A three-dimensional collaborative observation system for the sea area in the exchange zone based on submarines and buoys

Abstract

The present invention relates to the field of marine environmental monitoring technology, and specifically to a three-dimensional collaborative observation system for sea areas in an exchange zone based on submarines and buoys, comprising a data acquisition module, an instruction generation module, a buoy dynamic response module, and a spatiotemporal alignment and fusion module. The system collects raw water body observation data at multiple depth layers through the submarine buoy system to construct a structured feature data set; identifies marine anomalies and generates trigger instructions based on an anomaly detection model; the buoy system analyzes the instructions and drives the lifting mechanism to accurately position itself at the target depth layer, performing multi-sensor synchronous acquisition according to the matching observation mode; the fusion module uses time calibration and Kriging interpolation algorithms to achieve three-dimensional data reconstruction, and outputs a spatiotemporal synchronized three-dimensional data set. The present invention has the characteristics of intelligent recognition, dynamic response, and multi-parameter fusion, which improves the capture accuracy and observation efficiency of complex processes in the exchange zone, and provides technical support for the study of marine dynamic processes and ecological monitoring.

Description

Technical Field

[0001] The present invention relates to the technical field of marine environment monitoring, and in particular to a three-dimensional collaborative observation system for sea areas in an exchange zone based on submarines and buoys. Background Art

[0002] Against the backdrop of climate change and the frequent evolution of regional ocean dynamics, exchange zones (such as shelf edges and gyre convergence zones) serve as crucial nodes for material transport and energy conversion. Their water structure and dynamics have significant impacts on marine ecosystems, carbon cycles, and even disaster warning. Because exchange zones typically exhibit significant water stratification, intense flow disturbances, and the coexistence of multi-scale phenomena, there is an urgent need for high-resolution, stereoscopic observation data on a vertical scale to capture the spatial structure and evolutionary dynamics of key processes such as frontoclines, mesoscale eddies, and internal wave excitation. In recent years, submersibles and buoys, as complementary deep and shallow ocean observation platforms, have demonstrated tremendous potential for regional collaborative monitoring and have become an important technical path for the development of high-precision three-dimensional observation systems.

[0003] However, existing submersible and buoy observation methods are mostly based on parallel and independent deployment, lacking an effective information linkage mechanism. On the one hand, although submersible buoys can be fixed at a specific depth for a long time to conduct high-frequency monitoring, they have difficulty responding to complex abnormal events in real time. On the other hand, although buoys have mobility and controllable lifting and lowering capabilities, most sampling is still carried out at a fixed frequency, lacking the ability to actively trigger and command-driven key deep-seated phenomena. In addition, the existing buoy response system lacks precise time alignment and spatial fusion mechanisms, making it difficult to construct a continuous three-dimensional observation field in a unified grid based on submersible data, limiting the comprehensive utilization and scientific interpretation of the data. Summary of the Invention

[0004] The present invention provides a three-dimensional collaborative observation system for the exchange zone sea area based on submarines and buoys, constructing a three-dimensional collaborative observation system that integrates anomaly detection, response regulation and multi-source data fusion to improve the deep-seated anomaly recognition capability and three-dimensional process reconstruction accuracy of the exchange zone sea area.

[0005] A three-dimensional collaborative observation system for the sea area in the exchange zone based on submarines and buoys, including a data acquisition module, a command generation module, a buoy dynamic response module, and a time-space alignment and fusion module, wherein;

[0006] The data acquisition module collects the original observation data stream of the preset depth layer through the buoy system, divides the original observation data stream into data blocks according to the time window, extracts the feature vector of each data block and generates the buoy observation data set;

[0007] The instruction generation module inputs the buoy observation data set into a preset anomaly detection model, and if the target ocean phenomenon characteristics are identified, generates a trigger instruction including the phenomenon type, occurrence depth and timestamp;

[0008] The buoy dynamic response module transmits the trigger command 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 command, drives the lifting mechanism to locate to the target depth layer, and starts multi-sensor synchronous acquisition to form a three-dimensional observation data packet.

[0009] The spatiotemporal alignment and fusion module performs time calibration on the stereo observation data packet based on the timestamp in the trigger instruction, performs three-dimensional interpolation based on the spatial coordinates of the latent buoy observation data set, and generates a spatiotemporally synchronized stereo data set.

[0010] Optionally, the data acquisition module includes:

[0011] The temperature, salinity, depth, and acoustic Doppler current profiler and dissolved oxygen sensor carried by the buoy system continuously collect raw observation data streams at multiple preset depth layers. The raw observation data streams include:

[0012] Temperature, salinity and depth values ​​output by the temperature, salinity and depth instrument;

[0013] Eastward, northward, and vertical current velocities output by the acoustic Doppler current profiler;

[0014] Dissolved oxygen concentration and dissolved oxygen saturation output by the dissolved oxygen sensor;

[0015] Timestamps collected synchronously by all sensors.

