uploads massive amounts of multi-dimensional, low-level hydrological raw data to the shore-based control center, easily causing communication link congestion and significantly consuming the limited power of the ocean-going sensor nodes, leading to rapid battery depletion. While static edge processing reduces data transmission to some extent, each sensor node can only independently process local data from its surroundings. When anomalous water masses move spatially, statically deployed detection nodes cannot dynamically adjust their computing power allocation and sampling frequency according to hydrological evolution trends, resulting in redundant data in hydrologically stable areas and the loss of crucial physical boundary contours in areas of drastic environmental change, leading to poor accuracy and continuity in tracking the evolution trajectory of large-scale anomalous water masses. Therefore, it is still necessary to provide a cloud-edge collaborative dynamic marine hydrological monitoring method to improve the efficiency of computing resource allocation and the consistency of hydrological anomaly tracking during marine environmental monitoring. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a method and system for dynamic monitoring of marine hydrology based on cloud-edge collaboration to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a cloud-edge collaborative dynamic monitoring method for marine hydrology, comprising the following steps: S1, collecting time-series data sequences of temperature, salinity, and current from a preset water depth profile and constructing a three-dimensional temperature-salinity matrix; performing differential operations on the three-dimensional temperature-salinity matrix to extract the temperature-salinity spatiotemporal gradient vector; encapsulating the temperature-salinity spatiotemporal gradient vector and wave energy storage margin into a node status message; and triggering the formation of a self-organizing collaborative detection cluster via broadcast through an underwater acoustic channel; S2, parsing the node status message; performing clustering and segmentation based on the temperature-salinity spatiotemporal gradient vector to divide the core and edge regions of the variability water mass; and allocating detection nodes within the core region of the variability water mass based on the wave energy storage margin to perform contour tracing operations, extracting the vortex boundary topological point set and... S3. Transmit the data unidirectionally to the upper-level main network. The vortex boundary topology point set is fused with the pre-set ocean current vector field. Based on the vortex conservation condition, continuous displacement parameters are iteratively calculated to deduce the discrete drift coordinate sequence. The discrete drift coordinate sequence is extracted as geometric control points to construct a three-dimensional irregular triangular network. A temporal spatial manifold envelope is generated by sweeping along the time axis to enclose the hydrological anomaly evolution trajectory. S4. A spatial bounding box intersection test is performed on the temporal spatial manifold envelope and the global node three-dimensional coordinate library. The coordinate set of intersection nodes located inside the temporal spatial manifold envelope is selected. The temporal evolution depth parameters corresponding to the intersection node coordinate set are extracted. Based on the temporal evolution depth parameters, an asynchronous wake-up time slot table and frequency conversion sampling commands are compiled and sent to the corresponding detection nodes.

[0006] In a preferred embodiment, the specific process of acquiring time-series temperature, salinity, and current data sequences from a preset water depth profile and constructing a three-dimensional temperature, salinity, and current matrix, and performing differential operations on the three-dimensional temperature, salinity, and current matrix to extract the temperature, salinity, and current spatiotemporal gradient vector is as follows: Activating a distributed sensing array at a fixed underwater depth to acquire synchronous temperature scales, salinity, and current velocity parameters; combining the synchronous temperature scales, salinity, and current velocity parameters with corresponding depth coordinate metadata to generate a time-series temperature, salinity, and current data sequence; mapping the time-series temperature, salinity, and current data sequence along the depth axis and time axis to a preset spatial grid, filling the preset spatial grid to generate a three-dimensional temperature, salinity, and current matrix; applying discrete differential operators along the coordinate axes of the three-dimensional temperature, salinity, and current matrix to obtain spatial gradient components and temporal gradient components, and concatenating the spatial gradient components and temporal gradient components to synthesize a temperature, salinity, and current spatiotemporal gradient vector.

[0007] In a preferred embodiment, the specific process of encapsulating the temperature-salinity spatiotemporal gradient vector and wave energy storage margin into a node status message, and triggering the formation of a self-organizing cooperative detection cluster via underwater acoustic channel broadcast is as follows: read the real-time voltage monitoring data output by the hardware power management module, convert the real-time voltage monitoring data to generate wave energy storage margin; splice the temperature-salinity spatiotemporal gradient vector, wave energy storage margin, and detection node physical identifier to form a payload data frame, add a frame header check bit to encapsulate the payload data frame into a node status message; transmit the node status message to a preset underwater acoustic communication frequency band, receive the handshake confirmation frame returned by the adjacent detection node, bind the adjacent detection node that returned the handshake confirmation frame to establish a local communication link, and aggregate along the local communication link to form a self-organizing cooperative detection cluster.

[0008] In a preferred embodiment, the specific process of parsing node status messages and performing clustering segmentation based on temperature-salinity spatiotemporal gradient vectors to divide the core and edge regions of the mutated water mass is as follows: The temperature-salinity spatiotemporal gradient vectors are extracted from the node status messages within the self-organizing cooperative detection cluster through reverse unpacking. These vectors are then mapped to a feature vector space, and a high-density gradient clustering center is iteratively searched within the feature vector space. The characteristic Euclidean distance from each temperature-salinity spatiotemporal gradient vector to the high-density gradient clustering center is calculated, and the characteristic Euclidean distance is numerically compared with a preset spatial classification boundary value. If the comparison determines that the characteristic Euclidean distance is less than the preset spatial classification boundary value, the corresponding detection node is confirmed to have near-range aggregation characteristics and is classified into the core region of the mutated water mass. If the comparison determines that the characteristic Euclidean distance is greater than or equal to the preset spatial classification boundary value, the corresponding detection node is confirmed to have far-range attenuation characteristics and is classified into the edge region.

[0009] In a preferred embodiment, the specific process of allocating detection nodes within the core area of ​​the mutated water mass based on the wave energy storage capacity to perform contour tracing calculations, extracting the vortex boundary topology point set, and unidirectionally transmitting it to the upper-level main network is as follows: Within the core area of ​​the mutated water mass, a sorted sequence of wave energy storage capacity is selected, and the detection node corresponding to the highest wave energy storage capacity is designated as the master control tracking node; an edge computing scheduling command is issued to the master control tracking node, triggering the master control tracking node to perform point-by-point approximation calculations along the contour lines connecting the temperature-salinity-temporal gradient vectors within the core area of ​​the mutated water mass, locking the closed contour to generate the vortex boundary topology point set; the uplink communication bandwidth resources of the master control tracking node are allocated, the vortex boundary topology point set is modulated into an uplink carrier signal, and the uplink carrier signal is transmitted unidirectionally to the main network receiving terminal to transmit the vortex boundary topology point set.

[0010] In a preferred embodiment, the specific process of fusing the vortex boundary topological point set with a preset ocean current vector field and iteratively calculating continuous displacement parameters based on the vorticity conservation condition to deduce the discrete drift coordinate sequence is as follows: extract the velocity and direction attributes of the preset ocean current vector field, map the vortex boundary topological point set to the preset ocean current vector field to generate an initial boundary point set with ocean current forcing terms; substitute the initial boundary point set into the vorticity conservation condition, perform successive integration operations along the fluid streamline direction of the preset ocean current vector field to extract continuous displacement parameters; use the continuous displacement parameters to perform vector translation updates on the spatial coordinates of the initial boundary point set along the time axis to output a discrete drift coordinate sequence over multiple time spans.

