A marine environment data intelligent collection and analysis method and system

Abstract

The application discloses a kind of marine environment data intelligent collection analysis method and system, for the judgment of coral reef to be repaired area and the determination of repair time window, this method divides the rectangular grid of 10m×10m into coral reef monitoring area, and the picture information and environmental data of each grid are collected by autonomous underwater vehicle;Based on image recognition technology, extract coral coverage, whitening rate and species diversity index, combine environmental parameters to calculate coral reef health index and degradation rate, determine repair priority and mark the grid to be repaired;For the grid to be repaired, combined with marine environment and weather information to establish repair suitability evaluation model, determine the best repair time window through multi-objective optimization algorithm, the application realizes the intelligentization and refinement of coral reef monitoring, provides scientific decision-making basis for coral reef repair, and has important engineering application value.

Description

Technical Field

[0001] This invention relates to the field of marine environmental monitoring and analysis technology, and in particular to a method and system for intelligent acquisition and analysis of marine environmental data. Background Technology

[0002] Traditional coral reef monitoring methods primarily rely on manual surveys and periodic sampling at fixed monitoring stations, which suffer from low efficiency, high cost, and limited spatial coverage. Manual surveys are limited by diving depth and operating time, making it difficult to achieve large-scale continuous monitoring, and the timeliness and accuracy of data acquisition need improvement. Existing coral reef health assessment methods often focus on single indicators, lacking a comprehensive quantitative assessment system, particularly lacking intelligent assessment methods that combine image recognition technology and diverse environmental data. The assessment results are highly subjective and fail to provide a scientific basis for restoration decisions.

[0003] The timing of coral reef restoration directly impacts the success rate and ecological benefits of restoration projects. However, current methods primarily rely on expert experience and simple seasonal patterns, lacking scientific analysis of marine environmental conditions, meteorological factors, and dynamic ecosystem changes. In particular, the comprehensive impact assessment of key environmental factors such as typhoons, waves, and ocean currents is inaccurate, leading to significant environmental risks and unstable success rates for restoration projects. Furthermore, the lack of scientific methods for predicting restoration effects and a quantitative prioritization system hinders the rational allocation of restoration resources and the scientific formulation of restoration strategies.

[0004] Therefore, there is an urgent need for a method based on intelligent collection and analysis of marine environmental data to achieve intelligent assessment and scientific decision-making for coral restoration. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for intelligent acquisition and analysis of marine environmental data, in order to solve the technical problems of low accuracy and poor efficiency of existing coral reef monitoring technologies, lack of scientific identification of areas to be restored, and strong subjectivity in the selection of restoration timing, so as to realize refined monitoring, intelligent assessment and scientific restoration decision-making of coral reef ecosystems.

[0006] To achieve the above objectives, this invention provides an intelligent data acquisition and analysis method for marine environments, used for identifying coral reef restoration areas and determining restoration time windows. The method includes the following steps:

[0007] Step 1: Divide the coral reef monitoring area into 10m×10m rectangular grids, and construct the data collection path of the autonomous underwater vehicle using the center coordinates of each grid as the marking point;

[0008] The autonomous underwater vehicle is used to collect image information and environmental data from each grid, including water temperature, salinity, pH value, and dissolved oxygen.

[0009] Step S2: Based on the image information and environmental data of each grid, identify the grid to be repaired and calculate the coral reef health index and degradation rate of each grid, obtain the repair priority of each grid and mark the grid to be repaired.

[0010] Step S3: For the grid to be repaired, a repair suitability assessment model is established by combining weather information including wave height, ocean current speed and typhoon path forecast. The prediction result of the best repair time window is given by a multi-objective optimization algorithm.

[0011] Furthermore, the data acquisition path of the autonomous underwater vehicle is planned according to the grid row and column order, the travel time between adjacent grids is not less than 2 minutes, and the data acquisition time at the center of each grid is not less than 3 minutes.

[0012] The image information is used to identify coral coverage, bleaching rate, and species diversity index.

[0013] Furthermore, coral coverage C is extracted from image information using image recognition technology. c (i,j), whitening rate C b (i,j) and species diversity index S s (i,j), where (i,j) are the coordinates of the label point of the grid cell in the i-th row and j-th column, denoted as grid cell (i,j);

[0014] Where A c (i,j) represents the coral coverage area of ​​grid (i,j), A g =100m 2 Area of ​​a single grid cell; Where A b (i,j) represents the area of ​​bleached coral in grid (i,j); the Shannon-Wiener index was used as the species diversity index. Where, p k denoted as the relative abundance of the k-th coral species, and n as the number of coral species identified within the grid.

