Exchange area sea area three-dimensional collaborative observation system based on submerged buoy and buoy

Abstract

The invention relates to the technical field of marine environment monitoring, in particular to an exchange area sea area three-dimensional collaborative observation system based on submerged buoys, which comprises a data acquisition module, an instruction generation module, a buoy dynamic response module and a space-time alignment and fusion module. The method comprises the following steps: acquiring original water observation data in multiple depth layers through a subsurface buoy system, and constructing a structured feature data set; identifying an ocean abnormal phenomenon based on the abnormal detection model and generating a trigger instruction; the buoy system analyzes the instruction and then drives the lifting mechanism to be accurately positioned to a target depth layer, and multi-sensor synchronous acquisition is executed according to a matched observation mode; and the fusion module adopts time calibration and a Kriging interpolation algorithm to realize three-dimensional data reconstruction, and outputs a space-time synchronous three-dimensional data set. The method has the characteristics of intelligent recognition, dynamic response, multi-parameter fusion and the like, the capture precision and observation efficiency of the complex process of the exchange area are improved, and technical support is provided for ocean dynamic process research and ecological monitoring.

Description

Technical Field

[0001] The present invention relates to the technical field of marine environment monitoring, and particularly to a three-dimensional collaborative observation system for the exchange area sea area based on submersible buoys and surface buoys. Background Technique

[0002] Under the background of frequent evolution of climate change and regional ocean dynamic processes, the exchange area sea area (such as the continental shelf edge, circulation convergence zone, etc.), as an important node for material transport and energy conversion, its water body structure and dynamic processes have important impacts on the marine ecosystem, carbon cycle and even disaster warning. Due to the characteristics of significant water body stratification, intense flow field disturbance and coexistence of multi-scale phenomena in the exchange area, it is urgent to obtain high-resolution three-dimensional observation data in the vertical scale to capture the spatial structure and evolution dynamics of key processes such as frontal pycnocline, mesoscale eddy, internal wave excitation, etc. In recent years, submersible buoys and surface buoys, as complementary deep and shallow ocean observation platforms, have shown great potential in regional collaborative monitoring and have become an important technical path for developing high-precision three-dimensional observation systems.

[0003] However, the existing observation methods of submersible buoys and surface buoys mainly deploy in parallel independently and lack an effective information linkage mechanism. On the one hand, although submersible buoys can be fixed at a specific depth for long-term high-frequency monitoring, it is difficult to respond to complex abnormal events in real time; on the other hand, although surface buoys have mobility and controllable lifting ability, most samplings are still carried out at a fixed frequency, lacking the ability to actively trigger and command drive key deep-layer phenomena. In addition, the existing surface buoy response system lacks a fine time alignment and spatial fusion mechanism, making it difficult to construct a continuous three-dimensional observation field of submersible and surface buoy data in a unified grid, which limits the comprehensive utilization and scientific interpretation ability of the data. Summary of the Invention

[0004] The present invention provides a three-dimensional collaborative observation system for the exchange area sea area based on submersible buoys and surface buoys, and constructs a three-dimensional collaborative observation system integrating anomaly detection, response regulation and multi-source data fusion to improve the deep anomaly identification ability and three-dimensional process reconstruction accuracy of the exchange area sea area.

[0005] A three-dimensional collaborative observation system for the exchange area sea area based on submersible buoys and surface buoys includes a data acquisition module, an instruction generation module, a surface buoy dynamic response module, and a time-space alignment and fusion module, wherein; The data acquisition module collects the original observation data stream of the preset depth layer through the submersible buoy system, divides the original observation data stream into data blocks according to the time window, extracts the feature vectors of each data block and generates a submersible buoy observation data set; The instruction generation module inputs the submersible buoy observation data set into a preset anomaly detection model. When identifying the characteristics of the target ocean phenomenon, it generates a trigger instruction including the phenomenon type, occurrence depth and time stamp; The buoy dynamic response module transmits the trigger instruction to the buoy system through an underwater acoustic communication link. The buoy matches the preset observation mode according to the phenomenon type in the trigger instruction, drives the lifting mechanism to position to the target depth layer, and starts multi-sensor synchronous acquisition to form a three-dimensional observation data packet. The spatio-temporal alignment and fusion module performs time calibration on the three-dimensional observation data packet based on the timestamp in the trigger instruction, and performs three-dimensional interpolation in combination with the spatial coordinates of the mooring observation data set to generate a spatio-temporally synchronized three-dimensional data set.

[0006] Optionally, the data acquisition module includes: By using the CTD (Conductivity, Temperature, Depth), ADCP (Acoustic Doppler Current Profiler) and dissolved oxygen sensor carried by the mooring system, continuously collect the original observation data stream at multiple preset depth layers. The original observation data stream includes: The temperature value, salinity value and depth value output by the CTD; The eastward velocity, northward velocity and vertical velocity output by the ADCP; The dissolved oxygen concentration and dissolved oxygen saturation output by the dissolved oxygen sensor; The timestamps of all sensors' synchronous acquisitions.