[0016] Optionally, the data acquisition module further includes:

[0017] The original observation data stream is divided into time windows of fixed length of 30 minutes to generate continuous data blocks;

[0018] For each data block, time domain statistical features and frequency domain features are extracted respectively. Time domain statistical features include mean, variance and extreme value, and frequency domain features are energy features calculated based on wavelet transform.

[0019] The time domain statistical features and frequency domain features are integrated to form a feature vector, which is then bound to the depth layer number and acquisition start and end timestamps of the corresponding data block to construct a structured buoy observation dataset.

[0020] The submerged buoy observation data set is stored in real time by the underwater data storage unit of the submerged buoy system and is synchronously sent to the instruction generation module.

[0021] Optionally, the instruction generation module includes:

[0022] Receive a structured buoy observation dataset and input it into a preset anomaly detection model, the output of which includes the type identification of the target ocean phenomenon, the corresponding depth layer number, the corresponding confidence score, and the start and end timestamps of the data block to which the corresponding feature vector belongs;

[0023] If the confidence score exceeds a preset threshold of 0.85, a trigger instruction is generated based on the type identification, depth layer number, and acquisition start and end timestamps output by the model. The trigger instruction includes the phenomenon type, occurrence depth, and timestamp;

[0024] The trigger instruction is stored in the instruction cache queue and is ready to be sent to the buoy system through the underwater acoustic communication link.

[0025] Optionally, the buoy dynamic response module includes:

[0026] Receive the trigger command from the command generation module through the underwater acoustic communication link, and parse and obtain the phenomenon type, occurrence depth and timestamp;

[0027] Matching a preset observation mode in the buoy system according to the phenomenon type, wherein the observation mode includes different sensor combination schemes and sampling frequencies;

[0028] The lifting mechanism of the driving buoy locates the target depth layer 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.

[0029] Optionally, after completing the depth positioning, the buoy dynamic response module starts the following multi-sensor synchronous acquisition process according to the matched observation mode:

[0030] Start the temperature, salinity and depth instrument to obtain the temperature, salinity and depth data of the current layer;

[0031] Start the turbulence profiler to obtain turbulent energy distribution and shear rate data;

[0032] Start the bio-optical sensor to obtain particle scattering intensity and chlorophyll fluorescence parameters;

[0033] Continuously sample the target depth layer for no less than 5 minutes, and bind the sampled data with the corresponding timestamp and depth layer coordinates to generate a stereo observation data package.

[0034] Optionally, the buoy system transmits the generated stereoscopic observation data packets back to the shore-based center in real time through its water wireless transmission unit, and retains a local cache for fault recovery mechanism.

[0035] Optionally, the spatiotemporal alignment and fusion module includes:

[0036] Receive the stereo observation data packet and extract its timestamp and depth layer coordinates;

[0037] Perform time calibration based on timestamps to correct the clock deviation between the buoy system and the submerged buoy system to within ±0.1 seconds;

[0038] The observation records of the corresponding time window and depth layer in the buoy observation dataset are called to perform spatial position matching.

[0039] Optionally, the spatiotemporal alignment and fusion module further includes:

[0040] After completing the spatial position matching, the Kriging interpolation algorithm is used to perform three-dimensional interpolation operations on the same observation parameters of the buoy observation data set and the stereo observation data set to generate a three-dimensional grid data volume of temperature, salinity, flow velocity and dissolved oxygen concentration;

[0041] The interpolation results are combined with the time calibration results to output a temporally and spatially synchronized three-dimensional dataset including temperature field, salinity field, velocity field and dissolved oxygen concentration field. The three-dimensional dataset is then stored and synchronously transmitted to a shore-based analysis platform.

[0042] Beneficial effects of the present invention:

[0043] This invention, by introducing a "time window + multi-feature fusion" buoy observation modeling approach into the data acquisition module, combined with a deep anomaly detection model and confidence calculation mechanism in the command generation module, enables in-depth, high-confidence, real-time identification of complex target phenomena in the ocean, such as fronts, vortices, and hybrid jumps. After the response command is generated, the buoy's dynamic response module automatically matches the observation mode based on the phenomenon type and rapidly locates the target layer through a deep closed-loop control algorithm, achieving a response delay of less than 2 minutes, ensuring the spatiotemporal linkage observation requirements of sudden phenomena.