[0015] Furthermore, the reef health index H of grid (i,j) I The formula for calculating (i,j,t) is:

[0016] H I (i,j,t)=[C c (i,j)·(1―0.8·C b (i,j))·S s (i,j)] 0.6 ·[E t (i,j,t)·E c(i,j,t)] 0.4 ;

[0017] Among them, E t (i,j,t) and E c (i,j,t) represent the temperature suitability index and chemical environment suitability index of grid (i,j), respectively;

[0018]

[0019] Where, (T(i,j,t) is the water temperature of grid (i,j) at time t; E p (i,j,t),E d (i,j,t),E s (i,j,t) represent the pH suitability, saturated dissolved oxygen concentration suitability, and salinity suitability of grid (i,j) at time t, respectively.

[0020]

[0021] (pH(i,j,t), DO(i,j,t), and Sal(i,j,t) represent the pH value, saturated dissolved oxygen concentration, and salinity of the grid (i,j) at time t, respectively.

[0022] Furthermore, the degradation rate R of the raster (i,j) d (i,j,t) is calculated based on changes in health indices monitored over 30 consecutive days:

[0023] Where Δt = 30 days.

[0024] Furthermore, fix priority P r (i,j,t) is: P r (i,j,t)=(1―H I (i,j,t))(1+0.5·P d (i,j,t))·A f (i,j), where the area factor A f (i,j): N c (i,j) represents the number of grid cells to be repaired that are connected to grid cell (i,j);

[0025] When the raster satisfies H I (i,j,t)<0.6 and P r When (i,j,t)>0.3, it is marked as a raster to be repaired.

[0026] Furthermore, the optimal repair time window [t] s ,t eDetermined through a remediation suitability assessment model;

[0027] The model is represented by the objective function as follows:

[0028] The constraints are:

[0029]

[0030] Where Ω represents the set of all graticles to be repaired, and S min =0.7 is the minimum requirement for repair suitability, S I (i,j,t) represents the repair suitability of the raster (i,j) at time t, where T min The shortest time required to complete a repair cycle; P t (t) represents the probability of a typhoon occurring at time t; H w (t) represents the wave height at time t, V c (i,j,t) represents the ocean current velocity at time t. Solve for the objective function and maximize its value under the constraints. The corresponding time window is the optimal time window.

[0031] Furthermore, the repair suitability S of grid (i,j) at time t I The formula for calculating (i,j,t) is:

[0032]

[0033] Furthermore, the shortest time T required to complete one repair cycle. min The timeframe is determined based on the type of restoration work: 14 days for coral transplantation and restoration, and 21 days for artificial base construction.

[0034] Based on the same inventive concept, in another aspect, the present invention provides a marine environmental data intelligent acquisition and analysis system, the system comprising: a data acquisition module, a grid identification module to be repaired, and a repair window prediction module;

[0035] The data acquisition module divides the coral reef monitoring area into 10m×10m rectangular grids, and constructs the acquisition path of the autonomous underwater vehicle using the center coordinates of each grid as the annotation point.

[0036] The autonomous underwater vehicle is used to collect image information and environmental data from various grids, including water temperature, salinity, pH value, and dissolved oxygen.

[0037] The grid identification module is used to identify the grids to be repaired based on the image information and environmental data of each grid, calculate the coral reef health index and degradation rate of each grid, obtain the repair priority of each grid, and mark the grids to be repaired.

[0038] The repair window prediction module is used to establish a repair suitability assessment model for the grid to be repaired, combining weather information including wave height, ocean current speed and typhoon path forecast, and to give the prediction result of the best repair time window through a multi-objective optimization algorithm.

[0039] Compared with the prior art, the beneficial effects of the present invention are:

[0040] This invention establishes a comprehensive health assessment system based on image recognition and multivariate environmental data, combined with degradation rate analysis and connectivity assessment, achieving an accuracy rate of over 95% in identifying areas to be restored. It also constructs a multi-factor restoration suitability assessment model, comprehensively considering marine environment and meteorological conditions, particularly the impact of extreme weather such as typhoons, significantly improving the scientific rigor of restoration timing selection and increasing the restoration success rate by over 35%. Furthermore, it achieves intelligent management of the entire process from data collection and health assessment to restoration decision-making, providing a complete technical solution for coral reef protection and restoration, and possessing significant theoretical value and promising engineering applications. Attached Figure Description

[0041] Figure 1 This is a flowchart of an intelligent marine environmental data acquisition and analysis method according to the present invention;

[0042] Figure 2 This is a schematic diagram of the composition of a marine environmental data intelligent acquisition and analysis system according to the present invention. Detailed Implementation

[0043] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention are described clearly and completely below. Obviously, the described embodiments are only a part of the embodiments of this invention, not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0044] Example 1

[0045] like Figure 1 The diagram shown is a flowchart of an intelligent marine environmental data acquisition and analysis method according to the present invention, used for determining the coral reef restoration area and the restoration time window. The method includes the following steps:

[0046] Step 1: Divide the coral reef monitoring area into 10m×10m rectangular grids, and construct the data collection path of the autonomous underwater vehicle using the center coordinates of each grid as the marker point.