[0011] In a preferred embodiment, the specific process of extracting discrete drift coordinate sequences as geometric control points to construct a three-dimensional irregular triangular mesh, and sweeping along the time axis to generate a temporal spatial manifold envelope that encapsulates the evolution trajectory of hydrological anomalies, is as follows: extracting spatial nodes within the discrete drift coordinate sequence as geometric control points, performing Delaunay triangulation on the geometric control points in a three-dimensional Cartesian coordinate system, and connecting the geometric control points to construct a three-dimensional irregular triangular mesh; configuring the time axis stretching step of the three-dimensional irregular triangular mesh, driving the three-dimensional irregular triangular mesh to perform a lofting and sweeping operation along the time axis direction, and recording the outer edge contour surface of the three-dimensional irregular triangular mesh during the sweeping process; stitching the outer edge contour surface to generate a closed three-dimensional geometric surface entity, and confirming the three-dimensional geometric surface entity as the temporal spatial manifold envelope.

[0012] In a preferred embodiment, the specific process of performing a spatial bounding box intersection test on the envelope of the temporal spatial manifold and the global node 3D coordinate library to filter out the intersection node coordinate set located inside the envelope of the temporal spatial manifold is as follows: extract the 3D extreme boundary of the envelope of the temporal spatial manifold, and construct an orthogonal axial bounding box based on the 3D extreme boundary; read the node physical coordinates in the global node 3D coordinate library, perform a spatial ray intersection test on the node physical coordinates and the orthogonal axial bounding box, and determine the spatial intersection state of the node physical coordinates and the orthogonal axial bounding box; capture the node physical coordinates whose spatial intersection state is internal intersection, and aggregate the internal intersection node physical coordinates to generate the intersection node coordinate set.

[0013] In a preferred embodiment, the specific process of extracting the time evolution depth parameters corresponding to the intersection node coordinate set, and compiling the asynchronous wake-up time slot table and frequency conversion sampling command based on the time evolution depth parameters and sending them to the corresponding probe nodes is as follows: The intersection node coordinate set is back-projected onto the time axis profile of the temporal space manifold envelope, and the time evolution depth parameters corresponding to the intersection node coordinate set on the time axis profile are extracted; the time evolution depth parameters are mapped to a preset hardware scheduling matrix, and the target wake-up delay period and target sampling frequency are extracted from the preset hardware scheduling matrix; the target wake-up delay period and target sampling frequency are merged to compile the asynchronous wake-up time slot table and frequency conversion sampling command; the hardware network identifier bound to the intersection node coordinate set is parsed, a downlink control channel is established based on the hardware network identifier, and the asynchronous wake-up time slot table and frequency conversion sampling command are sent to the corresponding probe nodes through the downlink control channel.

[0014] The cloud-edge collaborative marine hydrological dynamic monitoring system is used to execute the aforementioned cloud-edge collaborative marine hydrological dynamic monitoring method. It includes: a sensing network module, used to collect time-series data sequences of temperature, salinity, and current from a preset water depth profile and construct a three-dimensional temperature-salinity matrix; performing differential operations on the three-dimensional temperature-salinity matrix to extract the temperature-salinity spatiotemporal gradient vector; encapsulating the temperature-salinity spatiotemporal gradient vector and wave energy storage margin into a node status message; and broadcasting this message via an underwater acoustic channel to trigger the formation of a self-organizing collaborative detection cluster. A boundary extraction module is used to parse the node status message, perform clustering and segmentation based on the temperature-salinity spatiotemporal gradient vector to divide the core and edge regions of the anomalous water mass; and allocate detection nodes within the core region of the anomalous water mass based on the wave energy storage margin to perform contour tracing operations, extracting the vortex boundary topological point set and... The system transmits data unidirectionally to the upper-level main network. The spatiotemporal simulation module integrates the vortex boundary topology point set with a pre-set ocean current vector field, iteratively calculates continuous displacement parameters based on vorticity conservation conditions to deduce discrete drift coordinate sequences, extracts these sequences as geometric control points to construct a three-dimensional irregular triangular network, and sweeps along the time axis to generate a temporal spatial manifold envelope that encapsulates the hydrological anomaly evolution trajectory. The closed-loop scheduling module performs spatial bounding box intersection tests on the temporal spatial manifold envelope and the global node three-dimensional coordinate library, filters out the intersection node coordinate set located within the temporal spatial manifold envelope, extracts the temporal evolution depth parameters corresponding to the intersection node coordinate set, and compiles an asynchronous wake-up time slot table and frequency conversion sampling commands based on the temporal evolution depth parameters, sending them to the corresponding detection nodes.

[0015] The technical effects and advantages of this invention are as follows: (1) The cloud-edge collaborative dynamic monitoring method for marine hydrology extracts a detection cluster by performing differential operations on the three-dimensional temperature-salinity matrix during the edge collaborative sensing and topology extraction stage. The vector is then encapsulated and broadcast with the remaining wave energy storage capacity. Subsequently, clustering is performed based on the temperature-salinity spatial gradient vector to divide the core area into the edge area. Based on the remaining wave energy storage capacity, detection nodes are allocated in the core area to extract the topological point set of eddy boundary by contour tracing. This overcomes the communication congestion and rapid energy depletion of nodes caused by the direct transmission of massive amounts of raw data in the existing centralized processing mode. Feature dimensionality reduction is completed at the edge, and edge computing power is dynamically allocated based on the actual sea state gradient and the remaining power of physical nodes. This significantly reduces the uplink communication pressure on the main network while ensuring the accurate extraction of local abnormal boundaries.

[0016] (2) The cloud-edge collaborative marine hydrological dynamic monitoring system integrates the vortex boundary topology point set transmitted from the edge side with the ocean current vector field during the cloud-based dynamics deduction and reverse scheduling stage. Based on the vortex conservation condition, it deduces the temporal spatial manifold envelope of the hydrological anomaly evolution trajectory. Then, it performs spatial intersection tests on the global node three-dimensional coordinate library, compiles the asynchronous wake-up time slot table and frequency conversion sampling command based on the time evolution depth parameters corresponding to the intersection node coordinates, and sends them to the detection nodes. As a result, it can overcome the observation bottleneck of limited field of view of local edge nodes, use the hydrodynamic deduction results in the cloud to guide the working status of downstream marine physical detection nodes, avoid the equipment from maintaining ineffective high-frequency sampling in stable waters, and thus improve the continuity of the system in tracking abnormal hydrological events and the overall endurance of the global monitoring network.

[0017] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0018] Figure 1 This is a flowchart of the cloud-edge collaborative marine hydrological dynamic monitoring method of the present invention; Figure 2 This is a three-dimensional scatter diagram of the feature space segmented by clustering of detection nodes in an embodiment of the present invention; Figure 3 This is a flowchart of the cloud-edge collaborative marine hydrological dynamic monitoring system of the present invention. Detailed Implementation

[0019] This application's embodiments address the problems in existing marine hydrological monitoring, such as communication channel congestion caused by uploading all raw data from the bottom layer, uneven distribution of computing power among edge devices, and difficulty in continuously tracking the evolution path of large-scale hydrological anomalies through a cloud-edge collaborative marine hydrological dynamic monitoring method and system.