[0044] This invention utilizes a multi-sensor collaborative acquisition mechanism deployed by the buoy system to acquire high-frequency parameters from multiple sources, including temperature, salinity, current velocity, dissolved oxygen, turbulence, and bio-optics, in real time when observation is triggered. Furthermore, the spatiotemporal alignment and fusion module, after acquisition, incorporates Kriging interpolation to reconstruct the buoy's historical data and buoy response data into a three-dimensional grid at a unified time and depth. This generates high-resolution, highly consistent temperature, salinity, current velocity, and dissolved oxygen fields, providing high-quality, raw support for multiscale dynamic process research, ecological anomaly diagnosis, and data assimilation.

[0045] This invention utilizes a modular architecture, integrating four subsystems: data acquisition, command generation, response execution, and data fusion. This system, supplemented by FIFO command scheduling, confidence threshold control, closed-loop correction of lifting mechanism errors, interpolation error control, and feedback confirmation mechanisms, forms a complete observation closed loop. The system boasts engineering capabilities for task tracking, parameter configuration, and anomaly recovery. It is suitable for scalable deployment and operation across multiple snorkeling platforms, multi-layer response strategies, and complex exchange zone environments, significantly improving the practicality and stability of the observation system. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0047] Figure 1 A schematic diagram of a system flow diagram of an embodiment of the present invention;

[0048] Figure 2 This is a flow chart of the buoy dynamic response module according to an embodiment of the present invention. DETAILED DESCRIPTION

[0049] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments. It is also noted that, to provide a more detailed description, the following embodiments are best and preferred embodiments, and those skilled in the art may employ alternative methods for implementing certain known technologies. Furthermore, the accompanying drawings are intended only to provide a more detailed description of the embodiments and are not intended to limit the present invention.

[0050] like Figure 1-Figure 2 As shown, a three-dimensional collaborative observation system for the sea area in the exchange zone based on submarines and buoys includes a data acquisition module, an instruction generation module, a buoy dynamic response module, and a time-space alignment and fusion module, wherein;

[0051] The data acquisition module collects the original observation data stream of the preset depth layer through the buoy system, divides the original observation data stream into data blocks according to the time window, extracts the feature vector of each data block and generates the buoy observation data set;

[0052] The command generation module inputs the buoy observation data set into the preset anomaly detection model. If the target ocean phenomenon characteristics are identified, a trigger command including the phenomenon type, occurrence depth and timestamp is generated.

[0053] The buoy's dynamic response module transmits the trigger command to the buoy system via the underwater acoustic communication link. The buoy matches the preset observation mode according to the phenomenon type in the trigger command, drives the lifting mechanism to locate to the target depth layer, and starts multi-sensor synchronous acquisition to form a three-dimensional observation data packet.

[0054] The spatiotemporal alignment and fusion module performs time calibration on the stereo observation data packets based on the timestamp in the trigger instruction, and performs three-dimensional interpolation based on the spatial coordinates of the buoy observation data set to generate a spatiotemporally synchronized stereo data set.

[0055] The data acquisition module includes:

[0056] 1. Multi-sensor collaborative acquisition unit: The data acquisition module integrates the following ocean environment monitoring sensors into the buoy system to achieve real-time acquisition of raw observation data:

[0057] Temperature-salinity-depth meter (CTD): used to measure the temperature (T), salinity (S) and depth (D) of water bodies;

[0058] Acoustic Doppler Current Profiler (ADCP): obtains three-component velocity data at each measuring point on the profile, namely the eastward velocity u, northward velocity v, and vertical velocity w;

[0059] Dissolved oxygen sensor (DO Sensor): provides dissolved oxygen concentration DO and dissolved oxygen saturation DOS in water;

[0060] System clock module: ensures that all sensor output data are accompanied by a precise and unified timestamp t.

[0061] The above-mentioned sensors sample synchronously at preset time intervals (the default sampling period is 10 seconds) and form a raw observation data sequence in the following format:

[0062] ;

[0063] Each set of observation data points Each is bound to a unique timestamp to ensure the temporal consistency of multi-parameter data.

[0064] 2. Time window segmented response mechanism, the collected data is segmented and processed according to a fixed time window. The default time window is 30 minutes processing unit: to support subsequent anomaly detection and, it means that every 30 minutes of observation data is classified into a data block. Suppose the buoy observation period is , can generate Data blocks:

[0065] , ;

[0066] Each data block Corresponding to a collection depth layer number and time range .

[0067] in, The result is rounded down.

[0068] 3. Feature extraction and vector construction unit To achieve efficient anomaly detection, the system performs Extract time domain and frequency domain features to form feature vectors. The specific steps are as follows:

[0069] (1) Time domain statistical feature extraction: For each physical quantity , extract the following statistical features:

[0070] Mean: ;

[0071] variance: ;

[0072] extremum: , ;

[0073] (2) Frequency domain wavelet energy feature extraction: is the number of data points in the data block, usually 180 (30 minutes × 60 seconds ÷ 10 seconds / sampling point). Perform discrete wavelet transform (DWT) and extract the wavelet subband energy as frequency domain features. Taking the Daubechies-4 wavelet basis as an example, decompose the signal into 3 layers and calculate the energy distribution at each scale:

[0074] ;

[0075] in Represents the i-th coefficient after the j-th layer wavelet decomposition.