[0047] The autonomous underwater vehicle is used to collect image information and environmental data from various grids, including water temperature, salinity, pH value, and dissolved oxygen.

[0048] The coral reef monitoring area was divided into 10m × 10m rectangular grids, and the data collection path of the autonomous underwater vehicle (AUV) was constructed using the center coordinates of each grid as the markers. Taking a typical 1km × 1km monitoring area as an example, it was divided into 100 × 100 = 10,000 grids, each with an area of ​​100 square meters. Grid numbers were assigned using a row-column coordinate system; for example, grid (1,1) was located in the upper left corner of the monitoring area, and grid (100,100) was located in the lower right corner. The AUV operated along a zigzag path, scanning row by row starting from the first row and first column, ensuring coverage of all grids.

[0049] The data collection path of the autonomous underwater vehicle is planned according to the grid row and column order, with a travel time of no less than 2 minutes between adjacent grids and a collection time of no less than 3 minutes at the center of each grid; the image information is used to identify coral coverage, bleaching rate and species diversity index.

[0050] For example, the autonomous underwater vehicle (AUV) uses the BlueROV2 model, equipped with a 4K high-definition camera and a YSI EXO2 multi-parameter water quality monitor, operating at a depth range of 0-100 meters. At the center coordinates of each grid, the AUV hovers at a height of 2-3 meters above the seabed, using the camera to photograph the coral reef area on the bottom, covering the entire 10m × 10m grid. Simultaneously, the water quality monitor measures the water temperature, salinity, pH value, and dissolved oxygen concentration at that point in real time. The required accuracy for environmental data acquisition is: water temperature ±0.1℃, salinity ±0.1 psu, pH value ±0.01, and dissolved oxygen ±0.1 mg / L.

[0051] The data acquisition path of the autonomous underwater vehicle is planned according to the grid row and column order. The travel time between adjacent grids is set to be no less than 2 minutes, mainly considering ocean current resistance and turning time. For example, moving from grid (1,1) to grid (1,2) is a travel distance of 10 meters. The average speed of the vehicle is 0.5 m / s, and the theoretical travel time is 20 seconds. Including the turning and stabilization time, the total time is no less than 2 minutes. The data acquisition time at the center of each grid is no less than 3 minutes, including 1 minute for taking pictures (taking 20-30 pictures from different angles) and 2 minutes for measuring environmental parameters (data is recorded every 30 seconds, for a total of 4 records, and the average value is taken).

[0052] The image information is primarily used for subsequent image recognition and analysis, identifying coral coverage, the proportion of bleached corals, and species diversity. The images are in JPEG format with a resolution of at least 4096×2160 pixels to ensure the identification of coral details down to the centimeter level. Each grid contains 20-30 images, captured from various angles including vertical overhead and 45-degree oblique views, providing ample visual information for the image recognition algorithm.

[0053] Step S2: Based on the image information and environmental data of each grid, identify the grid to be repaired and calculate the coral reef health index and degradation rate of each grid, obtain the repair priority of each grid and mark the grid to be repaired.

[0054] Image recognition processing employs a deep learning-based semantic segmentation algorithm, using a pre-trained DeepLab V3+ model for coral species identification and region segmentation. The algorithm can identify different categories such as hard corals, soft corals, anemones, seaweed, rocks, and sand, and automatically calculates the pixel area for each category. Coral coverage C is extracted from the image information using image recognition technology. c (i,j), whitening rate C b (i,j) and species diversity index S s (i,j), where (i,j) are the coordinates of the label point of the grid cell in the i-th row and j-th column, denoted as grid cell (i,j);

[0055] Where A c (i,j) represents the coral coverage area of ​​grid (i,j), A g =100m 2 Area of ​​a single grid cell; Where A b (i,j) represents the area of ​​bleached coral in grid (i,j); the Shannon-Wiener index was used as the species diversity index. Where, p k Let represent the relative abundance of the k-th coral species, and n be the number of coral species identified within the raster. For example, if a raster image has 4 million pixels, and 1.2 million pixels are marked as coral, then the coral coverage area is 30 square meters, and the coral coverage rate is 0.3. Albino coral identification is based on color characteristics. Normal corals exhibit rich colors such as brown and green, with RGB values ​​ranging from (80-180, 100-200, 60-150), while albino corals appear white or pale yellow, with RGB values ​​ranging from (200-255, 200-255, 180-255). Color thresholding segmentation technology is used to distinguish between normal and albino coral areas.

[0056] Image recognition algorithms can distinguish common coral species such as staghorn coral, brain coral, plate coral, and surface-porous coral. For example, if a grid identifies three coral species: staghorn coral (60%), brain coral (30%), and plate coral (10%), the diversity index is 0.90.

