System for monitoring sediment plume dynamics in a deep-sea mining area and multi-module analysis method thereof

Abstract

This invention discloses a dynamic monitoring system for sediment plumes in deep-sea mining areas and its multi-module fusion analysis method, belonging to the field of deep-sea environmental monitoring and marine engineering technology. The system includes fixed and mobile monitoring units. The fixed unit consists of an in-situ optical and three-dimensional sonar observation base station, a sediment trap array, and a multi-parameter sensor array, deployed in the mining area. The mobile unit includes an AUV equipped with a multispectral lidar and an ROV equipped with a binocular vision system and a miniature sampler. The analysis method uses multi-source data fusion, combined with dark channel defogging and pyramid optical flow method to extract the plume motion field, constructing a physics-data dual-driven model to simulate the plume diffusion and sedimentation process, generating a three-dimensional dynamic heat map in real time, and setting intelligent early warning thresholds based on diffusion radius and sedimentation flux. This invention achieves multi-scale, high-precision, and real-time dynamic monitoring and evaluation of plumes, effectively supporting the optimization of deep-sea mining operations and environmental compliance supervision.

Description

Technical Field

[0001] This invention relates to the field of deep-sea environmental monitoring and marine engineering equipment technology, and more specifically, to a dynamic monitoring system for sediment plumes in deep-sea mining areas and its multi-module analysis method. Background Technology

[0002] During deep-sea mining operations, seabed disturbance and fracturing transport create sediment plumes primarily composed of suspended fine particles. These plumes are characterized by large spatial scales (centimeter to meter level near the surface, extending to hundreds of meters to kilometers in the outer region), rapid temporal changes (mutual abrupt changes from seconds to minutes and evolutions on hourly to daily scales), wide particle size distribution, and strong coupling with the flow field and topography. Therefore, they place higher demands on the spatiotemporal resolution, quantitative capabilities, and engineering usability of monitoring systems. Existing technologies mainly focus on the following methods and processes:

[0003] I. Current Status of Typical Monitoring Methods

[0004] Optical imaging and video monitoring: Utilizing deep-sea cameras, laser speckle / beam and structured light for plume visualization and velocity estimation. This type of method can acquire relatively clear images under clear water or slightly turbid conditions, but the imaging contrast and effective visibility decrease significantly under strongly turbid, low-light, and long-distance conditions. Furthermore, the stability of 3D reconstruction and quantitative inversion is limited due to the influence of refraction, multiple scattering, and suspended matter attachment.

[0005] Acoustic observations: Doppler current profilers (ADCP / ADP) and multibeam / imaging sonar can provide velocity profiles and echo intensities over a wide range, which can be used to infer plume diffusion and relative concentration distribution. However, the conversion between acoustic echoes and actual volume fractions / mass concentrations depends on empirical calibration and is easily affected by particle size, shape, and gas content. Side-scan and imaging sonar have high near-field resolution, but there are uncertainties in identifying small-scale turbulence and boundary fronts.

[0006] In-situ physicochemical sensing: Sensors such as turbidimeters, optical backscattering (OBS), and conductivity-temperature-depth (CTD) can continuously record the physicochemical characteristics of plume water masses, providing indicative variables for the diffusion process. However, their limitations include the reliance on point or line sampling, making it difficult to directly obtain the three-dimensional volume field; inconsistencies in dimensions and response curves between different sensors, and difficulties in cross-device data fusion.

[0007] Sediment capture and sampling: Sediment traps, water samplers, and membrane filters can obtain key information such as sedimentation flux, particle size, and mineralogical composition, which are important bases for model calibration and environmental assessment. However, most of them are passive or time-triggered, making it difficult to correspond to the spatiotemporal dynamics of plume processes; the sampling points are sparse and the recovery cycle is long, making it difficult to support event-based diagnosis.

[0008] Mobile platform cruise: ROVs / AUVs equipped with multimodal sensors can achieve close-range observation and profile scanning in the near field, improving local spatial resolution. However, the trajectory planning and payload power consumption of mobile platforms are limited, and time synchronization, coordinate alignment, and data backhaul with fixed base stations often become bottlenecks affecting the accuracy of fusion reconstruction.

[0009] II. Current Status of Data Processing and Modeling Methods

[0010] Optical flow / feature tracking algorithms are used to estimate the velocity and deformation fields of a plume plane from video. Traditional CPU implementations are computationally intensive and struggle to meet online processing requirements for medium to high resolution videos. Furthermore, texture loss, occlusion, and strong scattering can cause motion estimation drift, with errors increasing significantly with distance.

[0011] Data assimilation and numerical simulation: Based on RANS / LES or geofluid models, concentration field and sedimentation flux inversion can be performed using observational data, which can improve the characterization of mid- and far-field diffusion. However, assimilation has high requirements for spatiotemporal synchronization and error statistics. Without a unified framework for heterogeneous data collaboration, the model still has biases in fitting microscopic processes such as near-field turbulent mixing and particle aggregation / flocculation.

[0012] Multi-source fusion and 3D reconstruction: Existing multi-source fusion methods are mostly "offline-batch processing" methods, which use acoustic profiles and sparse optical beams for volume field interpolation or visualization based on volume rendering. The real-time 3D boundary recognition and uncertainty assessment capabilities are relatively insufficient. The observation equations, noise models and calibration relationships of different sensors are not consistent, which leads to systematic errors in fusion.

[0013] III. Common Challenges in Project Deployment and Operation

[0014] Spatiotemporal synchronization and common reference coordinates: Fixed base stations and mobile platforms are usually driven by different clocks and inertial navigation / Doppler odometry, lacking a unified time base and acoustic / visual common reference coordinates, which leads to misalignment and superposition errors in data stitching and model driving.

[0015] Low temperature and high pressure and long-term reliability: Requirements such as pressure resistance at tens of thousands of meters, long-term corrosion protection and prevention of biofouling increase the design difficulty of optical windows, acoustic transducers and mechanical seals; continuous high-bandwidth data acquisition and edge computing under limited power supply also restrict the online degree of the system.

[0016] Triggering strategies and sample representativeness: Sampling triggers often rely on single-threshold sensors (such as turbidity), which are prone to false triggers or missed detections; the lack of sampling strategies that are tied to the dynamic events of the plume makes it difficult for samples to represent key stages (such as the initial rise, frontal passage and fall stages).

[0017] Data services and compliance requirements: Inconsistent standards for operational and regulatory indicators, non-standard event logs, and insufficient long-term baselines affect the quantitative assessment and comparative analysis of environmental impacts.

[0018] In the above context, the limitations of existing technologies are specifically manifested as follows:

[0019] —Single observation method: It mainly uses a single optical or acoustic method, which makes it difficult to simultaneously take into account near-field details and mid-to-far-field range, resulting in large errors in the reconstruction of three-dimensional motion trajectory.

[0020] —Data spatiotemporal mismatch: Fixed and mobile platforms lack a unified clock and spatial reference, making it difficult to support the fusion of multi-source data in the same spatiotemporal framework, and the accuracy of model-driven reconstruction drifts over time.

[0021] —Insufficient quantification of sedimentation: Passive capture and timed sampling are difficult to correspond to the dynamic process of plume, and cannot form a causal chain of "diffusion-sedimentation", resulting in a non-closed loop in flux assessment.

[0022] —Poor real-time performance: Traditional optical flow methods take a long time to process 1080P video, making it difficult to meet the needs of online monitoring, early warning and operation linkage in mining operations.

[0023] In summary, there is an urgent need for a multi-module integrated dynamic monitoring system for deep-sea mining scenarios. This system should achieve fixed-mobile collaboration and spatiotemporal co-reference at the deployment level; unified observation equations and real-time fusion of visual, acoustic, and physicochemical data at the algorithm level; and event-driven sampling and edge computing closed-loop at the engineering level. Furthermore, it should meet the dual requirements of operational optimization and environmental compliance with verifiable quantitative indicators (diffusion radius, settling flux, boundary positioning error, etc.). These pain points are precisely the technical problems that this invention aims to solve. Summary of the Invention

[0024] To overcome the shortcomings of existing technologies, this invention provides a dynamic monitoring system for sediment plumes in deep-sea mining areas and its multi-module analysis method. The invention, based on optical imaging, sediment capture, and multi-parameter sensing, is applicable to the quantitative assessment of plume diffusion behavior in scenarios such as deep-sea mining and subsea pipeline leaks.