[0007] Optionally, the data acquisition module further includes: Divide the original observation data stream into continuous data blocks according to a time window with a fixed duration of 30 minutes; Extract the time-domain statistical features and frequency-domain features for each data block respectively. The time-domain statistical features include mean, variance and extreme value, and the frequency-domain feature is the energy feature calculated based on wavelet transform; Integrate the time-domain statistical features and frequency-domain features to form a feature vector, and bind it to the depth layer number and the acquisition start and end timestamps of the corresponding data block to construct a structured mooring observation data set; Real-time store the mooring observation data set through the underwater data storage unit of the mooring system and synchronously send it to the instruction generation module.

[0008] Optionally, the instruction generation module includes: Receive the structured mooring observation data set and input it into a preset anomaly detection model. The anomaly detection model outputs the type identifier of the target ocean phenomenon, the corresponding depth layer number, the corresponding confidence score, and the acquisition start and end timestamps of the data block to which the corresponding feature vector belongs;

[0009] When the confidence score exceeds the preset threshold of 0.85, generate a trigger instruction based on the type identifier, depth layer number and acquisition start and end timestamps output by the model. The trigger instruction includes the phenomenon type, occurrence depth and timestamp; Store the trigger instruction in the instruction cache queue, and prepare to send it to the buoy system through the underwater acoustic communication link.

[0010] Optionally, the buoy dynamic response module includes: Receive the trigger instruction from the instruction generation module through the underwater acoustic communication link, and parse to obtain the phenomenon type, occurrence depth, and timestamp therein; Match the preset observation mode in the buoy system according to the phenomenon type, and the observation mode includes different sensor combination schemes and sampling frequencies; Drive the lifting mechanism of the buoy to perform target depth layer positioning according to the occurrence depth, and in the positioning process, real-time feedback the current depth information and perform closed-loop correction control to ensure that the deviation does not exceed the set error range.

[0011] Optionally, after the buoy dynamic response module completes the depth positioning, it starts the following multi-sensor synchronous acquisition process according to the matched observation mode: Start the CTD profiler to obtain the temperature, salinity, and depth data of the current layer; Start the turbulence profiler to obtain the turbulence energy distribution and shear rate data; Start the bio-optical sensor to obtain the particle scattering intensity and chlorophyll fluorescence parameters; Continuously sample in the target depth layer for no less than 5 minutes, and bind the sampling data with the corresponding timestamp and depth layer coordinates to generate a three-dimensional observation data packet.

[0012] Optionally, the buoy system transmits the generated three-dimensional observation data packet back to the shore-based center in real time through its waterborne wireless transmission unit, and retains a local cache for the fault retransmission mechanism.

[0013] Optionally, the spatio-temporal alignment and fusion module includes: Receive the three-dimensional observation data packet and extract its timestamp and depth layer coordinates; Perform time calibration based on the timestamp, and correct the clock deviation between the buoy system and the mooring system to within ±0.1 seconds; Call the observation records in the mooring observation dataset corresponding to the time window and depth layer for spatial position matching.

[0014] Optionally, the spatio-temporal alignment and fusion module further includes: After completing the spatial position matching, use the Kriging interpolation algorithm to perform three-dimensional interpolation operations on the same type of observation parameters in the mooring observation dataset and the three-dimensional observation data packet to generate a three-dimensional grid data volume of temperature, salinity, flow velocity, and dissolved oxygen concentration; Combine the interpolation result with the time calibration result, output a spatio-temporally synchronized three-dimensional dataset including temperature field, salinity field, flow velocity field and dissolved oxygen concentration field, and store and synchronously transmit this three-dimensional dataset to the shore-based analysis platform.

[0015] Advantages of the present invention: In the present invention, by introducing the mooring observation modeling method of "time window + multi-feature fusion" into the data acquisition module, combining the depth anomaly detection model and the confidence calculation mechanism in the instruction generation module, it is possible to achieve deep and high-confidence real-time recognition of complex target phenomena such as fronts, vortices, and mixed layers in the ocean. After the response instruction is generated, the buoy dynamic response module automatically matches the observation mode based on the phenomenon type, and quickly locates to the target layer through the depth closed-loop control algorithm, realizing a response delay of less than 2 minutes to ensure the spatio-temporal linkage observation requirements of sudden phenomena.