[0020] Example 1; please refer to Figure 1This invention provides a technical solution: a cloud-edge collaborative method for dynamic monitoring of marine hydrology, comprising the following steps: S1, collecting time-series data sequences of temperature, salinity, and current from a preset water depth profile and constructing a three-dimensional temperature-salinity matrix; performing differential operations on the three-dimensional temperature-salinity matrix to extract the temperature-salinity spatiotemporal gradient vector; encapsulating the temperature-salinity spatiotemporal gradient vector and wave energy storage margin into a node status message; and triggering the formation of a self-organizing collaborative detection cluster via broadcast through an underwater acoustic channel; S2, parsing the node status message; performing clustering and segmentation based on the temperature-salinity spatiotemporal gradient vector to divide the core and edge regions of the variability water mass; and allocating detection nodes within the core region of the variability water mass based on the wave energy storage margin to perform contour tracing operations, extracting the vortex boundary topological point set and upward... The main network transmits data unidirectionally; S3, the vortex boundary topology point set is fused with the pre-set ocean current vector field, and the continuous displacement parameters are iteratively calculated based on the vortex conservation condition to deduce the discrete drift coordinate sequence. The discrete drift coordinate sequence is extracted as geometric control points to construct a three-dimensional irregular triangular network, and a temporal spatial manifold envelope that wraps the hydrological anomaly evolution trajectory is generated by sweeping along the time axis; S4, the spatial bounding box intersection test is performed on the temporal spatial manifold envelope and the global node three-dimensional coordinate library, the coordinate set of the intersection nodes located inside the temporal spatial manifold envelope is selected, the time evolution depth parameters corresponding to the intersection node coordinate set are extracted, and the asynchronous wake-up time slot table and frequency conversion sampling command are compiled and sent to the corresponding detection nodes based on the time evolution depth parameters.

[0021] In this implementation scheme, in step S1, the underwater detection node acquires data on the continuous changes of temperature, salinity, and current velocity at a specified depth over time using its built-in sensors. This data is then arranged and combined according to both depth and time dimensions to form a three-dimensional temperature-salinity matrix, which is the fundamental data structure reflecting the state of the ocean's vertical water column. Locally, the node uses differential operations to compare the numerical changes at different times and depths within the matrix, extracting a spatiotemporal gradient vector representing the intensity and direction of hydrological changes. Subsequently, the node reads its remaining electrical energy stored through its wave power generation device, i.e., its wave energy storage reserve. The node packages the environmental feature vector and its own electrical energy value into a status message, which is broadcast among neighboring nodes using an underwater acoustic channel. Upon receiving the message, surrounding nodes establish local communication connections based on the status information, forming a temporary, self-organized, cooperative detection cluster.

[0022] In step S2, after the nodes within the self-organizing collaborative detection cluster parse each other's status messages, they classify the water areas according to their attributes based on the similarity of the temperature-salinity spatiotemporal gradient vectors. Water areas with drastic and concentrated gradient changes are designated as the core region of the anomalous water mass, while transitional water areas with gentle gradient changes are designated as the edge region. Within the core region, the system extracts the wave energy storage capacity of each node and selects the physical node with the most abundant power to undertake the high-consumption computational tasks. The detection nodes assigned tasks perform point-by-point calculations along the lines connecting equal temperature-salinity changes, i.e., contour tracing calculations, to find the closed physical contour surrounding the anomalous water mass and generate a set of vortex boundary topological points. Finally, the edge nodes only send the set of coordinate points representing the boundary shape unidirectionally to the cloud platform via the uplink network.

[0023] In step S3, after receiving the topological point set reported by the edge side, the cloud platform inputs it into a pre-set ocean current vector field containing historical data on global ocean current direction and velocity. Following the vorticity conservation condition that maintains equilibrium of rotational intensity during fluid motion, the cloud platform progressively calculates the relatively independent displacement parameters of each boundary point under the forcing action of the ocean current, thereby obtaining the specific location coordinates of the anomalous water mass at different future time periods, forming a discrete drift coordinate sequence. The cloud platform extracts these discrete coordinate points and connects them in three-dimensional space, constructing a three-dimensional irregular triangular network composed of multiple irregular triangles. Following the temporal progression, the cloud platform performs solid stitching on the outer edge of the trajectory generated by the movement of this triangular network in space, generating a geometry that completely envelops the movement path of the anomalous water mass in both spatial and temporal dimensions—the temporal spatial manifold envelope.

[0024] In step S4, the cloud platform retrieves a global node 3D coordinate library containing the physical locations of all sensor deployments. It then performs a spatial geometric intersection and collision comparison between the static coordinates of these nodes and the generated temporal spatial manifold envelope—a spatial bounding box intersection test. Through this test, the cloud platform accurately identifies the coordinates of nodes whose locations fall within the envelope and aggregates them to generate a set of intersection node coordinates. The cloud platform calculates the time difference required for the envelope's leading edge to reach these intersection nodes and extracts the temporal evolution depth parameter. Based on this parameter, the cloud platform creates an asynchronous wake-up time slot table to control the sleep and startup schedules of the target nodes and generates corresponding frequency-adjusting sampling commands to adjust the data acquisition frequency. These commands are then precisely transmitted via the downlink channel to the detection nodes located on the future evolution path of the anomalous water mass.

[0025] Specifically, the process of collecting time-series temperature, salinity, and current data sequences from a preset water depth profile and constructing a three-dimensional temperature, salinity, and current matrix, and then performing differential operations on the three-dimensional temperature, salinity, and current matrix to extract the temperature, salinity, and current spatiotemporal gradient vector is as follows: Activating a distributed sensing array at a fixed underwater depth to acquire synchronous temperature scales, salinity, and current velocity parameters; combining these parameters with corresponding depth coordinate metadata to generate a time-series temperature, salinity, and current data sequence; mapping the time-series data sequence along the depth axis and time axis to a preset spatial grid; filling the preset spatial grid to generate a three-dimensional temperature, salinity, and current matrix; applying discrete differential operators along the coordinate axes of the three-dimensional temperature, salinity, and current matrix to obtain spatial gradient components and temporal gradient components; and concatenating the spatial gradient components and temporal gradient components to synthesize a temperature, salinity, and current spatiotemporal gradient vector.

[0026] In this implementation scheme, the system first synchronously triggers a conductivity meter and an acoustic Doppler device mounted at a fixed underwater depth via a low-level hardware control bus to read the raw analog electrical signals collected by each physical probe and convert them into actual physical parameters. A three-dimensional temperature-salinity matrix is ​​constructed by arranging the temperature scale, salinity, and current velocity parameters along the depth and time axes and filling a preset spatial grid. The specific calculation logic is as follows: ;in, This represents the vertical depth discrete index of the preset spatial grid; Discrete index representing the sampling period along the time axis; Indicates the first The depth profile and the first The actual temperature scale parameters measured at each sampling period; This represents the salinity parameter in the corresponding coordinate system; This represents the flow velocity parameter in the corresponding coordinate system; This represents the matrix transpose operation. After the matrix is ​​constructed, the microprocessor unit calls the arithmetic logic module to extract the spatial gradient components along the depth axis using the central difference algorithm. The temporal gradient component is extracted along the time axis using a forward difference algorithm. Here, we take temperature scale parameters as an example, and the calculation process is as follows: ; ;in, This parameter represents the vertical physical spacing between adjacent probes when deploying a physical sensor array. This parameter represents the time interval between two consecutive data acquisitions controlling the crystal oscillator output. After calculating the physical rates of change of temperature and salinity in each dimension, the system performs a one-dimensional flattening and stitching operation on the spatial and temporal gradient components according to the preset feature dimension direction, directly synthesizing the corresponding temperature-salinity spatiotemporal gradient vector in the memory register. .