[0025] The present invention is achieved through the following technical solution: a dynamic monitoring system for sediment plumes in deep-sea mining areas, the dynamic monitoring system for sediment plumes includes a fixed monitoring unit and a mobile monitoring unit;

[0026] The fixed monitoring unit includes an in-situ optical and three-dimensional sonar observation base station, as well as a sediment trap array and a multi-parameter sensor array arranged in a ring around the in-situ optical and three-dimensional sonar observation base station.

[0027] The in-situ optical and three-dimensional sonar observation base station includes a main frame, inside which are assembled a 75k acoustic Doppler profiler, an underwater camera, a 2M acoustic Doppler profiler, an acoustic Doppler current meter, and a multi-frequency digital imaging sonar; the underwater camera is a high-sensitivity CMOS camera in the blue-green light band, equipped with laser-assisted illumination and a narrowband optical filter; acoustic receivers are assembled on both the left and right sides of the main frame;

[0028] The sediment trap array includes three sediment trap stations, which are arranged in a non-uniform ring array with a radius of 50m centered on the optical and three-dimensional sonar observation base. Each sediment trap station includes a recovery beacon and a sediment trap. The recovery beacon and the sediment trap are connected by a first Kevlar cable. Several first floats are installed inside the recovery beacon. An acoustic transmitter is installed at the bottom of the recovery beacon. A counterweight is connected to the bottom of the sediment trap through an acoustic release device.

[0029] The multi-parameter sensor array includes three multi-parameter water monitoring stations, which are arranged in a non-uniform ring array with a radius of 50m centered on the optical and three-dimensional sonar observation base station. Each multi-parameter water monitoring station includes a float group consisting of several second floats and a counterweight. The float group and the counterweight are connected by a second Kevlar cable. A multi-parameter water quality meter is mounted on the second Kevlar cable. The multi-parameter water quality meter integrates sensors including a turbidity sensor, a conductivity sensor, and a dissolved oxygen sensor.

[0030] The mobile monitoring unit includes an autonomous underwater vehicle (AUV) and an unmanned remotely operated vehicle (ROV). The AUV is equipped with a multispectral lidar for plume boundary identification; the ROV is equipped with a binocular stereo vision system and a miniature sediment trap deployed with the robotic arm.

[0031] As a preferred option, the high-sensitivity CMOS camera in the blue-green light band has a frame rate of ≥60fps; the laser-assisted illumination has a wavelength of 532nm, a power of 5W, and an effective penetration distance of 20m; and the narrowband optical filter has a bandwidth of ±5nm.

[0032] As a preferred option, the sediment trap is a built-in pressure-triggered sampling tube with a trigger threshold of flow velocity > 0.15 m / s or turbidity change > 30%, supporting timed / event-driven dual-mode sampling.

[0033] As a preferred option, the multi-parameter water quality meter has a sampling frequency of 1Hz and a pressure resistance rating of 120MPa.

[0034] As a preferred option, the multispectral lidar has a scanning angle of 120° and a spatial resolution of 5cm.

[0035] As a preferred option, the binocular stereo vision system has a baseline distance of 30cm, a depth resolution of 1cm at a distance of 10m, and synchronously records RGB-D video streams.

[0036] A multi-module fusion analysis method for a dynamic monitoring system of sediment plumes in deep-sea mining areas, specifically including the following steps:

[0037] Step S0, Deployment Phase:

[0038] Centered on the in-situ optical and three-dimensional sonar observation base station, six base stations for sediment trap array and multi-parameter sensor array are positioned in a circular pattern with a spacing of 60°, with a radius of 50m. Three sediment trap station devices and three multi-parameter water body monitoring station devices are then deployed sequentially.

[0039] The mobile monitoring unit's autonomous underwater vehicle (AUV) and remotely operated vehicle (ROV) were launched into the water and set off along a preset cruise path, simultaneously establishing underwater acoustic communication with the fixed monitoring unit.

[0040] Step S1, Multivariate Data Processing:

[0041] Step S1-1: Processing steps for deep-sea images acquired by the camera: Apply dark channel prior dehazing to compensate for image degradation caused by water scattering; First, estimate the global ambient light (A) based on the observed image, then calculate the transmittance map (t(x)) characterizing the degree of water scattering based on this ambient light, and finally restore the clear scene image (J(x)) by solving the inverse problem of the underwater imaging model; The mathematical model describing the image degradation process is as follows:

[0042] I(x) = J(x)t(x) + A(1−t(x))

[0043] in:

[0044] I(x): is the observed foggy / degraded image.

[0045] J(x): is the desired clear, haze-free image to be recovered.

[0046] t(x): This is a transmittance map, representing the proportion of light reaching the camera. t(x)→0 means that the medium at this location is very dense, and the scene is completely invisible; t(x)→1 means that the medium is very thin, and the scene is clearly visible.

[0047] A: It is ambient light, which is usually assumed to be a global constant, representing the brightness of the water body at infinity;

[0048] The goal is to estimate A and t(x) from the known I(x) to obtain a clear image J(x); Step S1-2: Extract the plume motion vector field using optical flow method:

[0049] The pyramid LK optical flow method is used to extract high-precision plume motion vector fields in real time from consecutive frames of deep-sea optical images;

[0050] The extraction of the feather motion vector field is based on the fundamental constraint equation of the optical flow method, which assumes that the pixel intensity of objects in the scene remains constant between consecutive frames:

[0051]

[0052] in, I Represents the intensity of image pixels, ( x , y () represents pixel coordinates. t For time, (Δ x ,Δ y The displacement vector to be determined is the optical flow.

[0053] By performing a first-order Taylor expansion on the above equation and introducing a spacetime gradient, the problem can be transformed into solving the following linear system:

[0054]

[0055] in:

[0056] These are the spatial gradients of the image in the x and y directions, respectively;

[0057] The temporal gradient of the image;

[0058] (u,v) is the motion vector at the pixel (x,y) to be solved;

[0059] To handle the potentially large displacement movements of plumes, an image pyramid model is used to optimize the above solution process from coarse to fine; at each layer of the pyramid... L The optical flow vector obtained by solving will be used as the next layer. L The initial value of -1 is passed and refined:

[0060]

[0061] Finally, the final motion vector field is output at the bottom layer of the image pyramid (original resolution). ;

[0062] Step S2, Construction of the transport and diffusion model:

[0063] Step S2-1, Physical-Data Dual-Driven Modeling:

[0064] Step S2-1-1, Physical Constraint Layer: The numerical model used is MITgcm. A basic plume diffusion model is established based on the incompressible Navier-Stokes equations. The model's computational domain covers an actual seabed topographic area of ​​approximately 3×3 kilometers, with a horizontal resolution of 35 meters. 200 vertical layers are set to cover the entire overlying water body. The vertical layers are densified near the seabed, with the lowest layer having a resolution of 1 meter. The 35-meter horizontal resolution used in the model matches the best available measured topographic data, ultimately generating a topographic grid with a size of 35×35 meters.

[0065] Sediment categories were determined using the advection-diffusion scheme provided by MITgcm, employing a flux-limited direct spatiotemporal method for transport in the form of suspended particulate matter (SPM). The background water viscosity coefficient and sediment diffusion coefficient were both set to 10⁻⁻⁴. 4 m² / s and 10⁻ 5 m² / s; The vertical mixing process is solved using a nonlocal parameterization scheme. The SPM in the bottom layer grid cells can be separated from the water body and accumulated on the seabed through sedimentation. When the bottom shear velocity of the bottom grid reaches the critical value for particle resuspension, the deposited material can be resuspended and enter the water body.

[0066] Step S2-1-2, Data Fusion Layer:

[0067] The diffusion of sediment plumes from deep-sea mining was simulated with a time step of 5 seconds until the system reached stability.

[0068] Step S2-2, Calculation of dynamic settlement:

[0069] Establish the settlement formula:

[0070]

[0071]

[0072]

[0073] in, For the first i Layer sensor concentration (mg / L) The value represents the settlement velocity of the corresponding layer (m / s). The cross-sectional area of ​​the plume. Particle density (measured by sediment trap) (g / cm³), measured range 2.5-2.8 (clay-sand).