[0016] In the present invention, through the multi-sensor collaborative acquisition mechanism deployed by the buoy system, multi-source high-frequency parameters such as temperature, salinity, flow velocity, dissolved oxygen, turbulence, and bio-optics can be obtained in real time when the observation is triggered. At the same time, after the acquisition is completed, the spatio-temporal alignment and fusion module introduces the Kriging interpolation method to perform three-dimensional reconstruction of the mooring historical data and the buoy response data in the unified time and unified depth grid, generating a temperature field, salinity field, flow velocity field, and dissolved oxygen field with high resolution and strong consistency, providing high-quality original support for multi-scale dynamic process research, ecological anomaly diagnosis, and data assimilation.

[0017] In the present invention, a modular architecture design is adopted, and the four subsystems of data acquisition, instruction generation, response execution, and data fusion are linked in series. Supplemented by the FIFO instruction scheduling, confidence threshold regulation, lifting mechanism error closed-loop correction, interpolation error control, and feedback confirmation mechanism, a complete observation closed-loop is formed. The system has the engineering capabilities of task traceability, parameter configurability, and anomaly backfilling, and is suitable for extended deployment and operation in multi-type floating and diving joint platforms, multi-water layer response strategies, and complex exchange area environments, significantly improving the practicability and stability of the observation system. Description of the drawings

[0018] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings in the following description are only those of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0019] Figure 1 It is a schematic diagram of the system flow of the embodiment of the present invention; Figure 2 It is a schematic diagram of the flow of the buoy dynamic response module of the embodiment of the present invention. Detailed implementation manners

[0020] The present invention will be described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments. For some well-known technologies, those skilled in the art can also adopt other alternative methods for implementation; moreover, the accompanying drawings are only for more specifically describing the embodiments, and are not intended to specifically limit the present invention.

[0021] As Figure 1 - Figure 2 shown, a three-dimensional collaborative observation system for the exchange area sea area based on submersibles and buoys includes a data acquisition module, an instruction generation module, a buoy dynamic response module, and a spatio-temporal alignment and fusion module, wherein; The data acquisition module acquires the original observation data stream of the preset depth layer through the submersible system, divides the original observation data stream into data blocks according to the time window, extracts the feature vectors of each data block, and generates a submersible observation data set; The instruction generation module inputs the submersible observation data set into a preset anomaly detection model. When the characteristics of the target ocean phenomenon are recognized, a trigger instruction including the phenomenon type, the occurrence depth, and the timestamp is generated; The buoy dynamic response module transmits the trigger instruction to the buoy system through the underwater acoustic communication link. The buoy matches the preset observation mode according to the phenomenon type in the trigger instruction, drives the lifting mechanism to position to the target depth layer, and starts the multi-sensor synchronous acquisition to form a three-dimensional observation data packet; The spatio-temporal alignment and fusion module performs time calibration on the three-dimensional observation data packet based on the timestamp in the trigger instruction, and performs three-dimensional interpolation in combination with the spatial coordinates of the submersible observation data set to generate a spatio-temporally synchronized three-dimensional data set.

[0022] The data acquisition module includes: 1. Multi-sensor collaborative acquisition unit: The data acquisition module realizes the real-time acquisition of the original observation data by integrating the following marine environment monitoring sensors in the submersible system: Conductivity-temperature-depth profiler (CTD): used to measure the temperature (T), salinity (S), and depth (D) of the water body; Acoustic Doppler current profiler (ADCP): obtains the three-component flow velocity data of each measurement point on the profile, namely the eastward flow velocity u, the northward flow velocity v, and the vertical flow velocity w; Dissolved oxygen sensor (DO Sensor): provides the dissolved oxygen concentration DO and the dissolved oxygen saturation DOS of the water body; System clock module: ensures that all sensor output data are attached with accurate and unified timestamps t.

[0023] The above-mentioned various sensors synchronously sample at a preset time interval (the default sampling period is 10 seconds), and form an original observation data sequence in the following format: ; Each set of observed data points is bound to a unique timestamp to ensure the temporal consistency of multi-parameter data.

[0024] 2. Time window segmented response mechanism. The collected data is processed in segments according to a fixed time window. The default time window is 30 minutes. Processing unit: To support subsequent anomaly detection, it means that the observed data for every consecutive 30 minutes is grouped into a data block. Suppose the mooring observation period is , data blocks can be generated: , ; Each data block corresponds to a collection depth layer number and the time range .

[0025] Among them, is the floor function.

[0026] 3. Feature extraction and vector construction unit. To achieve efficient anomaly detection, the system extracts time-domain and frequency-domain features from each data block to form a feature vector. The specific steps are as follows: (1) Time-domain statistical feature extraction: For each physical quantity , the following statistical features are extracted: Mean: ; Variance: ; Extreme value: , ; (2) Frequency-domain wavelet energy feature extraction: Among them is the number of data points in the data block, usually 180 (30 minutes × 60 seconds ÷ 10 seconds / sampling point). By performing discrete wavelet transform (DWT) on the time series signal , the wavelet sub-band energy is extracted as the frequency-domain feature. Taking the Daubechies-4 wavelet basis as an example, the signal is decomposed into 3 layers, and the energy distribution of each scale is calculated: ; Among them represents the i-th coefficient after the j-th layer of wavelet decomposition.