[0027] Specifically, the process of encapsulating the temperature-salinity spatiotemporal gradient vector and wave energy storage margin into a node status message, and triggering the formation of a self-organizing cooperative detection cluster via underwater acoustic channel broadcast is as follows: Read the real-time voltage monitoring data output by the hardware power management module, convert the real-time voltage monitoring data to generate wave energy storage margin; splice the temperature-salinity spatiotemporal gradient vector, wave energy storage margin, and detection node physical identifier to form a payload data frame, add a frame header check bit to encapsulate the payload data frame into a node status message; transmit the node status message to the preset underwater acoustic communication frequency band, receive the handshake confirmation frame returned by the adjacent detection node, bind the adjacent detection node that returned the handshake confirmation frame to establish a local communication link, and aggregate along the local communication link to form a self-organizing cooperative detection cluster.

[0028] In this implementation scheme, the underlying main control chip continuously polls the digital-to-analog converter interface inside the hardware power management module to obtain the real-time potential difference signal across the supercapacitor energy storage group, and substitutes the real-time potential difference signal into the energy conversion equation to generate wave energy storage margin. The specific transformation equation is as follows: ;in, This indicates the equivalent rated capacitance of the wave power generation energy storage capacitor component. This represents the current instantaneous potential difference read by the power management module; This refers to the physical cutoff potential difference that maintains the minimum communication wake-up state between the main control chip and the underwater acoustic emission module. This indicates the system's power output conversion efficiency of the entire wave energy harvesting board. Based on this, the main control chip will store the temperature-salinity spatiotemporal gradient vector in the memory register from the previous steps. Wave energy storage capacity and the node physical address burned into the read-only memory The system performs bit-by-bit splicing to generate payload data frames, and appends cyclic redundancy check bits generated by polynomial division to the end of the data frames, thus encapsulating them into node status messages conforming to the underwater communication protocol. The underlying main control chip drives the underwater acoustic transducer to transmit node status messages to a specific frequency channel. When the underwater acoustic receiver captures a handshake confirmation frame from a neighboring detection node, the system triggers a link bonding mechanism to calculate signal quality and set a link admission threshold. The method for determining the link admission threshold is as follows: ;in, This indicates the received physical strength of the underwater acoustic signal in the handshake confirmation frame extracted after analog-to-digital conversion. This represents the average measured amplitude of the underwater acoustic background noise floor. This represents the signal gain coefficient pre-calibrated by the system based on the current hydrological conditions of the sea area; This represents the preset environmental acoustic attenuation penalty coefficient at the hardware level. The system compares the actual communication link quality score with the link admission threshold. Numerical comparison calculations are performed, and only when the actual calculated score is greater than or equal to the link admission threshold, the physical addresses corresponding to the adjacent nodes are written into the local routing table to establish a stable local communication link, thereby aggregating at the physical layer to form a self-organizing cooperative detection cluster.

[0029] Specifically, the process of parsing node status messages and performing clustering segmentation based on temperature-salinity spatiotemporal gradient vectors to divide the core and edge regions of the mutated water mass is as follows: The temperature-salinity spatiotemporal gradient vectors are extracted from the node status messages within the self-organizing cooperative detection cluster through reverse unpacking. These vectors are then mapped to a feature vector space, and high-density gradient cluster centers are iteratively searched within this space. The characteristic Euclidean distance from each temperature-salinity spatiotemporal gradient vector to the high-density gradient cluster center is calculated, and this distance is compared numerically with a preset spatial classification boundary value. If the comparison determines that the characteristic Euclidean distance is less than the preset spatial classification boundary value, the corresponding detection node is confirmed to have near-range aggregation characteristics and is classified into the core region of the mutated water mass. If the comparison determines that the characteristic Euclidean distance is greater than or equal to the preset spatial classification boundary value, the corresponding detection node is confirmed to have far-range attenuation characteristics and is classified into the edge region.

[0030] In this implementation scheme, the underlying processing unit decomposes the received status messages from each node and extracts the temperature-salinity spatiotemporal gradient vector of the corresponding node location. The system then projects all vectors within its communication cluster into a multidimensional feature vector space. Within this feature vector space, the system calculates the local feature density at each node's location. The solution is obtained through iterative equations as follows: ;in, and Both represent the physical indexes of the probe nodes within a self-organizing collaborative probe cluster; Indicates the first The spatiotemporal gradient vector of temperature and salinity extracted from each detection node; Indicates the first The temperature-salinity spatiotemporal gradient vectors of the adjacent nodes; This represents the smoothing scale parameter of the search kernel in the feature vector space. The system traverses the local feature cluster density of all nodes, extracts the coordinates of nodes with global maxima, and updates their feature vectors with high-density gradient cluster centers. Subsequently, the system calculates the characteristic Euclidean distance from each node to the cluster center. : ;in, Represents the high-density gradient cluster center vector; This represents the matrix transpose operation. After calculation, the system performs distance comparison to classify water cluster attributes, comparing them against the preset spatial classification boundary values. The dynamic determination logic is as follows: ;in, This represents the mathematical expectation of the Euclidean distances of all features within the current cluster; The statistical standard deviation of the Euclidean distance between the corresponding features; This represents a pre-set disturbance tolerance adjustment factor based on historical sea state and hydrological conditions. The system compares each factor sequentially; if a decision is made... If the detection node is confirmed to meet the near-range aggregation characteristic, its hardware address is assigned to the memory table of the mutated water mass core region; if determined If this is confirmed, the node exhibits long-distance attenuation characteristics, and its hardware address is assigned to the edge region. Combined with... Figure 2 As shown in the figure, the three-dimensional coordinate axes map to the independent feature dimensions of the temperature-salinity spatiotemporal gradient vector. The system iteratively searches for high-density gradient cluster centers in the feature vector space. (As shown by the rhombus coordinate points in the figure) are used as the reference anchor points for distance measurement. Preset spatial classification boundary values ​​are used. For distance threshold, the characteristic Euclidean distance Detection nodes with a feature distance less than this boundary value (as shown by the densely distributed circular dots in the inner layer of the figure) are classified into the core region of the mutated water mass, representing water areas with drastic changes in hydrological characteristics; while nodes with a feature Euclidean distance greater than or equal to this boundary value (as shown by the radiating circular dots in the outer layer of the figure) are classified into the edge region. Through the three-dimensional clustering visualization of this feature space, the dynamic allocation benchmark of computing power nodes on the edge side of the mutated water mass can be intuitively verified.

[0031] Specifically, the process of allocating detection nodes within the core area of ​​the mutated water mass to perform contour tracing calculations based on the remaining wave energy storage capacity, extracting the vortex boundary topology point set, and unidirectionally transmitting it to the upper-level main network is as follows: Within the core area of ​​the mutated water mass, a sorted sequence of remaining wave energy storage capacity is selected, and the detection node corresponding to the highest position of the remaining wave energy storage capacity is designated as the master control tracking node; an edge computing scheduling command is issued to the master control tracking node, triggering the master control tracking node to perform point-by-point approximation calculations along the contour lines connecting the temperature-salinity-temporal gradient vectors within the core area of ​​the mutated water mass, locking the closed contour to generate the vortex boundary topology point set; the uplink communication bandwidth resources of the master control tracking node are allocated, the vortex boundary topology point set is modulated into an uplink carrier signal, and the uplink carrier signal is transmitted unidirectionally to the main network receiving terminal to transmit the vortex boundary topology point set.