[0074] Step S3, 3D dynamic heat map: Map the model output to RGB transparency channels and overlay it on the seabed topography DEM data to display the plume core area, diffusion front and dense settlement area in real time;

[0075] Step S3-1, Data Preparation and Normalization:

[0076] Step S3-1-1: Obtain plume state data calculated by the aforementioned motion vector field or concentration inversion model, including sediment particle concentration distribution, motion velocity amplitude, or sedimentation flux; define this data as intensity matrix M(x, y), where (x, y) are image pixel coordinates or grid points in the geographic reference coordinate system;

[0077] Step S3-1-2: Normalize the intensity matrix M, linearly mapping its numerical range to the interval [0, 1], to obtain the normalized intensity matrix Mnorm:

[0078]

[0079] Step S3-2, Color Mapping:

[0080] Step S3-2-1: Define a color lookup table (Color Map) C, which maps intensity values ​​to specific RGB colors; including:

[0081] Low-value areas are mapped to cool colors: blue, RGB(0, 0, 255);

[0082] High-value areas are mapped to warm colors: red, RGB(255, 0, 0);

[0083] The intermediate value region is mapped to a transitional color: yellow, RGB(255, 255, 0);

[0084] Step S3-2-2: Based on each value in Mnorm(x, y), interpolate from the color lookup table C to generate the corresponding RGB color matrix RGB. plume (x, y).

[0085] Step 3-3, Alpha channel mapping for transparency:

[0086] The normalized intensity matrix Mnorm, either directly or after passing through an enhancement function, is used as the alpha channel matrix A. plume (x,y),

[0087]

[0088] Where γ is an adjustable parameter;

[0089] Steps 3-4: Terrain overlay and 3D rendering:

[0090] Step 3-4-1: Obtain the seabed digital elevation model (DEM) data D(x, y) for the same area;

[0091] Step 3-4-2: Create the feather RGBA layer [RGB...] plume A plume ] and the RGB of the seabed terrain texture layer seafloor Perform alpha blending; for each pixel, the final blended color is RGB. final Calculated by the following formula:

[0092]

[0093] Step 3-4-3: Apply the blended texture map to the 3D terrain mesh D(x, y), and use a computer graphics library to perform real-time 3D rendering, thereby generating a 3D dynamic heat map that updates in real time with the model data.

[0094] Step S4, Intelligent Early Warning Threshold:

[0095] When the plume diffusion radius is greater than the predetermined safe distance or the sedimentation flux is greater than the ecological carrying capacity threshold, an audible and visual alarm is triggered and uploaded to the water surface control center; this part defines the logical judgment conditions and system response process for the early warning.

[0096] Step 4-1, Calculation of early warning indicators:

[0097] The system calculates the following two key indicators in real time:

[0098] Plume spread radius R: Calculates the maximum distance from the plume source point to the current plume front;

[0099]

[0100] Settlement flux F: In key sensitive areas, calculate the average settlement flux value within the area, in g / m² / day; this value can be directly output from the model or derived from parameters such as concentration and velocity field.

[0101]

[0102] Step 4-2, Threshold Comparison and Alarm Triggering:

[0103] The system compares the real-time calculated metrics with preset security thresholds:

[0104] Preset safe distance threshold: Ralert; Preset ecological carrying capacity threshold: Falert; Alarm triggering logic is an "OR" relationship: if (R(t)>Ralert) or (F(t)>Falert), then an alarm is triggered;

[0105] As a preferred embodiment, step S1-1 specifically includes the following steps:

[0106] Step S1-1-1, Ambient Light Estimation:

[0107] Ambient light A is determined by finding the average brightness of a specific region in the observed image I(x):

[0108]

[0109] Where Φ represents the observed image I The set of pixels with the highest brightness (e.g., the original image pixel positions corresponding to the top 0.1% of the brightest pixels in the dark channel image). |Φ| represents the number of pixels in set Φ.

[0110] Step S1-1-2, Transmittance Map Estimation:

[0111] Transmittance diagram t ( x ) represents the light at a specific pixel point x The transmittance is estimated using the following formula:

[0112]

[0113] in:

[0114] ω It is an adjustable parameter (0 < ω ≤1), used to preserve the natural sense of depth in underwater scenes.

[0115] Ω( x ) is based on pixels x A local image patch centered on the image.

[0116] I c ( y ) is the observed image at the pixel level y Color channel c The intensity value.

[0117] A c Ambient light A In color channels c The amount.

[0118] Inner layer The operation targets the color channel. R , G , B conduct.

[0119] outer layer Operations on local image patch Ω ( x (To be carried out within)

[0120] Step S1-1-3, Clear Image Restoration:

[0121] Finally, the estimated transmittance map was used. t ( x ) and ambient light A A clear image J(x) is reconstructed through the inverse process of the underwater optical imaging model:

[0122]

[0123] in, t 0 is a preset lower limit threshold for transmittance, used to prevent noise amplification caused by an excessively small denominator.

[0124] As a preferred embodiment, step S2-1-2 specifically includes the following steps:

[0125] Step S2-1-2-1, Hydrodynamic Boundary Conditions: The lateral boundary is determined using hydrodynamic data measured at an in-situ optical and three-dimensional sonar observation base. Based on the bottom flow velocity, the peak and average velocity values ​​are calculated. Velocity troughs correspond to transition periods. Through spectral analysis, the semi-diurnal tide M2 ​​is dominant, followed by the quarter-diurnal tide M4. The average bed shear velocity ranges from 0.07 to 0.8 cm / s, corresponding to bed shear stresses between approximately 0 and 0.06 Pa. The average shear rate is 0.25 s⁻¹, as shown in the following formula:

[0126]

[0127] The bottom flow velocity, It is the von Kármán constant (0.41). The roughness length of the seabed. The height above the seabed. This represents the bottom shear velocity. The density of seawater is 1050 kg / m³. This is the bottom shear force. Seawater viscosity m 2 / s, For shear rate, The turbulent dissipation rate per unit mass;

[0128] Step S2-1-2-2: Salinity and temperature are based on in-situ measurements from a multi-parameter water quality analyzer;

[0129] Step S2-1-2-3: SPM concentration model based on in-situ optical and three-dimensional sonar observation base at a height of 2-3 meters above the ground (mab) - observational verification;

[0130] Step S2-1-2-4: Set the sediment plume release height to a range of 0–2 meters above the bottom.

[0131] Step S2-1-2-5, Sediment Particle Size Classification: Integrate the particle size distribution data (D50, sorting coefficient) of the sediment trap as the particle settling rate parameter in the model;

[0132] As a preferred option, the alarm response action in step 4-2 specifically includes the following steps:

[0133] When any of the conditions in step 4-2 are met, the system automatically executes the following alarm response actions:

[0134] Step 4-3-1, Local audible and visual alarm: Triggers the alarm lights and buzzer in the control room to alert the operator;

[0135] Step 4-3-2, Visual Warning on Interface: On the software monitoring interface, the areas exceeding the standard are highlighted with a conspicuous flashing outline;

[0136] Step 4-3-3, Data Upload: Generate alarm event messages, including alarm type (diffusion / settlement), exceedance value, occurrence time, and geographical location information, and automatically upload them to the surface control center or remote monitoring platform via underwater acoustic communication or network link;

[0137] Step 4-3-4: Log recording: Store all relevant data snapshots of this alarm event into the database for subsequent auditing and retrospective analysis.

[0138] By employing the above technical solutions, this invention has the following beneficial effects compared to existing technologies:

[0139] Deployment mode:

[0140] 1. Fixed-mobile integrated architecture: Fixed base stations (optical / imaging sonar / ADCP / multi-parameter) and mobile platforms such as AUV / ROV form a three-dimensional observation network, which takes into account both wide-area coverage and near-field high-precision details, and realizes multi-scale continuous monitoring of plumes at the "area-line-point" scale.

[0141] 2. Non-uniform array and self-adaptation to the main diffusion direction: The trap and multi-parameter probe are arranged in a dense-sparse non-uniform pattern, with the density increased along the inferred main diffusion direction, which improves the ability to characterize anisotropy and the representativeness of the samples.

[0142] 3. High-precision spatiotemporal synchronization and common reference coordinates: Underwater acoustic reference + clock synchronization enables the fusion of cross-platform data under the same spatiotemporal reference (time synchronization can be better than millisecond level), ensuring that multi-source data can be stitched together and traced back.

[0143] 4. Event / Timed Dual-Mode Trigger: Supports event-driven sampling for "turbidity / flow rate / motion vector anomalies → automatic approach and sampling", as well as timed sampling according to the task cycle. Combined with duration threshold to suppress false triggers and improve the effective sample ratio.