[0027] (3)Example of feature vector construction: Taking temperature as an example, its feature vector is as follows: ; Extract the above features for all observed variables respectively, and splice them to obtain a complete feature vector: ; 4. Dataset Structuring and Caching Unit: Each feature vector is bound to the data block number k it belongs to, the acquisition depth layer number , and the time range , forming the following structured observation data item: ; All are arranged in chronological order to form the moored observation dataset : ; Finally, the data acquisition module stores the moored observation dataset in real time through the underwater data storage unit and transmits it to the instruction generation module through the internal communication interface for subsequent ocean phenomenon recognition and trigger response processing to call.

[0028] Example of the process of extracting data at a specific depth layer: Taking the 75-meter depth layer as an example, the whole process operation of the data acquisition module is as follows: Original acquisition: The mooring buoy deploys a CTD, an ADCP, and a dissolved oxygen sensor at a depth of 75 meters. The sampling period is 10 seconds, and a total of 180 data points are collected within 30 minutes. Each group of data contains: Temperature value (e.g., 12.3 °C), salinity value (e.g., 33.8 PSU), depth value (fixed at 75 meters), three-dimensional flow velocity (e.g., 0.42 m / s eastward, -0.17 m / s northward, 0.03 m / s vertically), dissolved oxygen concentration (e.g., 6.2 mg / L) and saturation (e.g., 92.4%), timestamp (e.g., 2025-06-07 10:00:00); Data chunking: The above 180 groups of data are combined into a data block. Its corresponding time window is "2025-06-07 10:00:00 – 10:29:59", numbered B1, and the depth identifier is Z = 75.

[0029] Feature extraction: Extract the following statistical and frequency domain features for each physical quantity in this data block: The average value of the temperature signal within these 30 minutes is 12.4 °C, the standard deviation is 0.08, and the extreme values are [12.2 °C, 12.6 °C]; Analyze the temperature sequence through wavelet transform to obtain the wavelet energy distribution at each scale, reflecting the degree of periodic oscillation; Similarly process variables such as salinity, each component of flow velocity, and dissolved oxygen to obtain feature vector segments of multiple variables.

[0030] Feature vector assembly: Concatenate the feature segments of all variables into a complete feature vector with a total length of approximately 60 to 80 dimensions, serving as the characterization result of this time period at this depth layer.

[0031] Dataset generation: Combine this feature vector with the labels "Z = 75" and "start and end timestamps = 10:00:00–10:29:59" into a structured data item; This item is then stored in the local cache of the submersible buoy and simultaneously transmitted to the instruction generation module for use as the discriminant basis for whether to trigger the buoy response.

[0032] The instruction generation module includes: 1. Submersible buoy observation data reception and buffer processing: This module first receives the structured submersible buoy observation dataset transmitted from the data acquisition module. Each record in the dataset includes: Feature vector : Reflecting multi-parameter statistics and frequency domain characteristics within a single time window (e.g., 30 minutes); Depth layer number : Indicating the sampling depth of the submersible buoy to which the data belongs; Time range : The acquisition time period of this data block; All received data first enters the input buffer for integrity verification and timing rearrangement processing to ensure that the input data is continuous, packet loss-free, and in the correct order.

[0033] 2. Anomaly detection model and recognition logic: Each received feature vector will be fed into a preset anomaly detection model. This model has a semi-supervised neural network structure, uses a variational autoencoder as an anomaly detector, and is supplemented by a posterior probability scoring function to judge the degree of anomaly.

[0034] (1) Model structure description: The encoder maps the feature vector to a latent space distribution , and reconstructs the original input; The decoder samples from the latent space to generate a reconstructed vector , and calculates the reconstruction error; The degree of anomaly is judged using a dual index of reconstruction error and distribution deviation: ; Among them, is the weighting coefficient, is the Kullback-Leibler divergence metric.

[0035] (2) Anomaly confidence calculation: Map the loss function to a normalized confidence score , indicating the suspicious degree of the ocean phenomenon corresponding to the current data block: ; where are the minimum and maximum errors within the sliding window, used for normalization processing.

[0036] 3. Confidence determination and event classification: When the confidence score exceeds the preset threshold (0.85), the system marks this data block as "potential target ocean phenomenon" and further conducts event type identification.

[0037] (1) Use a lightweight multi-layer perceptron (MLP) to classify abnormal samples and output the corresponding phenomenon type labels (such as: mesoscale eddy, frontal crossing, mixed layer), and the phenomenon type is denoted as ; (2) Event determination conditions: When the following conditions are met, the response process is triggered: (confidence determination threshold, default ; the classifier output is not empty and has a clear physical meaning; the mooring buoy is within the subjective measurement water layer range (such as 50 - 150 meters).