[0076] (3) Example of constructing eigenvectors: Taking temperature as an example, its eigenvectors are as follows:

[0077] ;

[0078] Extract the above features for all observed variables separately and concatenate them to obtain a complete feature vector:

[0079] ;

[0080] 4. Dataset structuring and cache unit: Each feature vector All of them are related to the data block number k and the acquisition depth layer number , time range Binding, forming the following structured observation data items:

[0081] ;

[0082] all Composing buoy observation data sets in chronological order :

[0083] ;

[0084] Finally, the data acquisition module stores the buoy observation data set 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 identification and trigger response processing.

[0085] Example of data extraction process at a specific depth layer:

[0086] Taking the depth layer of 75 meters as an example, the entire process of the data acquisition module is as follows:

[0087] Original data collection: The submersible deploys a temperature, salinity, depth, ADCP, and 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 data set contains:

[0088] Temperature (e.g., 12.3°C), salinity (e.g., 33.8 PSU), depth (fixed at 75 meters), three-dimensional current velocity (e.g., 0.42 m / s east, -0.17 m / s north, 0.03 m / s vertically), dissolved oxygen concentration (e.g., 6.2 mg / L) and saturation (e.g., 92.4%), and timestamp (e.g., 2025-06-07 10:00:00).

[0089] Data blocking: The above 180 groups of data are grouped into a data block, whose corresponding time window is "June 7, 2025 10:00:00–10:29:59", numbered B1, and depth marked as Z=75.

[0090] Feature extraction: Extract the following statistical and frequency domain features for each physical quantity in the data block:

[0091] The average value of the temperature signal in the 30 minutes is 12.4℃, the standard deviation is 0.08, and the extreme values ​​are [12.2℃, 12.6℃];

[0092] The temperature series is analyzed by wavelet transform to obtain the energy distribution of wavelets at each scale, which reflects the degree of periodic oscillation.

[0093] Similarly, variables such as salinity, various components of flow velocity, and dissolved oxygen are processed to obtain characteristic vector fragments of multiple variables.

[0094] Feature vector assembly: The feature fragments of all variables are spliced ​​into a complete feature vector with a total length of approximately 60 to 80 dimensions, which serves as the representation result of the depth layer in the time period.

[0095] Dataset generation: Combine the feature vector with the label "Z=75" and "start and end timestamps = 10:00:00–10:29:59" into a structured data item;

[0096] This item is then stored in the local cache of the buoy and transmitted to the instruction generation module to be used as a basis for determining whether to trigger a buoy response.

[0097] The instruction generation module includes:

[0098] 1. Buoy observation data reception and buffering: This module first receives the structured buoy observation data set transmitted from the data acquisition module. Each record in the data set includes:

[0099] Eigenvector : Reflects the multi-parameter statistics and frequency domain characteristics within a single time window (such as 30 minutes);

[0100] Depth layer number : Indicates the sampling depth of the latent marker to which the data belongs;

[0101] Time Range : The collection time period of the data block;

[0102] All received data first enters the input buffer for integrity checking and timing reordering to ensure that the input data is continuous, without packet loss, and in consistent order.

[0103] 2. Anomaly detection model and recognition logic: Each feature vector received The data will be fed into the preset anomaly detection model. This model is a semi-supervised neural network structure that uses a variational autoencoder as an anomaly detector and uses a posterior probability scoring function to determine the degree of anomaly.

[0104] (1) Model structure description:

[0105] The encoder transforms the feature vector Mapping to latent space distribution , and reconstruct the original input;

[0106] The decoder samples the latent space to generate a reconstruction vector ,Calculate the reconstruction error;

[0107] The degree of abnormality is determined by using the dual indicators of reconstruction error and distribution deviation:

[0108] ;

[0109] in, is the weighting coefficient, is the Kullback-Leibler divergence measure.

[0110] (2) Abnormal confidence calculation: The loss function Mapping to normalized confidence score , indicating the suspicious degree of the ocean phenomenon corresponding to the current data block:

[0111] ;

[0112] in is the minimum and maximum error within the sliding window, used for normalization.

[0113] 3. Confidence determination and event classification: When confidence scoring When it exceeds the preset threshold (0.85), the system marks the data block as a "potential target ocean phenomenon" and further identifies the event type.

[0114] (1) A lightweight multi-layer perceptron (MLP) is used to classify abnormal samples and output corresponding phenomenon type labels (such as mesoscale vortex, frontal crossing, mixed jump layer). The phenomenon type is recorded as ;

[0115] (2) Event judgment conditions: When the following conditions are met, the response process is triggered:

[0116] (Confidence determination threshold, default ;

[0117] Classifier output It is not empty and has a clear physical meaning;

[0118] The buoy is within the main observation water layer range (such as 50-150 meters).