[0032] In this implementation plan, the system traverses the memory table in the core area of ​​the mutated water mass and sequentially retrieves the wave energy storage capacity carried in the reported messages of each node. The main control logic constructs a resource scheduling queue by sorting the remaining wave energy storage capacity in descending order of numerical value. It directly extracts the detection node corresponding to the first node in the queue (i.e., the node with the most abundant energy reserves) and configures its physical network identifier as the main control tracking node. Upon receiving the edge computing scheduling command, the main control tracking node begins executing contour line tracing calculations within its local microprocessor. (Using the current coordinates...) To trace the starting point, search for the next discrete topological coordinate along the tangent direction where the temperature and salinity eigenvalues ​​are equal. The physical space conversion formula for point-by-point approximation is: ;in, represents the node sequence retrieval index within the core region of the mutated water mass; s represents the discrete topological coordinate iteration index on the contour line tracing path; Indicates the first contour line on the contour tracing path. A discrete topological coordinate system; This represents the next discrete topological coordinate generated after one directional iteration calculation; It represents the fixed tracking step size coefficient in the pointwise approximation operation of space; This represents an orthogonal rotation matrix in the probe two-dimensional plane, used to convert the gradient direction into the calculation direction of tangent contour lines; This indicates the hydrological gradient scalar field at the current coordinates. The directional derivative vector is then calculated. The master tracking node continuously iterates through the above physical calculations until the latest coordinate point is generated. When the absolute value of the spatial deviation from the tracking starting point is less than the step size coefficient, the trajectory achieves closed-loop locking in physical space. The system then aggregates and outputs all discrete coordinate points recorded on the tracking path as a vortex boundary topology point set. Finally, the master node reads its own communication baseband parameters, allocates corresponding uplink communication bandwidth resources according to the data packet throughput of the vortex boundary topology point set, modulates it into an uplink carrier signal through a digital-to-analog converter circuit, and drives the physical antenna to continuously transmit to the main network satellite receiving terminal, completing the one-way transparent transmission of boundary data.

[0033] Specifically, the process of fusing the vortex boundary topological point set with the preset ocean current vector field and iteratively calculating continuous displacement parameters based on the vorticity conservation condition to deduce the discrete drift coordinate sequence is as follows: Extract the velocity and direction attributes of the preset ocean current vector field, map the vortex boundary topological point set to the preset ocean current vector field, and generate an initial boundary point set with ocean current forcing terms; substitute the initial boundary point set into the vorticity conservation condition, perform successive integration operations along the fluid streamline direction of the preset ocean current vector field, and extract continuous displacement parameters; use the continuous displacement parameters to perform vector translation updates on the spatial coordinates of the initial boundary point set along the time axis, and output the discrete drift coordinate sequence under multiple time spans.

[0034] In this implementation, the cloud server reads the uplink transparent vortex boundary topology point set from its cache queue, maps and overlays its spatial coordinates onto a pre-stored global ocean current vector field layer, and extracts the velocity and direction attributes of each boundary node using an interpolation algorithm. These are then merged to generate an initial boundary point set with ocean current forcing. To deduce the dynamic migration path of the boundary point set, the server substitutes the initial boundary point set into a discretized vorticity conservation evolution equation. Successive integration along the fluid streamline direction is performed to obtain the continuous displacement parameters of each boundary point. The specific numerical integration logic is as follows: ;in, Spatial node index representing the boundary topological point set; The index representing the outlier of the time evolution; Indicates the first The node at the th Three-dimensional spatial coordinates under the step-by-step simulation cycle; Represents the preset ocean current vector field in spatial coordinates The background ocean current vector mapped at that location; This represents the core conserved vorticity scalar of the water mass; Indicates the first Coordinates of the vortex geometric center under the step-by-step derivation period; Represents the tangential unit direction vector of the current boundary node; This represents the time step parameter for forward time integration. Based on this, the system uses continuous displacement parameters to perform vector translation updates on the spatial coordinates to output the node positions for the next simulation cycle. ;in, The fluid viscosity-friction coefficient, representing the marine environment, is dynamically determined by a method that relies on the ratio of kinematic properties to physical scale. ; This represents the preset kinematic viscosity constant of the fluid in this sea area; Indicates the density of the background seawater; This represents the average equivalent physical diameter of the vortex boundary. The server iterates through multiple time spans, storing the continuously updated coordinate sets into a memory queue, and finally combines them to output a continuous discrete drift coordinate sequence. .

[0035] Specifically, the process of extracting discrete drift coordinate sequences as geometric control points to construct a three-dimensional irregular triangular network and sweeping along the time axis to generate a temporal spatial manifold envelope that encapsulates the evolution trajectory of hydrological anomalies is as follows: Spatial nodes within the discrete drift coordinate sequence are extracted as geometric control points. Delaunay triangulation is performed on the geometric control points in a three-dimensional Cartesian coordinate system, and the geometric control points are connected to construct a three-dimensional irregular triangular network. The time axis stretching step of the three-dimensional irregular triangular network is configured, and the three-dimensional irregular triangular network is driven to perform a lofting and sweeping operation along the time axis direction. The outer edge contour surfaces of the three-dimensional irregular triangular network are recorded during the sweeping process. The outer edge contour surfaces are stitched together to generate a closed three-dimensional geometric surface entity, which is then identified as the temporal spatial manifold envelope.

[0036] In this implementation scheme, the system outputs the aforementioned discrete drift coordinate sequence. Spatial nodes for each iteration cycle are extracted and transformed into geometric control points in a three-dimensional Cartesian coordinate system. The central processing unit (CPU) executes the Delaunay triangulation algorithm on the geometric control points within each cycle, thereby constructing a bottom-level three-dimensional irregular triangular mesh by connecting spatial straight lines. Subsequently, the system configuration drives the triangular mesh to extend upwards along the time axis with a time-axis stretching step size. The method for calculating and setting this step size is as follows: ;in, This indicates the system's allowed spatial resolution tolerance limit for mesh deformation; This represents the average spatial drift velocity vector as the discrete drift coordinate sequence evolves along the trajectory. During the lofting sweep operation, the system records the sweep parameter surface generated by the outer edge segments of the triangulated mesh during spatiotemporal extension along the evolution direction. The specific extended parameterized equation is as follows: ;in, The index representing the numbering of the boundary line segments of the outer contour of a three-dimensional irregular triangular mesh; and The normalized parameters representing the two-dimensional extension of the control plane all take values ​​in a closed real interval between zero and one. Indicates the first Spatial coordinates of the starting endpoints of the outline boundary line segment; Indicates the first The geometric direction vector of the contour boundary line segment. The system calls the graphics computing unit to repeatedly execute the above equations to calculate the swept outer edge contour surface in all evolution cycles. Finally, the initial and final triangulation meshes at both ends are extracted as endpoint caps, and vertex coincidence checks and geometric normal consistency stitching are performed on all edge contour surfaces and endpoint caps. After determining that there are no topological breaks at each joint, the underlying logic generates a physically closed three-dimensional geometric surface entity. The system then identifies it as the temporal spatial manifold envelope that encapsulates the evolution trajectory of hydrological anomalies.

[0037] Specifically, the process of performing a spatial bounding box intersection test on the envelope of the temporal spatial manifold and the global node 3D coordinate library to filter out the set of intersection node coordinates located inside the envelope of the temporal spatial manifold is as follows: extract the 3D extreme boundary of the envelope of the temporal spatial manifold, and construct an orthogonal axis bounding box based on the 3D extreme boundary; read the node physical coordinates in the global node 3D coordinate library, perform a spatial ray intersection test on the node physical coordinates and the orthogonal axis bounding box to determine the spatial intersection state of the node physical coordinates and the orthogonal axis bounding box; capture the node physical coordinates that are internally intersecting in the spatial intersection state, and aggregate the internally intersecting node physical coordinates to generate the set of intersection node coordinates.