[0038] 4. Trigger instruction generation mechanism: After determining the existence of an ocean anomaly, the system immediately constructs a trigger instruction. Each trigger instruction includes the phenomenon type, occurrence depth, and timestamp, and the complete instruction structure is: , is the timestamp; This trigger instruction will be stored in the instruction cache queue in the form of a structured message, along with a unique identification code and a transmission status label (such as "to be sent").

[0039] Example illustration: The following is a schematic of an actual working process of the instruction generation module: Input: The mooring buoy system collects 30 minutes of data at a depth of 75 meters, and the feature vector extraction is completed; Model output: The anomaly detection model determines that the confidence = 0.91, and the phenomenon type is frontal crossing; Generate instruction: Construct the instruction = ⟨frontal crossing, 75m, 2025 - 06 - 07 11:00:00>; Output status: The instruction is sent to the cache queue and marked as the "to be acoustic transmission" status, waiting for the opportunity of downlink acoustic communication.

[0040] Instruction scheduling logic: It consists of four parts: a cache queue mechanism, a scheduling priority rule, a communication window management strategy, and a fault-tolerant feedback control mechanism.

[0041] 1. Instruction cache queue mechanism: All trigger instructions output by the anomaly detection and recognition process enter the instruction cache queue uniformly. Each instruction is stored in a structured format, including: instruction number (unique identifier), instruction content (phenomenon type, depth, timestamp), enqueue time, and instruction status (pending transmission, transmitted, waiting for confirmation, resending pending). The system adopts the first-in-first-out (FIFO) mechanism for enqueue management to ensure that the instructions generated first are scheduled first; for multiple instructions generated simultaneously, they can be adjusted for queue jumping according to the priority rule (see below).

[0042] 2. Instruction scheduling priority rule: To meet the response time requirements of different types of ocean phenomena, the system assigns a dynamic priority identifier to each instruction. The priority is calculated considering the following factors: Phenomenon urgency : Determined by the event type (e.g., frontal surface > turbulent mixing > mesoscale vortex); Depth adaptability of occurrence : Weighted near the buoy observable area (e.g., 0 - 100m); Time decay weight : The closer to the event occurrence time, the higher the priority.

[0043] The comprehensive priority calculation formula is as follows: ; Where: , a higher level represents a more urgent event, is the depth weighting coefficient, with priority given to shallow layer observations, is the trigger event timestamp, is the current system time, is the weight coefficient, default , is the maximum allowable response delay, default 30 minutes.

[0044] The queue scheduler performs dynamic sorting based on the value calculated in real time, and always places the instruction with the highest priority at the head of the queue.

[0045] 3. Communication window management strategy: Since the underwater acoustic communication link is relied on for instruction transmission between the submersible and the buoy, its communication has strong intermittency and environmental instability. Therefore, the scheduler introduces a communication window mechanism to ensure that instructions are transmitted only when the following conditions are met simultaneously: The buoy system is in the communication receiving window opening period; There are no other high-priority communication tasks currently; The previous instruction has been confirmed received or the timeout retransmission has been completed; The current instruction status is "pending transmission".

[0046] The instruction scheduling logic automatically checks whether the head instruction of the queue meets the above conditions every time the communication window arrives. If it does, the instruction is packaged into an acoustic signal frame and sent after being encoded and modulated by the underwater acoustic communication module.

[0047] 4. Instruction fault tolerance and feedback control mechanism: Considering that communication in the ocean environment is vulnerable to interference, the instruction scheduling logic integrates a feedback confirmation and retransmission mechanism. The process is as follows: Each issued instruction requires the buoy system to feedback a reception confirmation frame. If the confirmation is not received within the maximum waiting time (such as 30 seconds), the system will mark this instruction as "waiting for retransmission"; The maximum number of retransmission attempts is 3. After exceeding, it is marked as "failed"; The instruction status of the successfully received confirmation is updated to "sent" and dequeued from the cache queue.

[0048] In addition, for the convenience of shore-based tracking, the scheduling system maintains an instruction scheduling log, recording the generation time, sending time, confirmation time and response status of each instruction to ensure the full process traceability of task execution.

[0049] 5. Example illustration Scenario: The submersible buoy detects a "front crossing" event at 75 meters and generates a trigger instruction: Number: T20250607-001; Content: Phenomenon = front crossing, depth = 75m, timestamp = 11:00:00; Urgency level = 3, depth adaptability = 0.9, 5 minutes from the current time; Calculated: = 1.937; When the communication window is open, the system preferentially selects this instruction for issuance. The buoy system confirms reception within 30 seconds, and the instruction status is updated to "sent", and the scheduling is completed.