[0144] 5. Monitoring-Early Warning-Control Closed Loop: Real-time calculation of threshold indicators such as diffusion radius and settlement flux at the edge, triggering local and upper computer linkage early warning, and outputting speed / path adjustment suggestions to the operation end, forming an operation closed loop.

[0145] Algorithm fusion:

[0146] 1. Real-time visual field calculation: GPU parallel implementation of pyramid LK optical flow / feature tracking, enabling near real-time (e.g., ≥30 fps) vector field extraction in 4K format, achieving "online" quantization of plume motion.

[0147] 2. Multimodal 3D boundary recognition: Jointly model the point cloud density gradient of AUV multispectral lidar, the texture features of imaging sonar, and the visual optical flow field to robustly output the 3D boundary and volume classification of the plume.

[0148] 3. Physics-Data Dual-Driven Model: Fluid models such as MITgcm are used as physical priors and assimilated or constrained inverted with measured optical flow / acoustic velocity data to ensure that diffusion and sedimentation both follow the mechanism and fit the actual measurements, thereby improving extrapolability.

[0149] 4. Anomaly detection and adaptive sampling: Rapid anomaly identification is achieved based on time-series statistics and small sample detection (change points / mutations); the strategy layer adaptively allocates AUV / ROV approach trajectories and dwell times according to uncertainty and gradient information.

[0150] 5. Task visualization and decision support: The vector field, concentration field and seabed DEM are integrated into a three-dimensional dynamic thermo-streamline map, and the front, backflow and settling hotspots are marked in real time to assist in operation optimization.

[0151] Engineering practicality:

[0152] 1. Deep-sea environment reliability: The entire unit is adapted to pressure and low-temperature environments at depths of 10,000 meters. Long-term continuous operation and temperature difference / hybrid power supply architecture ensure mission endurance and stability.

[0153] 2. Low-light, high-turbidity imaging: High-sensitivity imaging in the blue-green band, narrow-band filtering and coaxial laser arrangement improve signal-to-noise ratio and penetration distance, balancing long-distance detection and close-range detail acquisition.

[0154] 3. Near-field quantitative accuracy: The ROV binocular RGB-D provides centimeter-level depth resolution, and dry / wet calibration and refraction compensation reduce systematic errors, supporting near-field observation of fine structures and sediment particles.

[0155] 4. Sampling effectiveness and prevention of accidental triggering: Multi-threshold logic (amplitude + duration) and redundant triggering (pressure / turbidity / flow rate) design reduce the accidental triggering rate and improve sample effectiveness and representativeness.

[0156] 5. Modularity and maintainability: The modular design of sensing, computing, power supply, communication and sampling, with standardized interfaces, facilitates quick replacement / expansion and reduces maintenance costs and downtime.

[0157] Other aspects:

[0158] 1. Quantitative closed-loop and verifiable indicators: Establish a closed-loop evaluation system of "observation-inversion-actual measurement comparison" to measure performance with indicators such as settlement flux, diffusion radius, and boundary positioning error, which facilitates review and project acceptance.

[0159] 2. Dual optimization of operation and environment: While meeting compliance thresholds, provide dynamic deceleration / avoidance strategies for the operation end to reduce the risk of plume resuspension and secondary diffusion.

[0160] 3. Deployment efficiency and cost savings: Fixed and mobile platforms reuse the same baseline and communication link, reducing ship time and deployment frequency; algorithms run in real time at the edge, reducing the cost of transmitting raw big data back.

[0161] 4. Regulatory and data service friendly: Key indicators and event logs are output in a structured manner, which facilitates the formation of compliance reports and ESG disclosures and supports the continuous operation of long-term environmental baselines.

[0162] Additional aspects and advantages of the invention will become apparent in the following description or may be learned by practice of the invention. Attached Figure Description

[0163] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which:

[0164] Figure 1 This is a deployment distribution diagram of the present invention;

[0165] Figure 2 This is a schematic diagram of a fixed monitoring unit;

[0166] Figure 3 This is a schematic diagram of a mobile monitoring unit;

[0167] Figure 4 The camera and laser are arranged in a coaxial configuration.

[0168] Figure 5 To measure water flow velocity and direction data at fixed monitoring stations;

[0169] Figure 6 Binocular vision cameras were used to observe plume images at fixed observation stations;

[0170] Figure 7 This is a flowchart of the optical-deposit-sensor data fusion process.

[0171] Figure 8 This is an example of coupling an optical flow vector field with a CFD model (arrows indicate the direction of motion, and color indicates the magnitude of velocity).

[0172] in, Figures 1 to 3 The correspondence between the reference numerals and components in the attached drawings is as follows:

[0173] 1. In-situ optical and 3D sonar observation base station; 1-1 Main frame; 1-2 75k acoustic Doppler profiler; 1-3 Underwater camera; 1-4 2M acoustic Doppler profiler; 1-5 Acoustic Doppler current meter; 1-6 Multi-frequency digital imaging sonar; 1-7 Acoustic receiver; 2. Sediment trap station device; 2-1 Recovery beacon; 2-2 Sediment trap; 2-3 First Kevlar cable; 2-4 First buoy; 2-5 Acoustic transmitter; 2-6 Acoustic release device; 2-7 Counterweight; 3. Multi-parameter water monitoring station device; 3-1 Second buoy; 3-2 Counterweight; 3-3 Second Kevlar cable; 3-4 Multi-parameter water quality meter; 4. Autonomous underwater vehicle (AUV); 4-1 Multispectral lidar; 5. Remotely operated vehicle (ROV); 5-1 Binocular stereo vision system, 5-2 micro sediment trap. Detailed Implementation

[0174] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0175] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0176] The following is combined Figures 1 to 8 The present invention provides a detailed description of the deep-sea mining area sediment plume dynamic monitoring system and its multi-module analysis method according to an embodiment of the present invention.

[0177] like Figures 1 to 3 As shown, this invention proposes a dynamic monitoring system for sediment plumes in deep-sea mining areas. The dynamic monitoring system for sediment plumes includes a fixed monitoring unit and a mobile monitoring unit.

[0178] The fixed monitoring unit includes an in-situ optical and three-dimensional sonar observation base station, as well as a sediment trap array and a multi-parameter sensor array arranged in a ring around the in-situ optical and three-dimensional sonar observation base station.

[0179] The in-situ optical and three-dimensional sonar observation base station 1 includes a main frame 1-1, inside which are assembled a 75k acoustic Doppler profiler 1-2, an underwater camera 1-3, a 2M acoustic Doppler profiler 1-4, an acoustic Doppler current meter 1-5, and a multi-frequency digital imaging sonar 1-6; the underwater camera 1-3 is a high-sensitivity CMOS camera with a blue-green light band of 470-530nm, equipped with laser-assisted illumination and a narrowband optical filter; acoustic receivers 1-7 are assembled on both the left and right sides of the main frame 1-1, and the acoustic receivers can receive acoustic information from the sediment trap array and the multi-parameter sensor array;

[0180] Optical parameter optimization: The selection criteria for the 470-530nm blue-green light band: In turbid water bodies such as deep-sea turbid water and estuaries, scattering is mainly caused by suspended particles. The order of light intensity is: blue-green light (470-530 nm) > blue light (<470 nm) > green light (>530 nm) > red-orange light. Suspended particles such as silt and plankton will scatter light, and the scattering intensity is inversely proportional to the wavelength. The shorter the wavelength, the stronger the scattering. Blue light at 400-470nm has a shorter wavelength and is more severely scattered by particles, resulting in decreased penetration. Green light at 530-600nm has a slightly longer wavelength. Although scattering is reduced, the absorption of green light by the water body begins to increase. The 470-530nm blue-green light band balances the "scattering loss" and "absorption loss" and becomes the optimal penetration band. It is proven that this band has the optimal penetration attenuation coefficient ≤0.1 / m in deep-sea turbid water bodies.

[0181] Anti-interference design: The camera is equipped with a narrow-band optical filter with a bandwidth of ±5nm to suppress stray light other than 532nm laser and improve the signal-to-noise ratio.

[0182] Structural details: The camera and laser are arranged in a coaxial configuration (see attached). Figure 2 To avoid shadow effects; the pressure-resistant housing is equipped with a sapphire viewing window with a light transmittance of >95%.