[0038] In this implementation scheme, the cloud platform first traverses all the outer surface vertices of the temporal spatial manifold envelope, extracts geometric extrema along the coordinate axes of longitude, latitude, and water depth, and constructs an orthogonal axial bounding box by closing the extrema boundary of the Cartesian coordinate system. The logical expression for constructing the bounding box is: ;in, Represents the spatial physical dimension sequence of a three-dimensional coordinate system; Discrete indices representing the geometric vertices of the surface of the envelope of a temporal spatial manifold; Indicates the first The geometric vertex at the th Physical coordinate components in each dimension; This refers to the spatial expansion buffer margin that is forcibly introduced to accommodate errors caused by turbulence in the marine environment. This represents the Cartesian product operation across multiple intervals. After construction, the cloud platform reads the physical coordinates of all statically deployed nodes from the global node 3D coordinate library. First, fast spatial culling is performed using the upper and lower limits of the bounding box to filter out nodes that are clearly outside the abnormal region. For the nodes that are initially retained, the central processing unit performs a three-dimensional spatial ray cross-test, that is, using the physical coordinates of the node... A test ray extends outward from the base point into the bounding box. Calculate the total number of intersections between the ray and the outer surface of the manifold envelope. The intersection state determination function is expressed as: ;in, A unique serial number representing the global physical detection node; Indicates the first The actual deployment spatial coordinates of the physical node to be tested; This represents the total number of external triangular facets that constitute the envelope of the temporal spatial manifold; Indicates the index for traversing the triangle facet; The scalar representing the spatial extension of the test ray; Indicates the first The spatial geometric equation of the outer surface of the envelope; This represents the spatial Boolean intersection determination function between a ray and a surface. It outputs a constant of one when physical penetration occurs and a constant of zero when no penetration occurs. The central processing unit (CPU) retrieves the determination results and calculates the total number of intersections. For an odd number of detection nodes, the spatial intersection state is determined to be internal intersection according to the graph topology closure theorem. Then, the physical coordinates of all nodes exhibiting internal intersection are pushed into the memory stack and aggregated to generate the set of intersection node coordinates within the target warning area.

[0039] Specifically, the process of extracting the time evolution depth parameters corresponding to the intersection node coordinate set, and compiling the asynchronous wake-up time slot table and frequency conversion sampling command based on the time evolution depth parameters and sending them to the corresponding probe nodes is as follows: The intersection node coordinate set is back-projected onto the time axis profile of the time-series manifold envelope, and the time evolution depth parameters corresponding to the intersection node coordinate set on the time axis profile are extracted; the time evolution depth parameters are mapped to a preset hardware scheduling matrix, and the target wake-up delay period and target sampling frequency are extracted from the preset hardware scheduling matrix; the target wake-up delay period and target sampling frequency are merged to compile the asynchronous wake-up time slot table and frequency conversion sampling command; the hardware network identifier bound to the intersection node coordinate set is parsed, a downlink control channel is established based on the hardware network identifier, and the asynchronous wake-up time slot table and frequency conversion sampling command are sent to the corresponding probe nodes through the downlink control channel.

[0040] In this implementation scheme, the system substitutes each physical coordinate in the intersection node coordinate set into the principal direction equation of the temporal spatial manifold envelope, and calculates the temporal evolution depth parameter representing the predicted time taken for the anomalous water mass to arrive at the detection node by projecting a vertical projection line onto the evolution time axis profile. The specific projection conversion logic is as follows: ;in, This represents the three-dimensional geometric center coordinates of the anomalous water mass at the current calculation start time; This represents the average propagation direction normal vector of the manifold envelope as it expands and evolves along the time series. This represents the evolutionary conversion ratio of spatial physical distance to the temporal dimension. After obtaining the temporal evolution depth parameters of each node, the cloud platform uses these parameters as independent variables and maps them into a preset hardware scheduling matrix to solve for the target wake-up latency period allocated to each physical node. With target sampling frequency The underlying calculation equation for hardware scheduling parameters is as follows: ; ;in, This indicates an extremely low baseline sampling frequency when the underlying hardware sensors are in normal cruise mode; This represents the maximum high-frequency sampling gain coefficient in response to dramatic hydrological changes; It represents the exponential decay constant that causes the sensor sampling frequency to decrease as the prediction time margin increases; This represents the early wake-up safety margin deducted when the cloud sends instructions to the physical node. The dynamic determination method for this margin is strictly based on the actual cold start overhead of the underlying electronic hardware. ; here This represents the nominal value of the physical energy dissipation required for a sensor communication component to transition from a deep dormant state to an active working state during cold start. This indicates the current standby sustaining discharge power measured by the battery management module of the probe node; This represents the network tolerance constant used to compensate for delays in underwater acoustic communication channels. The central processing unit merges the target wake-up delay period and the target sampling frequency, assembles them into control command payloads according to the underlying communication protocol frame format, and finally establishes a downlink control channel from the cloud to the edge by parsing the underlying hardware network identifiers and routing tables bound to the intersection node coordinates. This accurately delivers the asynchronous wake-up time slot table containing the time schedule and the stepped frequency conversion sampling commands to the corresponding detection nodes.

[0041] Example 2; please refer to Figure 3 A cloud-edge collaborative marine hydrological dynamic monitoring system is used to execute the cloud-edge collaborative marine hydrological dynamic monitoring method described in the embodiments. It includes: a sensing network module, used to collect time-series data sequences of temperature, salinity, and current from a preset water depth profile and construct a three-dimensional temperature-salinity matrix; perform differential operations on the three-dimensional temperature-salinity matrix to extract the temperature-salinity spatiotemporal gradient vector; encapsulate the temperature-salinity spatiotemporal gradient vector and wave energy storage margin into a node status message; and trigger the formation of a self-organizing collaborative detection cluster via broadcast through an underwater acoustic channel. A boundary extraction module is used to parse the node status message, perform clustering and segmentation based on the temperature-salinity spatiotemporal gradient vector to divide the core and edge regions of the variability water mass, and allocate detection nodes within the core region of the variability water mass based on the wave energy storage margin to perform contour tracing operations and extract the vortex boundary topology. The point set is transmitted unidirectionally to the upper-level main network; the spatiotemporal extrapolation module is used to fuse the vortex boundary topology point set with the preset ocean current vector field, iteratively calculate the continuous displacement parameters based on the vortex conservation condition to extrapolate the discrete drift coordinate sequence, extract the discrete drift coordinate sequence as geometric control points to construct a three-dimensional irregular triangular network, and sweep along the time axis to generate a temporal spatial manifold envelope that encapsulates the hydrological anomaly evolution trajectory; the closed-loop scheduling module is used to perform spatial bounding box intersection tests on the temporal spatial manifold envelope and the global node three-dimensional coordinate library, screen out the intersection node coordinate set located inside the temporal spatial manifold envelope, extract the time evolution depth parameters corresponding to the intersection node coordinate set, compile an asynchronous wake-up time slot table and frequency conversion sampling commands based on the time evolution depth parameters, and send them to the corresponding detection nodes.