[0050] The buoy dynamic response module includes: 1. Instruction reception and parsing: Receive the structured trigger instruction from the submersible buoy system through the underwater acoustic communication receiver on the buoy. After the reception is completed, the system immediately writes the instruction content into the local execution task buffer and records the system reception time for feedback verification; 2. Observation Mode Matching: Based on the phenomenon type provided in the trigger instruction, the system retrieves the matching mode configuration file from the locally configured "Observation Mode Database". The configuration items include: Observation sensor combination (such as CTD + Turbulence Profiler + Bio - optical Sensor); Sampling frequency of each type of sensor (such as 2Hz, 5Hz, etc.); Sampling duration (such as 5 minutes, 10 minutes); Whether to enable the redundant backup sampling mechanism (such as high - power consumption dual - redundant channels); For example, when the phenomenon type is frontal crossing, the system matches the following configuration: Sensor type Enabled Sampling frequency Data granularity Conductivity, Temperature, Depth (CTD) Yes 2 Hz 0.1 m Turbulence profiler Yes 10 Hz Instantaneous shear Bio - optical sensor Yes 1 Hz Chlorophyll, turbidity ADCP (auxiliary) No — — The system loads the matched parameters into the acquisition plan controller and waits to be triggered for execution after the lifting is completed; 3. Closed - loop Control of the Lifting Mechanism: After receiving the target depth, the internal lifting mechanism of the buoy (such as a screw - propelled heave pod or a guide - rail counterweight system) starts and performs depth positioning control. To ensure accuracy, the following closed - loop control algorithm is adopted; Obtain the current depth in real - time ; Set the target depth ; Control increment Calculate according to the following formula: ; where, is the proportional - derivative control gain coefficient, is the adjustment action at the current moment; The control period is 1 second, and the execution upper - limit speed is 0.5m / s.

[0051] The system continuously feeds back the deviation between the current position and the target position until it meets: , where is the depth allowable error (default is ±0.3 meters).

[0052] 4. Multi - sensor Synchronous Acquisition: Once the target depth is reached and the steady - state condition is satisfied (speed is close to zero, depth error remains within the threshold for more than 5 seconds), the system immediately starts the matched observation mode for multi - sensor synchronous acquisition. The process includes: Start the synchronous scheduler and send the acquisition start instruction to all sensors; The acquisition process uses a unified clock, and all sensor data are bound with high - precision timestamps; Each type of sensor independently buffers and stores data, and the sampling process lasts for ≥5 minutes; If instantaneous drift or data loss occurs during data acquisition, the internal resampling mechanism is immediately activated or redundant sensor channels are enabled.

[0053] During the sampling process, environmental background information such as water body fluctuations and buoy attitude changes (metadata) is recorded simultaneously for later data calibration and fusion processing.

[0054] 5. Observation data encapsulation and transmission back: After the observation task is completed, the system encapsulates the output data streams of each sensor to form a three-dimensional observation data packet with a standard structure, as follows: Response task number (corresponding to Trigger ID); Response time period (actual start and end times of sampling); Observation depth coordinates (actual depth mean and fluctuation range); Data fields of each sensor (arranged in time series); Embedded sensor metadata and calibration parameters; Sampling status (whether it is complete, whether redundant supplementary sampling is triggered); This data packet will be transmitted back to the shore-based analysis platform through the surface communication module (such as satellite communication, LTE, or LoRa), and simultaneously written to local storage as a backup for fault retransmission.

[0055] Example illustration: Scene example: The moored buoy identifies the "frontal crossing" phenomenon at a depth of 75 meters, with a timestamp of 2025-06-07 11:00:00; After receiving the command, the floating buoy activates the lifting mechanism and reaches 74.9 meters within 1 minute; The CTD, turbulence profiler, and bio-optical sensor are started for synchronous acquisition and continue for 5 minutes; After the observation data is encapsulated, a data packet DP20250607-001 is formed, with a size of 8 MB; It is transmitted upstream through the 4G module with an average delay of 12 seconds.

[0056] The spatio-temporal alignment and fusion module includes: 1. Time alignment: Used to solve the problem of inconsistent time baselines between the moored buoy and floating buoy systems due to factors such as clock drift and transmission delay. The main steps are as follows: (1) Timestamp extraction and comparison: Extract the start and end sampling timestamps from the three-dimensional observation data packet generated by the floating buoy , and compare them with the records in the moored buoy observation dataset that overlap with this time period to extract the corresponding moored buoy time period .

[0057] (2) System clock difference estimation: The maximum cross - correlation position is extracted by using a sliding window to compare synchronous events (such as the temperature peak of the co - measured depth points, etc.) to estimate the clock offset between the two systems. ; ; where respectively represent the temperature sequences collected by the two systems.