[0183] The sediment trap array includes three sediment trap station devices 2, which are arranged in a non-uniform ring array with a radius of 50m centered on the optical and three-dimensional sonar observation base station 1. Each sediment trap station 2 includes a recovery beacon 2-1 and a sediment trap 2-2. The recovery beacon 2-1 and the sediment trap 2-2 are connected by a first Kevlar cable 2-3. Several first floats 2-4 are installed inside the recovery beacon 2-1. An acoustic transmitter 2-5 is installed at the bottom of the recovery beacon 2-1. A counterweight 2-7 is connected to the bottom of the sediment trap 2-2 through an acoustic release device 2-6.

[0184] Trigger mechanism optimization: Added anti-accidental trigger logic: Flow velocity > 0.15 m / s and turbidity abrupt change lasting > 30 seconds must be simultaneously met to avoid interference from biological movement. Sample preservation: A low-temperature preservative at 4℃ is pre-placed inside the sampling tube to ensure that the organic components of the sediment do not degrade.

[0185] Layout strategy: The ring array adopts a non-uniform distribution, doubling the density in the main diffusion direction of the plume to improve the representativeness of the sampling.

[0186] The multi-parameter sensor array includes three multi-parameter water monitoring station devices 3, which are arranged in a non-uniform ring array with a radius of 50m centered on the optical and three-dimensional sonar observation base station 1. Each multi-parameter water monitoring station device 3 includes a float group consisting of several second floats 3-1 and a counterweight 3-2. The float group and the counterweight 3-2 are connected by a second Kevlar cable 3-3. A multi-parameter water quality meter 3-4 is mounted on the second Kevlar cable 3-3. The multi-parameter water quality meter 3-4 integrates a turbidity sensor (NTU), a conductivity sensor (CTD), and a dissolved oxygen sensor (DO). It has a sampling frequency of 1Hz and a pressure resistance of 120MPa. Self-cleaning function: The sensor probe is equipped with an ultrasonic vibrating plate with a frequency of 40kHz to prevent biological adhesion.

[0187] The mobile monitoring unit includes an autonomous underwater vehicle (AUV4) and a remotely operated vehicle (ROV5). The AUV4 is equipped with a multispectral lidar (4-1) for plume boundary identification. Its plume boundary identification algorithm detects gradient changes in point cloud density exceeding 10% / cm and outputs the 3D point cloud boundary of the plume in real time. The ROV5 is equipped with a binocular stereo vision system (5-1) and a miniature sediment trap (5-2) deployed with its robotic arm, supporting instantaneous sampling response time of less than 5 seconds on dynamic paths. The robotic arm is adapted with a sampling head integrating a force feedback device with a threshold of 5N to prevent damage from collisions with rocks.

[0188] Spatiotemporal reference synchronization: The position alignment between fixed base stations and mobile units is achieved by using underwater acoustic beacon frequencies of 8-12kHz, with clock synchronization accuracy <1ms.

[0189] Adaptive triggering logic:

[0190] When a fixed base station detects a sudden change in turbidity, it activates an ROV / AUV to travel to the target area via underwater acoustic communication.

[0191] When the optical flow method of the moving unit detects an anomaly in the plume motion vector, it triggers a nearby sediment trap to perform fixed-point sampling.

[0192] As a preferred option, the high-sensitivity CMOS camera in the blue-green light band (470-530nm) has a frame rate of ≥60fps; the laser-assisted illumination has a wavelength of 532nm, a power of 5W, and an effective penetration distance of 20m; and the narrowband optical filter has a bandwidth of ±5nm.

[0193] As a preferred option, the sediment trap 2-2 is a built-in pressure-triggered sampling tube with a trigger threshold of flow velocity > 0.15 m / s or turbidity change > 30%, supporting timed / event-driven dual-mode sampling.

[0194] As a preferred option, the multi-parameter water quality meter 3-4 has a sampling frequency of 1Hz and a pressure resistance rating of 120MPa.

[0195] As a preferred option, the multispectral lidar 4-1 has a scanning angle of 120° and a spatial resolution of 5cm.

[0196] As a preferred option, the binocular stereo vision system 5-1 has a baseline distance of 30cm, a depth resolution of 1cm at a distance of 10m, and synchronously records RGB-D video streams. Calibration method: Dry calibration using a checkerboard pattern is performed before underwater, followed by real-time wet calibration underwater using acoustic beacon positions to compensate for refraction errors.

[0197] A multi-module fusion analysis method for a dynamic monitoring system of sediment plumes in deep-sea mining areas, specifically including the following steps:

[0198] Step S0, Deployment Phase: Six fixed base stations (100m apart) are arranged in a ring around the mining vehicle, and the ROV carries the mobile unit to cruise along the preset path.

[0199] Step S0-1: Complete the commissioning of 6 fixed base station equipment, including the assembly of in-situ optical and three-dimensional sonar observation base stations, sediment trap array and multi-parameter sensor array, verify the clock synchronization accuracy of acoustic beacons (8-12kHz) (<1ms), and test the power and sampling functions of the ROV equipped with mobile units (binocular stereo vision system, micro sediment trap).

[0200] Step S0-2: Centered on the in-situ optical and three-dimensional sonar observation base station, locate 6 base stations of sediment trap array and multi-parameter sensor array in a circular pattern with a spacing of 60°, with a radius of 50m. Sequentially lower 3 sediment trap station devices and 3 multi-parameter water monitoring station devices to ensure that the sediment traps (uniformly distributed in a ring with a radius of 50m) and multi-parameter sensor array are accurately positioned and start the equipment self-test.

[0201] Step S0-3: The autonomous underwater vehicle (AUV) and remotely operated vehicle (ROV) of the mobile monitoring unit are launched into the water and set off along the preset cruise path. They simultaneously establish underwater acoustic communication with the fixed monitoring unit, confirm that the data transmission is normal, and enter the cruise monitoring state.

[0202] Event Response: When a base station detects a sudden increase in turbidity from 5 NTU to 80 NTU, it initiates AUV close-up imaging and simultaneously triggers sampling by sediment traps No. 3 and No. 4.

[0203] Model output: The optical flow method identifies the plume spreading northeastward at 0.3 m / s. The model predicts that the influence radius will reach 350 m after 2 hours, and the system automatically generates suggestions for decelerating mining vehicles.

[0204] 2. Verification of key parameters

[0205] The relative error between the measured sedimentation of the sediment trap (105 g / m²) and the model inversion value (98 g / m²) is <7%.

[0206] The system operated continuously for 68 hours at a depth of 11,000 meters, with the thermoelectric power generation module maintaining an average power consumption of 180W.

[0207] Step S1, Multivariate Data Processing:

[0208] Step S1-1: Processing steps for deep-sea images acquired by the camera: The core is to estimate transmittance and restore the image using the following formulas: Apply dark channel prior dehazing to compensate for image degradation caused by water scattering (PSNR improvement ≥15dB); First, estimate the global ambient light (A) based on the observed image, then calculate the transmittance map (t(x)) representing the degree of water scattering based on this ambient light, and finally restore a clear scene image (J(x)) by solving the inverse problem of the underwater imaging model; Use the mathematical model describing the image degradation process, the model is as follows:

[0209] I(x) = J(x)t(x) + A(1−t(x))

[0210] in:

[0211] I(x): is the observed foggy / degraded image (the image directly captured by the camera).

[0212] J(x): is the clear, fog-free image to be recovered (the true brightness of the scene).

[0213] t(x): This is a transmittance map, representing the proportion of light reaching the camera. t(x)→0 means that the medium (water) at this location is very dense, and the scene is completely invisible; t(x)→1 means that the medium is very thin, and the scene is clearly visible.

[0214] A: It is ambient light (background light), which is usually assumed to be a global constant, representing the brightness of the water body at infinity (usually biased towards blue or green).

[0215] The goal is to estimate A and t(x) from the known I(x) to obtain a clear image J(x); specifically, this includes the following steps:

[0216] Step S1-1-1, Ambient Light Estimation:

[0217] Ambient light A is determined by finding the average brightness of a specific region in the observed image I(x):

[0218]

[0219] Where Φ represents the observed image I The set of pixels with the highest brightness (e.g., the original image pixel positions corresponding to the top 0.1% of the brightest pixels in the dark channel image). |Φ| represents the number of pixels in set Φ.

[0220] Step S1-1-2, Transmittance Map Estimation:

[0221] Transmittance diagram t ( x ) represents the light at a specific pixel point x The transmittance is estimated using the following formula:

[0222]

[0223] in:

[0224] ω It is an adjustable parameter (0 < ω ≤1), used to preserve the natural sense of depth in underwater scenes.