[0042] In this implementation scheme, the sensing network module is deployed at the underlying physical detection nodes, serving as the physical execution carrier for data acquisition and edge communication. This module drives the underwater sensor array via the underlying bus to acquire temperature, salinity, and current sampling values ​​at various depth profiles within a continuous time period, sequentially filling these values ​​into a preset grid to generate a three-dimensional temperature and salinity matrix. Subsequently, the module calls the built-in microprocessor to perform discrete difference operations along the coordinate axes of this matrix to obtain the spatiotemporal gradient vector of temperature and salinity, while simultaneously polling the node power management board to read the remaining energy stored from wave energy conversion. The module then concatenates the extracted gradient vector, remaining energy storage, and node hardware address bit-by-bit to generate a node status message, and controls the underwater acoustic communication transducer to broadcast this message to a specific frequency band. Upon receiving handshake confirmation signals from neighboring nodes, a local communication link is established, thereby physically aggregating to form a self-organizing cooperative detection cluster.

[0043] The boundary extraction module runs within the node microprocessors of the self-organizing cooperative detection cluster, responsible for determining the region of the environment and allocating computing power. This module reverse-engineers the received node status messages, extracts the temperature-salinity-temporal gradient vectors of adjacent nodes, projects them spatially, and calculates the characteristic Euclidean distance. By comparing this distance with the dynamically calculated classification boundary value, nodes that meet the proximity aggregation characteristics are stored in the core region memory table of the mutated water mass, while the remaining nodes are assigned to the edge region. For the core region memory table, this module reads the corresponding wave energy storage capacity, sorts it in descending order, and retrieves the computing resources of the node with the highest energy level as the master tracking node. The master tracking node performs point-by-point approximation calculations along the gradient isometry line direction. After the first and last coordinates are locked, a vortex boundary topology point set is generated. Finally, the point set is transmitted unidirectionally to the cloud main network via radio frequency or uplink acoustic communication unit.

[0044] The spatiotemporal extrapolation module is deployed on a cloud server, performing large-scale ocean dynamics calculations and geometric reconstructions. This module reads the transparent set of vortex boundary topological points from the data receiving port, substitutes their spatial coordinates into a pre-set ocean current vector field containing velocity and direction attributes, and performs forced term fusion. Combining the vorticity conservation condition, the module performs successive numerical integration calculations along the fluid streamline direction for each boundary point to obtain continuous displacement parameters, updates the original spatial coordinates with vector translation, and generates a discrete drift coordinate sequence through multi-time-span iterative iterations. Subsequently, the module uses this discrete coordinate sequence as control points to perform Delaunay triangulation to construct a three-dimensional irregular triangular network. Based on the calculated time stretching step, it drives the triangular network to perform a lofting and sweeping operation, stitching together the outer patches generated during the sweeping process, and finally generating a physically closed temporal spatial manifold envelope.

[0045] The closed-loop scheduling module, also running in the cloud, is responsible for the reverse timing control of the edge hardware's operating status. This module extracts the three-dimensional extrema of the temporal spatial manifold envelope to construct an orthogonal axial bounding box. It then reads the statically deployed physical coordinates from the global node three-dimensional coordinate library and emits test rays towards this bounding box. By statistically analyzing the intersection states, it filters out nodes located within the envelope to generate an intersection node coordinate set. For the intersection node coordinates, the module projects them backward onto the time axis profile of the envelope, calculates the temporal evolution depth parameter characterizing the arrival time of hydrological anomalies, and substitutes this parameter into a preset hardware scheduling matrix for mapping calculation, extracting the corresponding target wake-up delay period and frequency conversion sampling frequency. Finally, the module merges the above scheduling period and frequency values ​​to generate a command payload, establishes a downlink channel based on the underlying hardware network identifier bound to the intersection nodes, and accurately sends commands containing scheduling information to the corresponding detection nodes.

[0046] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for dynamic monitoring of marine hydrology based on cloud-edge collaboration, characterized in that, Includes the following steps: S1. Collect time-series data of temperature, salinity and current from a preset water depth profile and construct a three-dimensional temperature and salinity matrix. Perform differential operation on the three-dimensional temperature and salinity matrix to extract the temperature and salinity spatiotemporal gradient vector. Encapsulate the temperature and salinity spatiotemporal gradient vector and wave energy storage margin into a node status message. Trigger the formation of a self-organizing cooperative detection cluster by broadcasting through the underwater acoustic channel. S2. Parse the node status message, perform clustering and segmentation based on the temperature-salinity-temporal gradient vector to divide the core area and edge area of ​​the mutated water mass, and allocate detection nodes in the core area of ​​the mutated water mass according to the wave energy storage margin to perform contour tracing calculation, extract the vortex boundary topology point set and transmit it unidirectionally to the upper-layer main network. S3. The topological point set of the vortex boundary is fused with the pre-set ocean current vector field. Based on the vortex conservation condition, the continuous displacement parameters are iteratively calculated to deduce the discrete drift coordinate sequence. The discrete drift coordinate sequence is extracted as geometric control points to construct a three-dimensional irregular triangular network. The time-series spatial manifold envelope that wraps the evolution trajectory of hydrological anomalies is generated by sweeping along the time axis. S4. Perform a spatial bounding box intersection test on the envelope of the temporal spatial manifold and the global node 3D coordinate library, filter out the coordinate set of intersection nodes located inside the envelope of the temporal spatial manifold, extract the time evolution depth parameters corresponding to the coordinate set of intersection nodes, compile an asynchronous wake-up time slot table and frequency conversion sampling command based on the time evolution depth parameters and send them to the corresponding detection nodes.

2. The ocean hydrological dynamic monitoring method based on cloud-edge collaboration according to claim 1, characterized in that: The specific process of collecting time-series temperature, salinity, and current data sequences from a preset water depth profile and constructing a three-dimensional temperature, salinity, and current matrix, and then performing a difference operation on the three-dimensional temperature, salinity, and current matrix to extract the spatiotemporal gradient vector of temperature, salinity, and current is as follows: Activate a distributed sensing array at a fixed underwater depth to acquire synchronous temperature scales, salinity, and current velocity parameters. Combine the synchronous temperature scales, salinity, and current velocity parameters with the corresponding depth coordinate metadata to generate a time series of temperature, salinity, and current data. The time series data of temperature, salinity and flow is mapped to a preset spatial grid along the depth axis and the time axis, and the preset spatial grid is filled to generate a three-dimensional temperature, salinity matrix. Discrete differential operators are applied along the coordinate axes of the three-dimensional temperature-salinity matrix to obtain spatial and temporal gradient components. The spatial and temporal gradient components are then concatenated to synthesize the temperature-salinity spatiotemporal gradient vector.

3. The method for dynamic monitoring of marine hydrology based on cloud-edge collaboration according to claim 1, characterized in that: The specific process of encapsulating the temperature-salinity spatiotemporal gradient vector and the wave energy storage margin into a node status message, and triggering the formation of a self-organizing cooperative detection cluster via underwater acoustic channel broadcast is as follows: Read the real-time voltage monitoring data output by the hardware power management module, and convert the real-time voltage monitoring data to generate wave energy storage capacity. The splicing of temperature-salinity-temporal gradient vector, wave energy storage margin and physical identifier of detection node constitutes a payload data frame. The payload data frame is encapsulated into a node status message by adding frame header check bits. The system transmits node status messages to a preset underwater acoustic communication frequency band, receives handshake confirmation frames returned by adjacent detection nodes, establishes local communication links by binding adjacent detection nodes that returned handshake confirmation frames, and aggregates along the local communication links to form a self-organizing cooperative detection cluster.