[0119] 4. Trigger instruction generation mechanism: After determining the presence of an abnormal ocean phenomenon, the system immediately constructs a trigger instruction. Each trigger instruction includes the phenomenon type, occurrence depth and timestamp, and the complete instruction structure is as follows:

[0120] , is the timestamp;

[0121] The trigger instruction will be stored in the instruction cache queue in the form of a structured message, and will be accompanied by a unique identification code and a sending status tag (such as "to be sent").

[0122] Example description:

[0123] The following is an actual workflow diagram of the instruction generation module:

[0124] Input: The buoy system collects data for 30 minutes at a depth of 75 meters. The feature vector Extraction completed;

[0125] Model output: Anomaly detection model confidence level =0.91, phenomenon type For the front to pass through;

[0126] Generate instructions: Build instructions = ⟨frontal crossing, 75m, 2025-06-07 11:00:00>;

[0127] Output status: The command is sent to the cache queue and marked as "waiting for voice transmission" state, waiting for the opportunity of downlink voice communication.

[0128] Instruction scheduling logic: It consists of four parts: cache queue mechanism, scheduling priority rules, communication window management strategy and fault-tolerant return control mechanism.

[0129] 1. Instruction cache queue mechanism: All trigger instructions output by the anomaly detection and identification process are uniformly entered into the instruction cache queue. Each instruction is stored in a structured format, including: instruction number (unique identifier), instruction content (phenomenon type, depth, timestamp), entry time, and instruction status (pending, sent, awaiting confirmation, pending retransmission);

[0130] The system uses a first-in-first-out (FIFO) mechanism for queue management to ensure that the first-generated instructions are scheduled first. For multiple instructions generated simultaneously, queue adjustment can be made according to the priority rules (see below).

[0131] 2. Command scheduling priority rules: In order to adapt to the response time requirements of different types of ocean phenomena, the system assigns a dynamic priority identifier to each command. The calculation takes into account the following factors:

[0132] Urgency of the phenomenon : Determined by the event type (e.g., front > turbulent mixing > mesoscale eddy);

[0133] Deep adaptation occurs : Weighted near the observable area of ​​the buoy (e.g. 0-100m);

[0134] Time decay weight : The closer the time to the event occurs, the higher the priority.

[0135] The comprehensive priority calculation formula is as follows:

[0136] ;

[0137] in: , higher levels represent more urgent events, is the depth weighting coefficient, shallow observations are given priority, is the trigger event timestamp, is the current system time, is the weight coefficient, the default , To allow for maximum response delay, the default is 30 minutes.

[0138] The queue scheduler calculates the The values ​​are dynamically sorted, always placing the highest priority instruction at the head of the queue.

[0139] 3. Communication Window Management Strategy: Because the submarine and buoy rely on underwater acoustic communication links to issue commands, which are highly intermittent and subject to environmental instability, the scheduler introduces a communication window mechanism to ensure that commands are issued only when the following conditions are met simultaneously:

[0140] The buoy system is in the open communication reception window period;

[0141] There are currently no other high-priority communication tasks;

[0142] The previous command has been confirmed to be received or has timed out and resent;

[0143] The current command status is "pending".

[0144] The command scheduling logic automatically checks whether the head-of-team command meets the above conditions each time a communication window arrives. If so, the command is packaged into an acoustic signal frame and coded, modulated, and sent through the underwater acoustic communication module.

[0145] 4. Command fault tolerance and feedback control mechanism: Considering that communications in marine environments are susceptible to interference, the command scheduling logic integrates a feedback confirmation and retransmission mechanism. The process is as follows:

[0146] Each command sent requires the buoy system to send back a reception confirmation frame. If the maximum waiting time is If no confirmation is received within (e.g. 30 seconds), the system will mark the instruction as "waiting for resend";

[0147] The maximum number of resend attempts is 3, after which the message will be marked as "failed";

[0148] The status of the instruction that successfully receives the confirmation is updated to "sent" and is dequeued from the cache queue.

[0149] In addition, to facilitate shore-based tracking, the dispatching system maintains a command dispatch log, recording the generation time, sending time, confirmation time and response status of each command to ensure that the task execution is traceable throughout the entire process.

[0150] 5. Example

[0151] Scenario: The buoy detects a "front crossing" event at 75 meters and generates a trigger command:

[0152] No.: T20250607-001;

[0153] Content: Phenomenon = Frontal Crossing, Depth = 75m, Timestamp = 11:00:00;

[0154] Urgency = 3, Depth Adaptability = 0.9, 5 minutes from current time;

[0155] Calculation yields: =1.937;

[0156] When the communication window is open, the system prioritizes this command for dispatch. The buoy system confirms receipt within 30 seconds, and the command status is updated to "Sent," completing the dispatch.