[0225] Ω( x ) is based on pixels x A local image patch centered on the image.

[0226] I c ( y ) is the observed image at the pixel level y Color channel c The intensity value.

[0227] A c Ambient light A In color channels c The amount.

[0228] Inner layer The operation targets the color channel. R ,G , B conduct.

[0229] outer layer Operations on local image patch Ω ( x (To be carried out within)

[0230] Step S1-1-3, Clear Image Restoration:

[0231] Finally, the estimated transmittance map was used. t ( x ) and ambient light A A clear image J(x) is reconstructed through the inverse process of the underwater optical imaging model:

[0232]

[0233] in, t 0 is a preset lower limit threshold for transmittance (e.g.) t 0=0.1), used to prevent noise amplification caused by an excessively small denominator.

[0234] Step S1-2: Extract the plume motion vector field using optical flow method:

[0235] Employing the pyramid LK optical flow method (GPU accelerated), high-precision plume motion vector fields are extracted in real time from consecutive frames of deep-sea optical images (computation speed up to 30fps@4K resolution).

[0236] The extraction of the feather motion vector field is based on the fundamental constraint equations of the optical flow method, which assume that the pixel intensity of objects in the scene remains constant between consecutive frames:

[0237]

[0238] in, I Represents the intensity of image pixels, ( x , y () represents pixel coordinates. t For time, (Δ x ,Δ y The displacement vector to be determined is the optical flow.

[0239] By performing a first-order Taylor expansion on the above equation and introducing a spacetime gradient, the problem can be transformed into solving the following linear system:

[0240]

[0241] in:

[0242] These are the spatial gradients of the image in the x and y directions, respectively;

[0243] The temporal gradient of the image;

[0244] (u,v) is the motion vector at the pixel (x,y) to be solved;

[0245] To handle the potentially large displacement movements of plumes, an image pyramid model is used to optimize the above solution process from coarse to fine; at each layer of the pyramid... L The optical flow vector obtained by solving will be used as the next layer. L The initial value of -1 is passed and refined:

[0246]

[0247] Finally, the final motion vector field is output at the bottom layer of the image pyramid (original resolution). ;

[0248] To meet the performance requirements of real-time processing of 4K resolution sequences (≥30fps), all data-parallel operations in the pyramid optical flow calculation model, including but not limited to image pyramid construction, gradient calculation, matrix summation, and linear system solving, are executed in parallel on the GPU.

[0249] Step S2, Construction of the transport and diffusion model:

[0250] Step S2-1, Physical-Data Dual-Driven Modeling:

[0251] Step S2-1-1, Physical Constraint Layer: The numerical model used is MITgcm (General Circulation Model), which establishes a basic plume diffusion model based on the incompressible Navier-Stokes equations. The model's computational domain covers an actual seabed topography area of ​​approximately 3×3 kilometers, with a horizontal resolution of 35 meters. 200 vertical layers are set to cover the entire overlying water body. The vertical layers are densified near the seabed, with the lowest layer having a resolution of 1 meter. The 35-meter horizontal resolution used in the model matches the best available measured topographic data (from the shipborne EM122 12 kHz multibeam system), ultimately generating a topographic grid with a size of 35×35 meters.

[0252] This model can define and simulate any number of user-defined sediment categories based on settling velocity. In this patent application, we used three sediment particle size categories: 61.2 μm, 10.4 μm, and 1.65 μm (the laser diffraction volume fraction percentiles d90, d50, and d10 of the sediments, hereinafter referred to as the 61 μm category, 10 μm category, and 2 μm category, respectively). Considering the viscosity and density of deep-sea water, the Stokes settling velocities corresponding to the 61 μm, 10 μm, and 2 μm sediment categories are 1.67 mm / s, 0.048 mm / s, and 0.001 mm / s, respectively.

[0253] Sediment categories were determined using the advection-diffusion scheme provided by MITgcm, employing a flux-limited direct spatiotemporal method for transport in the form of suspended particulate matter (SPM). The background water viscosity coefficient and sediment diffusion coefficient were both set to 10⁻⁻⁴. 4 m² / s and 10⁻ 5 m² / s; a nonlocal parameterization scheme is used to solve the vertical mixing process to optimize the simulation accuracy of deep-sea mixing effects. SPM in the bottom layer grid cells can be separated from the water body and accumulated on the seabed through sedimentation; when the bottom shear velocity of the bottom grid reaches the particle resuspension critical value, the sediment can be resuspended and enter the water body; real-time CTD data is input to correct the density gradient parameters.

[0254] Step S2-1-2, Data Fusion Layer:

[0255] The diffusion of sediment plumes from deep-sea mining was simulated with a time step of 5 seconds until the system reached stability; the specific steps included:

[0256] Step S2-1-2-1, Hydrodynamic Boundary Conditions: The lateral boundary is determined using hydrodynamic data measured at an in-situ optical and three-dimensional sonar observation base station; based on the bottom flow velocity, the peak and average flow velocities are calculated, and the flow troughs correspond to the flow transition periods (e.g., ...). Figure 4 Spectral analysis showed that the semi-diurnal tide M2 ​​was dominant, followed by the quarter-diurnal tide M4; the average bed shear velocity ranged from 0.07 to 0.8 cm / s (mean 0.3 cm / s), corresponding to bed shear stresses between approximately 0 and 0.06 Pa (mean 0.01 Pa); the average shear rate was 0.25 s⁻¹ (1 / s), as shown in the following formula:

[0257]

[0258]

[0259]

[0260]

[0261] The bottom flow velocity, It is the von Kármán constant (0.41). The roughness length of the seabed. The height above the seabed. This represents the bottom shear velocity. The density of seawater is 1050 kg / m³. This is the bottom shear force. Seawater viscosity m 2 / s, For shear rate, The turbulent dissipation rate per unit mass;

[0262] Step S2-1-2-2: Salinity and temperature are based on in-situ measurements from a multi-parameter water quality analyzer;

[0263] Step S2-1-2-3: Based on the in-situ optical and three-dimensional sonar observation base at a height of 2–3 meters above the bottom (mab), the SPM concentration model was observed and verified, and the sediment release rate was determined to be 0.142 kg / s.

[0264] Step S2-1-2-4: Set the sediment plume release height to a range of 0–2 meters above the bottom.

[0265] Step S2-1-2-5, Sediment Particle Size Classification: The particle size distribution data (D50, sorting coefficient) of the sediment trap are used as the particle settling rate parameter in the model.

[0266] Step S2-2, Calculation of dynamic settlement:

[0267] Establish the settlement formula:

[0268]

[0269]

[0270]

[0271] in, For the first i Layer sensor concentration (mg / L) The value represents the settlement velocity of the corresponding layer (m / s). The cross-sectional area of ​​the plume. Particle density (measured by sediment trap) (g / cm³), measured range 2.5-2.8 (clay-sand).

[0272] Step S3, 3D dynamic heat map: Map the model output to RGB transparency channels and overlay it on the seabed topography DEM data to display the plume core area, diffusion front and dense settlement area in real time;

[0273] Step S3-1, Data Preparation and Normalization:

[0274] Step S3-1-1: Obtain plume state data calculated by the aforementioned motion vector field or concentration inversion model, including sediment particle concentration distribution, motion velocity amplitude, or sedimentation flux; define this data as intensity matrix M(x, y), where (x, y) are image pixel coordinates or grid points in the geographic reference coordinate system;

[0275] Step S3-1-2: Normalize the intensity matrix M, linearly mapping its numerical range to the interval [0, 1], to obtain the normalized intensity matrix Mnorm:

[0276]

[0277] Step S3-2, Color Mapping:

[0278] Step S3-2-1: Define a color lookup table (Color Map) C, which maps intensity values ​​to specific RGB colors; including:

[0279] Low-value areas (such as background density) are mapped to cool tones: blue, RGB(0, 0, 255);

[0280] High-value areas (such as the core area) are mapped to a warm color: red, RGB(255, 0, 0);

[0281] The intermediate value region is mapped to a transitional color: yellow, RGB(255, 255, 0);

[0282] Step S3-2-2: Based on each value in Mnorm(x, y), interpolate from the color lookup table C to generate the corresponding RGB color matrix RGB. plume (x, y).