4. The ocean hydrological dynamic monitoring method based on cloud-edge collaboration according to claim 1, characterized in that: The specific process of parsing node status messages and performing clustering segmentation based on temperature-salinity spatiotemporal gradient vectors to divide the core and edge regions of the anomalous water mass is as follows: The temperature-salinity spatiotemporal gradient vector is extracted from the node status messages in the self-organizing cooperative detection cluster by reverse unpacking. The temperature-salinity spatiotemporal gradient vector is then mapped to the feature vector space, and a high-density gradient cluster center is iteratively searched in the feature vector space. Calculate the characteristic Euclidean distance from each temperature-salinity spatiotemporal gradient vector to the high-density gradient aggregation center, and numerically compare the characteristic Euclidean distance with the preset spatial classification boundary value; If the Euclidean distance of the comparison and judgment feature is less than the preset spatial classification boundary value, then the corresponding detection node is confirmed to have close-range aggregation features and is classified into the core area of ​​the mutated water mass. If the Euclidean distance of the comparison and judgment feature is greater than or equal to the preset spatial classification boundary value, then the corresponding detection node is confirmed to have long-distance attenuation feature and is classified into the edge region.

5. The method for dynamic monitoring of marine hydrology based on cloud-edge collaboration according to claim 1, characterized in that: Based on the remaining wave energy storage capacity, the process of allocating detection nodes within the core area of ​​the mutated water mass to perform contour tracing calculations, extracting the vortex boundary topology point set, and unidirectionally transmitting it to the upper-level main network is as follows: Within the core area of ​​the mutated water mass, a sorting sequence of wave energy storage capacity was selected, and the detection node corresponding to the highest wave energy storage capacity was designated as the main control tracking node. The edge computing scheduling command is sent to the master tracking node, which triggers the master tracking node to perform point-by-point approximation calculations along the iso-line connecting the temperature-salinity-temporal gradient vector in the core area of ​​the mutated water mass, and locks the closed contour to generate the vortex boundary topology point set; Allocate uplink communication bandwidth resources of the master control tracking node, modulate the vortex boundary topology point set into an uplink carrier signal, and transmit the uplink carrier signal to the main network receiving terminal to unidirectionally transmit the vortex boundary topology point set.

6. The method for dynamic monitoring of marine hydrology based on cloud-edge collaboration according to claim 1, characterized in that: The specific process of fusing the vortex boundary topological point set with a pre-set ocean current vector field and iteratively calculating continuous displacement parameters based on the vortex conservation condition to deduce the discrete drift coordinate sequence is as follows: Extract the velocity and direction attributes of the preset ocean current vector field, map the vortex boundary topology point set to the preset ocean current vector field, and generate an initial boundary point set with ocean current forcing term; Substitute the initial boundary point set into the vorticity conservation condition, perform successive integration along the fluid streamline direction of the preset ocean current vector field, and extract the continuous displacement parameters; The spatial coordinates of the initial boundary point set are updated by vector translation using continuous displacement parameters along the time axis, and the discrete drift coordinate sequence under multiple time spans is output.

7. The ocean hydrological dynamic monitoring method based on cloud-edge collaboration according to claim 1, characterized in that: The specific process of extracting discrete drift coordinate sequences as geometric control points to construct a three-dimensional irregular triangular network, and then sweeping along the time axis to generate a temporal spatial manifold envelope that encapsulates the evolution trajectory of hydrological anomalies, is as follows: Spatial nodes within the discrete drift coordinate sequence are extracted as geometric control points. Delaunay triangulation is performed on the geometric control points in a three-dimensional Cartesian coordinate system, and three-dimensional irregular triangular meshes are constructed by connecting the geometric control points. Configure the time axis stretching step of the 3D irregular triangular mesh, drive the 3D irregular triangular mesh to perform lofting and sweeping operations along the time axis, and record the outer edge contour surface of the 3D irregular triangular mesh during the sweeping process; The outer edge contour surface is stitched together to generate a closed three-dimensional geometric surface entity, which is then identified as the temporal spatial manifold envelope.

8. The method for dynamic monitoring of marine hydrology based on cloud-edge collaboration according to claim 1, characterized in that: The specific process of performing a spatial bounding box intersection test on the envelope of the temporal spatial manifold and the global node 3D coordinate library to filter out the set of intersection node coordinates located inside the envelope of the temporal spatial manifold is as follows: Extract the three-dimensional extremum boundary of the temporal spatial manifold envelope, and construct an orthogonal axial bounding box by combining the three-dimensional extremum boundary. Read the physical coordinates of nodes in the global node 3D coordinate library, perform a spatial ray intersection test on the bounding box of the node physical coordinates and the orthogonal axis, and determine the spatial intersection status of the node physical coordinates and the orthogonal axis bounding box. Extract the physical coordinates of nodes that intersect internally in space, and aggregate the physical coordinates of these nodes to generate a set of intersection node coordinates.

9. The method for dynamic monitoring of marine hydrology based on cloud-edge collaboration according to claim 1, characterized in that: The specific process of extracting the time evolution depth parameters corresponding to the intersection node coordinate set, compiling the asynchronous wake-up time slot table based on the time evolution depth parameters, and issuing the frequency conversion sampling command to the corresponding detection node is as follows: The coordinate set of the intersection nodes is back-projected onto the time axis profile of the temporal space manifold envelope, and the temporal evolution depth parameters corresponding to the coordinate set of the intersection nodes on the time axis profile are extracted. The time evolution depth parameter is mapped to the preset hardware scheduling matrix, the target wake-up delay period and the target sampling frequency are extracted from the preset hardware scheduling matrix, and the target wake-up delay period and the target sampling frequency are merged to compile the asynchronous wake-up time slot table and the frequency conversion sampling instruction. The hardware network identifier bound to the intersection node coordinate set is parsed, and a downlink control channel is established based on the hardware network identifier. The asynchronous wake-up time slot table and frequency conversion sampling command are sent to the corresponding detection node through the downlink control channel.

10. A cloud-edge collaborative marine hydrological dynamic monitoring system, used to execute the cloud-edge collaborative marine hydrological dynamic monitoring method according to any one of claims 1-9, characterized in that, include: The sensing network module is used to collect time-series data of temperature, salinity and current from a preset water depth profile and construct a three-dimensional temperature and salinity matrix. It performs differential operations on the three-dimensional temperature and salinity matrix to extract the temperature and salinity spatiotemporal gradient vector. The temperature and salinity spatiotemporal gradient vector and the wave energy storage margin are encapsulated into a node status message, which is broadcast through the underwater acoustic channel to trigger the formation of a self-organizing cooperative detection cluster. The boundary extraction module is used to parse node status messages, perform clustering and segmentation based on temperature-salinity-temporal gradient vectors to divide the core area and edge area of ​​the mutated water mass, and allocate detection nodes within the core area of ​​the mutated water mass to perform contour tracing calculations based on the wave energy storage margin, extract the vortex boundary topology point set and transmit it unidirectionally to the upper-layer main network. The spatiotemporal extrapolation module is used to fuse the vortex boundary topological point set with the preset ocean current vector field, iteratively calculate the continuous displacement parameters based on the vorticity conservation condition to extrapolate the discrete drift coordinate sequence, extract the discrete drift coordinate sequence as geometric control points to construct a three-dimensional irregular triangular network, and sweep along the time axis to generate a temporal spatial manifold envelope that encapsulates the evolution trajectory of hydrological anomalies. The closed-loop scheduling module is used to perform spatial bounding box intersection tests on the envelope of the temporal spatial manifold and the global node 3D coordinate library, filter out the coordinate set of intersection nodes located inside the envelope of the temporal spatial manifold, extract the time evolution depth parameters corresponding to the coordinate set of intersection nodes, compile an asynchronous wake-up time slot table and frequency conversion sampling commands based on the time evolution depth parameters, and send them to the corresponding detection nodes.

Images