[0157] The buoy dynamic response module includes:

[0158] 1. Command reception and analysis: The underwater acoustic communication receiver on the buoy receives the structured trigger command from the submerged buoy system. After receiving the command, the system immediately writes the command content into the local execution task buffer and records the system reception time for feedback verification.

[0159] 2. Observation pattern matching: The system searches the locally configured "observation pattern database" for a matching pattern configuration file based on the phenomenon type provided in the trigger instruction. The configuration items include:

[0160] Observation sensor combination (such as CTD+turbulence profiler+bio-optical sensor);

[0161] The sampling frequency of each type of sensor (e.g., 2 Hz, 5 Hz, etc.);

[0162] Sampling duration (e.g., 5 minutes, 10 minutes);

[0163] Whether to enable redundant backup sampling mechanism (such as high-energy dual redundant channels);

[0164] For example, when the phenomenon type is front crossing, the system matches the following configuration:

[0165] Sensor Type Enable Sampling frequency Data granularity Temperature, Salinity and Depth (CTD) yes 2 Hz 0.1 m Turbulence Profiler yes 10 Hz Instantaneous shear Bio-optical sensors yes 1 Hz Chlorophyll, turbidity ADCP (auxiliary) no — —

[0166] The system loads the matched parameters into the acquisition plan controller and waits for the lifting to be completed before triggering the execution;

[0167] 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 lifting chamber or a guide rail counterweight system) is activated to perform depth positioning control. To ensure accuracy, the following closed-loop control algorithm is used;

[0168] Get current depth in real time ;

[0169] Set target depth ;

[0170] Control increment Calculated according to the following formula:

[0171] ;

[0172] in, is the proportional-derivative control gain coefficient, is the adjustment action at the current moment;

[0173] The control cycle is 1 second and the upper limit speed is 0.5m / s.

[0174] The system continuously feeds back the deviation between the current position and the target position until:

[0175] ,in The default is ±0.3 meters.

[0176] 4. Multi-sensor synchronous acquisition: Once the target depth is reached and steady-state conditions are met (velocity approaches zero, and depth error remains within the threshold for >5 seconds), the system immediately initiates the matching observation mode and performs multi-sensor synchronous acquisition. The process includes:

[0177] Start the synchronization scheduler and send the acquisition start instruction to all sensors;

[0178] The acquisition process uses a unified clock, and all sensor data are bound to high-precision timestamps;

[0179] Each type of sensor has independent buffer storage, and the sampling process lasts ≥5 minutes;

[0180] If instantaneous drift or dropout occurs during data acquisition, the internal resampling mechanism is immediately activated or redundant sensor channels are enabled.

[0181] During the sampling process, environmental background information, such as water fluctuations, buoy attitude changes and other metadata, is recorded for later data calibration and fusion processing.

[0182] 5. Observation data packaging and transmission: After the observation task is completed, the system packages the output data streams of each sensor to form a stereoscopic observation data packet with a standard structure. The structure is as follows:

[0183] Response task number (corresponding to Trigger ID);

[0184] Response time period (actual sampling start and end time);

[0185] Observed depth coordinates (actual depth mean and fluctuation range);

[0186] Each sensor data field (arranged in time series);

[0187] Embedded sensor metadata and calibration parameters;

[0188] Sampling status (whether it is complete, whether redundant sampling is triggered);

[0189] The data packet will be transmitted back to the shore-based analysis platform via a surface communication module (such as satellite communication, LTE or LoRa) and will be synchronously written to the local storage as a fault recovery backup.

[0190] Example description:

[0191] Example scenario:

[0192] The buoy identified the "frontal crossing" phenomenon at a depth of 75 meters and a timestamp of 2025-06-07 11:00:00.

[0193] After receiving the command, the buoy activated the lifting mechanism and reached 74.9 meters in 1 minute;

[0194] Start synchronous data acquisition of CTD, turbulence profiler and bio-optical sensor for 5 minutes;

[0195] The observation data is encapsulated to form data packet DP20250607-001, which is 8MB in size.

[0196] The average delay for uplink transmission via the 4G module is 12 seconds.

[0197] The spatiotemporal alignment and fusion module includes:

[0198] 1. Time alignment: This is used to resolve the time baseline inconsistency issue between the submerged buoy and the buoy system due to clock drift, transmission delay, and other factors. The main steps are as follows:

[0199] (1) Timestamp extraction and comparison: Extract the sampling start and end timestamps from the stereo observation data packet generated by the buoy , and compare the records in the latent buoy observation data set that overlap with this time period to extract the corresponding latent buoy time period .