[0283] Step 3-3, Alpha channel mapping for transparency:

[0284] The normalized intensity matrix Mnorm is used directly or after passing through an enhancement function (such as the exponential function f(x) = x^γ to highlight high-value regions) as the transparency channel matrix A. plume (x, y),

[0285]

[0286] Here, γ is an adjustable parameter (usually γ ≥ 1); this makes areas with higher concentration (larger values) more opaque, and areas with lower concentration (closer to 0 values) more transparent, thus achieving a natural fading effect for feather shapes.

[0287] Steps 3-4: Terrain overlay and 3D rendering:

[0288] Step 3-4-1: Obtain the seabed digital elevation model (DEM) data D(x, y) for the same area;

[0289] Step 3-4-2: Create the feather RGBA layer [RGB...] plume A plume ] and the RGB of the seabed terrain texture layer seafloor Perform alpha blending; for each pixel, the final blended color is RGB. final Calculated by the following formula:

[0290]

[0291] Step 3-4-3: Apply the blended texture map to the 3D terrain mesh D(x, y), and use computer graphics libraries (such as OpenGL, WebGL) for real-time 3D rendering to generate a 3D dynamic heat map that updates in real time with the model data; users can interactively observe the spatial distribution of plumes on the seabed terrain from any angle.

[0292] Step S4, Intelligent Early Warning Threshold:

[0293] When the plume diffusion radius is greater than the predetermined safe distance (e.g., 500m outside the mining area boundary) or the sedimentation flux is greater than the ecological carrying capacity threshold (e.g., 200g / m² / day), an audible and visual alarm is triggered and uploaded to the water surface control center; this part defines the logical judgment conditions and system response process for the early warning.

[0294] Step 4-1, Calculation of early warning indicators:

[0295] The system calculates the following two key indicators in real time:

[0296] Plume diffusion radius R: Calculates the maximum distance from the plume source point (such as a mining vehicle) to the current plume front (usually defined as the farthest boundary of all pixels with a concentration value greater than a certain threshold θradius);

[0297] ;

[0298] Settlement flux F: In key sensitive areas, including outside the mining area boundary and ecological protection zone, calculate the average settlement flux value within the area, in g / m² / day; this value can be directly output from the model or derived from parameters such as concentration and velocity field.

[0299] ;

[0300] Step 4-2, Threshold Comparison and Alarm Triggering:

[0301] The system compares the real-time calculated metrics with preset security thresholds:

[0302] Preset safe distance threshold: Raalert (e.g., 500 meters); Preset ecological carrying capacity threshold: Falert (e.g., 200 g / m² / day); Alarm triggering logic is an "OR" relationship: if (R(t)>Ralert) or (F(t)>Falert), then an alarm is triggered;

[0303] The alarm response action in step 4-2 specifically includes the following steps:

[0304] When any of the conditions in step 4-2 are met, the system automatically executes the following alarm response actions:

[0305] Step 4-3-1, Local audible and visual alarm: Triggers the alarm lights and buzzer in the control room to alert the operator;

[0306] Step 4-3-2, Visual Warning on Interface: On the software monitoring interface, the areas exceeding the standard are highlighted with a conspicuous flashing outline (such as a red circle);

[0307] Step 4-3-3, Data Upload: Generate alarm event messages, including alarm type (diffusion / settlement), exceedance value, occurrence time, and geographical location information, and automatically upload them to the surface control center or remote monitoring platform via underwater acoustic communication or network link;

[0308] Step 4-3-4: Log recording: Store all relevant data snapshots of this alarm event into the database for subsequent auditing and retrospective analysis.

[0309] The image represents the temporal gradient of the image. In the description of this invention, the term "multiple" refers to two or more. Unless otherwise explicitly defined, the terms "upper," "lower," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention. The terms "connect," "install," "fix," etc., should be interpreted broadly. For example, "connect" can be a fixed connection, a detachable connection, or an integral connection; it can be a direct connection or an indirect connection through an intermediate medium. For those skilled in the art, the specific meaning of the above terms in this invention can be understood according to the specific circumstances.

[0310] In the description of this specification, the terms "one embodiment," "some embodiments," "specific embodiment," etc., refer to a specific feature, structure, material, or characteristic described in connection with that embodiment or example, which is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0311] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A multi-module fusion analysis method for a dynamic monitoring system of sediment plumes in deep-sea mining areas, characterized in that, The sediment plume dynamic monitoring system includes fixed monitoring units and mobile monitoring units; The fixed monitoring unit includes an in-situ optical and three-dimensional sonar observation base station, and a sediment trap array and a multi-parameter sensor array arranged in a ring around the in-situ optical and three-dimensional sonar observation base station. The in-situ optical and three-dimensional sonar observation base station includes a main frame, inside which are assembled a 75k acoustic Doppler profiler, an underwater camera, a 2M acoustic Doppler profiler, an acoustic Doppler current meter, and a multi-frequency digital imaging sonar; the underwater camera is a high-sensitivity CMOS camera in the blue-green light band, equipped with laser-assisted illumination and a narrowband optical filter; acoustic receivers are assembled on both the left and right sides of the main frame. The sediment trap array includes three sediment trap stations. The three sediment trap stations are arranged in a non-uniform ring array with a radius of 50m centered on the in-situ optical and three-dimensional sonar observation base. Each sediment trap station includes a recovery beacon and a sediment trap. The recovery beacon and the sediment trap are connected by a first Kevlar cable. Several first floats are installed inside the recovery beacon. An acoustic transmitter is installed at the bottom of the recovery beacon. A counterweight is connected to the bottom of the sediment trap through an acoustic release device. The multi-parameter sensor array includes three multi-parameter water monitoring stations. These three stations are arranged in a non-uniform ring array with a radius of 50m centered on the in-situ optical and three-dimensional sonar observation base. Each multi-parameter water monitoring station includes a float group consisting of several second floats and a counterweight. The float group and the counterweight are connected by a second Kevlar cable. A multi-parameter water quality meter is mounted on the second Kevlar cable. The multi-parameter water quality meter integrates a turbidity sensor, a conductivity sensor, and a dissolved oxygen sensor. The mobile monitoring unit includes an autonomous underwater vehicle (AUV) and an unmanned remotely operated vehicle (ROV). The AUV is equipped with a multispectral lidar for plume boundary identification. The ROV is equipped with a binocular stereo vision system and a miniature sediment trap deployed with the robotic arm; The analytical method specifically includes the following steps: Step S0, Deployment Phase: Centered on the in-situ optical and three-dimensional sonar observation base station, three sediment trap stations and three multi-parameter water monitoring stations were deployed in sequence with a circular angular spacing of 60° and a radius of 50m. The mobile monitoring unit's AUV and ROV are launched into the water and set off along a preset cruise path, simultaneously establishing underwater acoustic communication with the fixed monitoring unit. Step S1, Multivariate Data Processing: Step S1-1: Processing steps for deep-sea images acquired by the camera: Apply dark channel prior dehazing to compensate for image degradation caused by water scattering; First, estimate the global ambient light A based on the observed image, then calculate the transmittance map t(x) representing the degree of water scattering based on the ambient light, and finally restore the clear scene image J(x) by solving the inverse problem of the underwater imaging model. Step S1-2: Extract the plume motion vector field using optical flow method: The pyramid LK optical flow method is used to extract high-precision plume motion vector fields in real time from consecutive frames of deep-sea optical images; Step S2, Construction of the transport and diffusion model: Step S2-1, Physical-Data Dual-Driven Modeling: Step S2-1-1, Physical Constraint Layer: The numerical model used is MITgcm. A basic plume diffusion model is established based on the incompressible Navier-Stokes equations. The model's computational domain covers a 3×3 km actual seabed topography area with a horizontal resolution of 35 meters. 200 vertical layers are set to cover the entire overlying water body. The vertical layers are densified near the seabed, with the lowest layer having a resolution of 1 meter. The 35-meter horizontal resolution used in the model matches the best measured topographic data obtained, and finally a topographic grid with a size of 35×35 meters is generated. Sediment categories were determined using the advection-diffusion scheme provided by MITgcm, employing a flux-constrained direct spatiotemporal method for transport as suspended particulate matter; the background water viscosity coefficient and sediment diffusion coefficient were set as follows: and The vertical mixing process is solved using a nonlocal parameterization scheme. Suspended particles in the bottom grid cells are separated from the water body and accumulate on the seabed through sedimentation. When the bottom shear velocity of the bottom grid reaches the critical value for particle resuspension, the deposited material is resuspended and enters the water body. Step S2-1-2, Data Fusion Layer: The diffusion of sediment plumes from deep-sea mining was simulated with a time step of 5 seconds until the system reached stability. Step S2-2, Calculation of dynamic settlement: Establish the settlement formula: in, For the first i Layer sensor concentration (mg / L) The corresponding settlement velocity is in m / s. The cross-sectional area of ​​the plume. The particle density is expressed in g / cm³, with a measured range of 2.5-2.