[0058] (3) Clock correction and calibration output: Unified correction is performed on the timestamps of all buoy data: ; After calibration, it is required that seconds to ensure the sampling - level alignment accuracy.

[0059] 2. Spatial registration: Buoy response data is often collected at specific depths, while the mooring system covers multiple fixed depth layers. Therefore, it is necessary to construct a spatial position mapping relationship; (1) Depth registration: Read the actual depth coordinates in the buoy observation data , and select the closest depth layer from the mooring dataset , satisfying: ; If the buoy depth is between two mooring depths, it is marked as the state to be interpolated.

[0060] (2) Spatial window definition: For each buoy sampling record, define a spatial registration window: Horizontal position error tolerance ; Vertical depth error tolerance ; Only when the buoy and mooring records simultaneously meet this spatial tolerance condition can they be included in the effective fusion window.

[0061] 3. 3D interpolation calculation: It is used to construct a continuous spatial distribution field, fuse mooring and buoy data into a unified 3D grid data structure, and the Kriging interpolation algorithm is used for value estimation; (1) Construction of spatial sampling point set: Let the current fusion time slice be , collect all mooring and buoy observation points within ±1 minute before and after this moment, and construct a point set: ; where is the spatial coordinate, is the variable to be interpolated (such as temperature).

[0062] (2) Kriging estimation formula: For any target grid point , the estimated value The calculation formula is: , where the weight coefficient satisfies: ; is the empirical semivariogram, and the spherical model is adopted, which is expressed as: ; where is the sill, C is the sill increment value, is the range of influence.

[0063] (3) Interpolation variable type: The interpolation variables include temperature, salinity, eastward / northward / vertical flow velocity, and dissolved oxygen concentration. Interpolation operations are performed on each variable separately to construct the corresponding three-dimensional physical field.

[0064] 4. Construction and output of the fused dataset: The final fused result will be organized into a spatio-temporal synchronized three-dimensional dataset with a unified structure, including the following fields: Timestamp, three-dimensional grid structure, physical parameter vector at each grid point, source label (indicating that the data comes from a mooring buoy, a floating buoy or interpolation estimation), interpolation error estimation matrix (supporting error control and post-processing analysis); This dataset will be output to: The shore-based analysis platform database (for real-time observation visualization and trend analysis); The local redundant storage unit (for continuous data transfer in case of disconnection); The data fusion log system (recording the interpolation path and algorithm parameters) Example illustration: The floating buoy collected three-dimensional data at 76.2 m at 11:05:00 on June 7, 2025. The mooring buoy provided data at two depth layers of 70 m and 80 m during the same period. The system operations are as follows: Perform time alignment and measure = -0.8 seconds and perform unified calibration; Locate the depth range of 75.5 - 76.5 m and identify it as the position to be interpolated; Extract a total of 48 sets of mooring / floating buoy sampling points within 2 minutes before and after; Apply Kriging interpolation to construct the temperature field, salinity field, etc. under this time slice; Organize the results into a three-dimensional data cube and output, including an interpolation error estimate of ±0.13 °C.

[0065] The present invention covers any alternatives, modifications, equivalent methods, and solutions made within the spirit and scope of the present invention. For the public to have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention, and those skilled in the art can fully understand the present invention without the description of these details. Additionally, to avoid unnecessary confusion with the essence of the present invention, well-known methods, processes, procedures, components, and circuits are not described in detail.

[0066] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A three-dimensional collaborative observation system for the exchange area sea area based on submerged and floating buoys, characterized in that, It includes a data acquisition module, an instruction generation module, a buoy dynamic response module, and a spatio-temporal alignment and fusion module, where; The data acquisition module collects the original observation data stream of a preset depth layer through a mooring system, divides the original observation data stream into data blocks according to a time window, extracts the feature vectors of each data block, and generates a mooring observation data set; The instruction generation module inputs the mooring observation data set into a preset anomaly detection model. When the characteristics of the target ocean phenomenon are recognized, it generates a trigger instruction including the phenomenon type, occurrence depth, and timestamp; The buoy dynamic response module transmits the trigger instruction to the buoy system through an underwater acoustic communication link. The buoy matches the preset observation mode according to the phenomenon type in the trigger instruction, drives the lifting mechanism to position to the target depth layer, and starts multi-sensor synchronous acquisition to form a three-dimensional observation data packet; The spatio-temporal alignment and fusion module performs time calibration on the three-dimensional observation data packet based on the timestamp in the trigger instruction, and performs three-dimensional interpolation in combination with the spatial coordinates of the mooring observation data set to generate a spatio-temporally synchronized three-dimensional data set.