[0200] (2) System clock difference estimation: Use a sliding window to compare synchronous events (such as temperature peaks at common depth points, etc.) to extract the maximum cross-correlation position and estimate the clock offset between the two systems. ;

[0201] ;

[0202] in Represent the temperature series collected by the two systems respectively.

[0203] (3) Clock correction and calibration output: Perform a unified correction on all buoy data timestamps:

[0204] ;

[0205] Post-calibration requirements seconds to ensure sample-level alignment accuracy.

[0206] 2. Spatial registration: Buoy response data is often collected at a specific depth, while the submersible system covers multiple fixed depth layers, so a spatial position mapping relationship must be established;

[0207] (1) Depth registration: Read the actual depth coordinates from the buoy observation data , and select the closest depth layer from the latent marker dataset ,satisfy:

[0208] ;

[0209] If the buoy depth is between the depths of two layers of submerged buoys, it is marked as pending interpolation state.

[0210] (2) Spatial window definition: For each buoy sampling record, define the spatial registration window:

[0211] Horizontal position error tolerance ;

[0212] Vertical depth error tolerance ;

[0213] Only when the buoy and submerged buoy records meet this spatial tolerance condition at the same time will they be included in the effective fusion window.

[0214] 3. 3D interpolation calculation: used to construct a continuous spatial distribution field, fuse the buoy and float data into a unified 3D raster data structure, and use the Kriging interpolation algorithm to estimate the value;

[0215] (1) Construction of spatial sampling point set: Let the current fusion time slice be , collect all the buoy and float observation points within ±1 minute before and after the moment, and construct the point set:

[0216] ;

[0217] in is the spatial coordinate, is the variable to be interpolated (such as temperature).

[0218] (2) Kriging estimation formula: For any target grid point , estimated value The calculation formula is:

[0219] , where the weight coefficient satisfy: ;

[0220] is the empirical semivariogram, using a spherical model, expressed as:

[0221] ;in is the base, C is the base rise value, is the influence radius.

[0222] (3) Interpolation variable type: Interpolation variables include temperature, salinity, eastward / northward / vertical flow velocity, and dissolved oxygen concentration. Interpolation operations are performed on each variable to construct the corresponding three-dimensional physical field.

[0223] 4. Construction and output of fusion dataset: The final fusion result will be organized into a unified structure of spatiotemporal synchronized stereo dataset, including the following fields:

[0224] Timestamp, 3D grid structure, physical parameter vector of each grid point, source label (indicating whether the data comes from a buoy, float, or interpolation estimate), interpolation error estimation matrix (supporting error control and post-processing analysis);

[0225] The dataset will be output to:

[0226] Shore-based analysis platform database (for real-time observation visualization and trend analysis);

[0227] Local redundant storage unit (for resuming transmission after link failure);

[0228] Data fusion log system (recording interpolation paths and algorithm parameters)

[0229] Example description:

[0230] The buoy collected 3D data at 76.2m at 11:05:00 on June 7, 2025. The submerged buoy provided data at two depth layers, 70m and 80m, during the same period. The system operates as follows:

[0231] Perform time alignment and measure = -0.8 seconds, perform unified calibration;

[0232] The positioning depth range is 75.5–76.5 m, which is identified as the position requiring interpolation;

[0233] A total of 48 groups of submerged / buoyant sampling points within 2 minutes before and after extraction;

[0234] Apply Kriging interpolation to construct the temperature field, salinity field, etc. under this time slice;

[0235] The results are organized into a three-dimensional data cube and output, including an interpolation error estimate of ±0.13°C.

[0236] The present invention encompasses any alternatives, modifications, equivalents, and solutions that fall within the spirit and scope of the present invention. To provide a thorough understanding of the present invention, specific details are described in detail below in connection with the preferred embodiments of the present invention, but those skilled in the art will be able to fully understand the present invention without these detailed descriptions. Furthermore, to avoid unnecessary confusion regarding the essence of the present invention, well-known methods, processes, procedures, components, and circuits have not been described in detail.

[0237] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A three-dimensional collaborative observation system for the exchange area sea area based on submarines and buoys, characterized by: It includes data acquisition module, instruction generation module, buoy dynamic response module, and time-space alignment and fusion module, among which; The data acquisition module collects the original observation data stream of the preset depth layer through the buoy system, divides the original observation data stream into data blocks according to the time window, extracts the feature vector of each data block and generates the buoy observation data set; The instruction generation module inputs the buoy observation data set into a preset anomaly detection model, and if the target ocean phenomenon characteristics are identified, generates a trigger instruction including the phenomenon type, occurrence depth and timestamp; The buoy dynamic response module transmits the trigger command 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 command, drives the lifting mechanism to locate to the target depth layer, and starts multi-sensor synchronous acquisition to form a three-dimensional observation data packet. The spatiotemporal alignment and fusion module performs time calibration on the stereo observation data packet based on the timestamp in the trigger instruction, performs three-dimensional interpolation based on the spatial coordinates of the latent buoy observation data set, and generates a spatiotemporally synchronized stereo data set.