8. Step S3, 3D dynamic heat map: Map the model output to RGB transparency channels and overlay it on the seabed topography DEM data to display the plume core area, diffusion front and dense settlement area in real time; Step S3-1, Data Preparation and Normalization: Step S3-1-1: Obtain plume state data calculated by the aforementioned motion vector field or transport and diffusion model, including sediment particle concentration distribution, motion velocity amplitude or sedimentation flux; define this data as intensity matrix M(x, y), where (x, y) are image pixel coordinates or grid points in the geographic reference coordinate system; Step S3-1-2: Normalize the intensity matrix M(x, y) by linearly mapping its numerical range to the interval [0, 1] to obtain the normalized intensity matrix M. norm (x, y); Step S3-2, Color Mapping: Step S3-2-1: Define a color lookup table C, which maps intensity values ​​to specific RGB colors; including: Low-value areas are mapped to cool colors: blue, RGB(0, 0, 255); High-value areas are mapped to warm colors: red, RGB(255, 0, 0); The intermediate value region is mapped to a transitional color: yellow, RGB(255, 255, 0); Step S3-2-2: Based on the normalized intensity matrix M norm For each value in (x, y), interpolation is performed from the color lookup table C to generate the corresponding RGB color matrix RGB. plume (x, y); Step S3-3, Alpha channel mapping: The normalized intensity matrix M norm (x, y) directly or after an enhancement function is used as the alpha channel matrix A. plume (x,y), ； Where γ is an adjustable parameter; Step S3-4, Terrain Overlay and 3D Rendering: Step S3-4-1: Obtain the seabed digital elevation model (DEM) data D(x, y) for the same area; Step S3-4-2: Transfer the feather RGB layer [RGB] plume A plume ] and the RGB of the seabed terrain texture layer seafloor Perform alpha blending; for each pixel, the final blended color is RGB. final Calculated by the following formula: ； Step S3-4-3: Apply the blended texture map to the 3D terrain mesh D(x, y), and use a computer graphics library to perform real-time 3D rendering, thereby generating a 3D dynamic heat map that updates in real time with the model data. Step S4, Intelligent Early Warning Threshold: When the plume diffusion radius is greater than the predetermined safe distance or the sedimentation flux is greater than the ecological carrying capacity threshold, an audible and visual alarm is triggered and uploaded to the water surface control center. Step S4-1, Calculation of early warning indicators: The system calculates the following two key indicators in real time: Plume spread radius R: Calculates the maximum distance from the plume source point to the current plume front; ； Settlement flux F: In key sensitive areas, calculate the average settlement flux value within that area, in g / m² / day; this value is directly output from the model or derived from parameters including concentration and velocity fields. ； Step S4-2, Threshold Comparison and Alarm Triggering: The system compares the real-time calculated metrics with preset security thresholds: Preset safety threshold: Ralert; Preset ecological carrying capacity threshold: Falert; Alarm triggering logic is an "OR" relationship: if R(t)>Ralert or F(t)>Falert, then an alarm is triggered.

2. The multi-module fusion analysis method for a deep-sea mining area sediment plume dynamic monitoring system according to claim 1, characterized in that, The high-sensitivity CMOS camera in the blue-green light band has a frame rate of ≥60fps; the laser-assisted illumination has a wavelength of 532nm, a power of 5W, and an effective penetration distance of 20m; the narrowband optical filter has a bandwidth of ±5nm.

3. The multi-module fusion analysis method for a deep-sea mining area sediment plume dynamic monitoring system according to claim 1, characterized in that, The sediment trap is a built-in pressure-triggered sampling tube with a trigger threshold of flow velocity > 0.15 m / s or turbidity mutation > 30%, supporting timed / event-driven dual-mode sampling.

4. The multi-module fusion analysis method for a deep-sea mining area sediment plume dynamic monitoring system according to claim 1, characterized in that, The multi-parameter water quality meter has a sampling frequency of 1Hz and a pressure resistance rating of 120MPa.

5. The multi-module fusion analysis method for a deep-sea mining area sediment plume dynamic monitoring system according to claim 1, characterized in that, The multispectral lidar has a scanning angle of 120° and a spatial resolution of 5cm.

6. The multi-module fusion analysis method for a deep-sea mining area sediment plume dynamic monitoring system according to claim 1, characterized in that, The baseline distance of the binocular stereo vision system is 30cm.

7. The multi-module fusion analysis method for a deep-sea mining area sediment plume dynamic monitoring system according to claim 1, characterized in that, Step S1-1 specifically includes the following steps: Step S1-1-1, Ambient Light Estimation: Ambient light A is determined by finding the average brightness of a specific region in the observed image I(x): ； Where Φ represents the observed image I The set of pixels with the highest brightness; |Φ| represents the number of pixels in set Φ; Step S1-1-2, Transmittance Map Estimation: Transmittance diagram t ( x ) represents the light at a specific pixel point x The transmittance is estimated using the following formula: ； in: ω It is an adjustable parameter 0< ω ≤1; Ω( x ) is based on pixels x A local image patch centered on the image; I c ( y ) is the observed image at the pixel level y Color channel c The intensity value; A c Ambient light A In color channels c The amount of; inner layer The operation targets the color channel. RGB Perform; outer layer Operations on local image patch Ω ( x (to be carried out within) Step S1-1-3, Clear Image Restoration: Finally, the estimated transmittance map was used. t ( x ) and ambient light A A clear image J(x) is reconstructed through the inverse process of the underwater optical imaging model: ； in, t 0 is a preset lower limit threshold for transmittance, used to prevent noise amplification caused by an excessively small denominator.

8. The multi-module fusion analysis method for a deep-sea mining area sediment plume dynamic monitoring system according to claim 1, characterized in that, Step S2-1-2 specifically includes the following steps: Step S2-1-2-1, Hydrodynamic Boundary Conditions: The lateral boundary was determined using hydrodynamic data measured at an in-situ optical and three-dimensional sonar observation base. Based on the bottom flow velocity, the peak and average velocity values ​​were calculated. Velocity troughs corresponded to transition periods. Spectral analysis showed that the semi-diurnal tide M2 ​​was dominant, followed by the quarter-diurnal tide M4. The average bed shear velocity ranged from 0.07 to 0.8 cm / s, corresponding to bed shear stresses between 0 and 0.06 Pa. The average shear rate was... The formula is as follows: ； Let be the bottom velocity, k be the von Kármán constant (0.41), z0 be the seabed roughness length, and z be the height above the seabed. This represents the bottom shear velocity. The density of seawater is 1050 kg / m³. This is the bottom shear force. Seawater viscosity m 2 / s, Where E is the shear rate and E is the turbulent dissipation rate per unit mass. Step S2-1-2-2: Salinity and temperature are based on in-situ measurements from a multi-parameter water quality analyzer; Step S2-1-2-3: Based on the in-situ optical and three-dimensional sonar observation base, the concentration model of suspended particulate matter at a height of 2-3 meters above the ground is observed and verified. Step S2-1-2-4: Set the sediment plume release height to a range of 0–2 meters above the bottom. Step S2-1-2-5, Sediment Particle Size Category: The particle size distribution data of the sediment trap is used as the particle settling rate parameter in the model.

9. The multi-module fusion analysis method for a deep-sea mining area sediment plume dynamic monitoring system according to claim 1, characterized in that, The alarm response action in step S4-2 specifically includes the following steps: When any of the conditions in step S4-2 are met, the system automatically executes the following alarm response actions: Step S4-3-1, Local audible and visual alarm: Triggers the alarm lights and buzzer in the control room to alert the operator; Step S4-3-2, Visual warning on the interface: On the software monitoring interface, the areas exceeding the standard are highlighted with a conspicuous flashing outline; Step S4-3-3, Data Upload: Generate an alarm event message, including alarm type, exceeding value, occurrence time and geographical location information, and automatically upload it to the water surface control center or remote monitoring platform via underwater acoustic communication or network link; Step S4-3-4, Log recording: Store all relevant data snapshots of this alarm event into the database for subsequent auditing and retrospective analysis.

Images