2. The three-dimensional collaborative observation system for the exchange area sea area based on submerged and floating buoys according to claim 1, wherein The data acquisition module includes: Using a conductivity-temperature-depth profiler, an acoustic Doppler current profiler, and a dissolved oxygen sensor carried by the mooring system to continuously collect the original observation data stream at multiple preset depth layers. The original observation data stream includes: The temperature value, salinity value, and depth value output by the conductivity-temperature-depth profiler; The eastward velocity, northward velocity, and vertical velocity output by the acoustic Doppler current profiler; The dissolved oxygen concentration and dissolved oxygen saturation output by the dissolved oxygen sensor; The timestamp of synchronous acquisition by all sensors.

3. The three-dimensional collaborative observation system for the exchange area sea area based on submerged and floating buoys according to claim 2, characterized in that The data acquisition module also includes: Dividing the original observation data stream into continuous data blocks according to a time window with a fixed duration of 30 minutes; Extracting time-domain statistical features and frequency-domain features for each data block respectively. The time-domain statistical features include mean, variance, and extreme values, and the frequency-domain features are energy features calculated based on wavelet transform; Integrating the time-domain statistical features and frequency-domain features to form a feature vector, and binding it to the depth layer number and acquisition start and end timestamps of the corresponding data block to construct a structured mooring observation data set; Real-time storing the mooring observation data set through the underwater data storage unit of the mooring system and synchronously sending it to the instruction generation module.

4. The three-dimensional collaborative observation system for the exchange area sea area based on submerged and floating buoys according to claim 3, characterized in that, The instruction generation module includes: Receiving the structured mooring observation data set and inputting it into a preset anomaly detection model. The anomaly detection model outputs the type identifier of the target ocean phenomenon, the corresponding depth layer number, the corresponding confidence score, and the acquisition start and end timestamps of the data block to which the corresponding feature vector belongs; When the confidence score exceeds the preset threshold of 0.85, based on the type identifier, depth layer number, and acquisition start and end timestamps output by the model, generating a trigger instruction, and the trigger instruction includes the phenomenon type, occurrence depth, and timestamp; Storing the trigger instruction in the instruction cache queue and preparing to send it to the buoy system through the underwater acoustic communication link.

5. The three-dimensional collaborative observation system for the exchange area sea area based on submerged and floating buoys according to claim 4, wherein, The buoy dynamic response module includes: Receiving the trigger instruction from the instruction generation module through the underwater acoustic communication link, and parsing to obtain the phenomenon type, occurrence depth, and timestamp therein; Match the preset observation modes in the buoy system according to the type of the phenomenon, where the observation modes include different sensor combination schemes and sampling frequencies; Drive the lifting mechanism of the buoy to perform target depth layer positioning according to the occurrence depth. During the positioning process, the current depth information is fed back in real time and closed-loop correction control is performed to ensure that the deviation does not exceed the set error range.

6. The three-dimensional collaborative observation system for the exchange area sea area based on submerged and floating buoys according to claim 5, characterized in that, After the depth positioning is completed, the buoy dynamic response module starts the following multi-sensor synchronous acquisition process according to the matched observation mode: Start the CTD profiler to obtain the temperature, salinity and depth data of the current layer; Start the turbulence profiler to obtain the turbulence energy distribution and shear rate data; Start the bio-optical sensor to obtain the particle scattering intensity and chlorophyll fluorescence parameters; Continuously sample at the target depth layer for no less than 5 minutes, and bind the sampling data with the corresponding timestamps and depth layer coordinates to generate a three-dimensional observation data packet.

7. The three-dimensional collaborative observation system for the exchange area sea area based on submerged and floating buoys according to claim 6, characterized in that, The buoy system transmits the generated three-dimensional observation data packet back to the shore-based center in real time through its waterborne wireless transmission unit, and retains a local cache for the fault retransmission mechanism.

8. The three-dimensional collaborative observation system for the exchange area sea area based on submerged and floating buoys according to claim 7, wherein The space-time alignment and fusion module includes: Receive the three-dimensional observation data packet and extract its timestamps and depth layer coordinates; Perform time calibration based on the timestamps to correct the clock deviation between the buoy system and the mooring system to within ±0.1 second; Call the observation records in the mooring observation dataset corresponding to the time window and depth layer for spatial position matching.

9. The three-dimensional collaborative observation system for the exchange area sea area based on submerged and floating buoys according to claim 8, characterized in that, The space-time alignment and fusion module further includes: After the spatial position matching is completed, use the Kriging interpolation algorithm to perform three-dimensional interpolation operations on the same type of observation parameters in the mooring observation dataset and the three-dimensional observation data packet to generate a three-dimensional grid data volume of temperature, salinity, flow velocity and dissolved oxygen concentration; Combine the interpolation result with the time calibration result, output a space-time synchronized three-dimensional dataset including the temperature field, salinity field, flow velocity field and dissolved oxygen concentration field, and store and synchronously transmit the three-dimensional dataset to the shore-based analysis platform.

Images