2. The three-dimensional collaborative observation system for the exchange area sea area based on submarines and buoys according to claim 1 is characterized in that: The data acquisition module includes: The temperature, salinity, depth, and acoustic Doppler current profiler and dissolved oxygen sensor carried by the buoy system continuously collect raw observation data streams at multiple preset depth layers. The raw observation data streams include: Temperature, salinity and depth values ​​output by the temperature, salinity and depth instrument; Eastward, northward, and vertical current velocities output by the acoustic Doppler current profiler; Dissolved oxygen concentration and dissolved oxygen saturation output by the dissolved oxygen sensor; Timestamps collected synchronously by all sensors.

3. The three-dimensional collaborative observation system for the exchange area sea area based on submarines and buoys according to claim 2 is characterized in that: The data acquisition module also includes: The original observation data stream is divided into time windows of fixed length of 30 minutes to generate continuous data blocks; For each data block, time domain statistical features and frequency domain features are extracted respectively. Time domain statistical features include mean, variance and extreme value, and frequency domain features are energy features calculated based on wavelet transform. The time domain statistical features and frequency domain features are integrated to form a feature vector, which is then bound to the depth layer number and acquisition start and end timestamps of the corresponding data block to construct a structured buoy observation dataset. The submerged buoy observation data set is stored in real time by the underwater data storage unit of the submerged buoy system and is synchronously sent to the instruction generation module.

4. The three-dimensional collaborative observation system for the exchange area sea area based on submarines and buoys according to claim 3 is characterized in that: The instruction generation module includes: Receive a structured buoy observation dataset and input it into a preset anomaly detection model, the output of which includes the type identification of the target ocean phenomenon, the corresponding depth layer number, the corresponding confidence score, and the start and end timestamps of the data block to which the corresponding feature vector belongs; If the confidence score exceeds a preset threshold of 0.85, a trigger instruction is generated based on the type identification, depth layer number, and acquisition start and end timestamps output by the model. The trigger instruction includes the phenomenon type, occurrence depth, and timestamp; The trigger instruction is stored in the instruction cache queue and is ready to be sent 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 submarines and buoys according to claim 4 is characterized in that: The buoy dynamic response module includes: Receive the trigger command from the command generation module through the underwater acoustic communication link, and parse and obtain the phenomenon type, occurrence depth and timestamp; Matching a preset observation mode in the buoy system according to the phenomenon type, wherein the observation mode includes different sensor combination schemes and sampling frequencies; The lifting mechanism of the driving buoy locates the target depth layer 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 submarines and buoys according to claim 5 is characterized in that: 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 temperature, salinity and depth instrument to obtain the temperature, salinity and depth data of the current layer; Start the turbulence profiler to obtain turbulent energy distribution and shear rate data; Start the bio-optical sensor to obtain particle scattering intensity and chlorophyll fluorescence parameters; Continuously sample the target depth layer for no less than 5 minutes, and bind the sampled data with the corresponding timestamp and depth layer coordinates to generate a stereo observation data package.

7. The three-dimensional collaborative observation system for the exchange area sea area based on submarines and buoys according to claim 6 is characterized in that: The buoy system transmits the generated stereoscopic observation data packets back to the shore-based center in real time through its water wireless transmission unit, and retains a local cache for fault recovery mechanism.

8. The three-dimensional collaborative observation system for the exchange area sea area based on submarines and buoys according to claim 7 is characterized in that: The spatiotemporal alignment and fusion module includes: Receive the stereo observation data packet and extract its timestamp and depth layer coordinates; Perform time calibration based on timestamps to correct the clock deviation between the buoy system and the submerged buoy system to within ±0.1 seconds; The observation records of the corresponding time window and depth layer in the buoy observation dataset are called to perform spatial position matching.

9. The three-dimensional collaborative observation system for the exchange area sea area based on submarines and buoys according to claim 8 is characterized in that: The spatiotemporal alignment and fusion module also includes: After completing the spatial position matching, the Kriging interpolation algorithm is used to perform three-dimensional interpolation operations on the same observation parameters of the buoy observation data set and the stereo observation data set to generate a three-dimensional grid data volume of temperature, salinity, flow velocity and dissolved oxygen concentration; The interpolation results are combined with the time calibration results to output a temporally and spatially synchronized three-dimensional dataset including temperature field, salinity field, velocity field and dissolved oxygen concentration field. The three-dimensional dataset is then stored and synchronously transmitted to a shore-based analysis platform.

Images