[0057] Coral reef health index H of grid (i,j) I The formula for calculating (i,j,t) is:

[0058] H I(i,j,t)=[C c (i,j)·(1―0.8·C b (i,j))·S s (i,j)] 0.6 ·[E t (i,j,t)·E c (i,j,t)] 0.4 ;

[0059] Among them, E t (i,j,t) and E c (i,j,t) represent the temperature suitability index and chemical environment suitability index of grid (i,j), respectively. According to coral ecology research, coral reef health is mainly determined by biological and environmental factors. Principal component analysis (PCA) was performed on five consecutive years of monitoring data from 15 typical coral reef areas worldwide. The results showed that biological factors (coverage, bleaching rate, diversity) contributed 61.2% to the health index, while environmental factors (temperature, chemical environment) contributed 38.8%. Considering the error range in practical applications, the weighting coefficients were set to 0.6 and 0.4, respectively. Different weighting combinations were used to evaluate 200 coral reef samples with known health status. When the weighting ratio was 0.6:0.4, the correlation coefficient between the evaluation results and expert evaluation reached 0.92, which was significantly higher than other weighting combinations.

[0060] (1―0.8·C b (i,j))·S s (i,j)] 0.6 The coefficient 0.8 represents the bleaching rate weighting coefficient. Coral bleaching is a manifestation of the breakdown of the relationship between corals and symbiotic algae, severely affecting the coral's photosynthesis and calcification ability. According to marine biological research, when the bleaching rate is <10%, coral function is basically normal, with an impact coefficient close to 0; when the bleaching rate is 10-30%, coral function declines significantly, with an impact coefficient of 0.2-0.6; and when the bleaching rate is >30%, coral function is severely damaged, with an impact coefficient >0.6. Through nonlinear regression analysis of the relationship between bleaching rate and coral function decline, the best fit coefficient was found to be 0.8 (R²). 2 =0.87).

[0061] The coral calcification rate was highest at 26℃, reaching 4.2 μmol CaCO3 / cm³. 2 / h; When the temperature deviates from 26℃, the calcification rate decreases in a Gaussian distribution. When the temperature exceeds 29℃ or falls below 23℃, corals begin to show stress responses.

[0062]

[0063] Where, (T(i,j,t) is the water temperature of grid (i,j) at time t; E p (i,j,t),E d (i,j,t),E s (i,j,t) represent the pH suitability, saturated dissolved oxygen concentration suitability, and salinity suitability of grid (i,j) at time t, respectively.

[0064] At pH 8.1, the aragonite saturation Ωarag≈3.5, which is the optimal condition for coral calcification; at pH < 7.9, aragonite begins to dissolve, and the coral skeletal structure is damaged; at pH > 8.3, the calcification process is inhibited.

[0065] (pH(i,j,t), DO(i,j,t), and Sal(i,j,t) represent the pH, saturated dissolved oxygen concentration, and salinity of grid (i,j) at time t, respectively. The oxygen saturation concentration in tropical seawater at 26℃ and 35 psu is approximately 8.2 mg / L. Respiration in corals and symbiotic algae requires sufficient dissolved oxygen. DO > 8 mg / L: sufficient oxygen supply, normal coral metabolism; DO = 6-8 mg / L: mild hypoxia, affecting coral vitality; DO < 6 mg / L: severe hypoxia, causing coral stress. The surface salinity of tropical oceans is typically 34-36 psu, with 35 psu being the standard seawater salinity. Corals adapt their osmoregulation mechanisms to this salinity environment: when salinity deviates by ±2 psu, the osmolarity regulation burden on corals increases; when salinity deviates by ±4 psu, cell function begins to be impaired.)

[0066] The degradation rate R of grid (i,j) d (i,j,t) is calculated based on changes in health indices monitored over 30 consecutive days:

[0067] Where Δt = 30 days.

[0068] Repair priority P r (i,j,t) is: P r (i,j,t)=(1―H I (i,j,t))·(1+0.5·R d (i,j,t))·A f (i,j), where the area factor is (i,j). N c (i,j) represents the number of grid cells to be repaired that are connected to grid cell (i,j);

[0069] When the raster satisfies H I (i,j,t)<0.6 and P rWhen (i,j,t)>0.3, it is marked as a grid to be repaired. Three hundred coral reef sampling points were tracked and monitored for five years. The optimal threshold was evaluated based on the repair effect: HI>0.8, healthy state, strong natural recovery ability; HI=0.6-0.8, sub-healthy state, requiring protection and management; HI<0.6, degraded state, requiring active repair intervention. The repair priority threshold analysis results are: Pr>0.5, high priority, immediate repair; Pr=0.3-0.5, medium priority, planned repair; Pr<0.3, low priority, monitoring and observation.

[0070] Step S3: For the grid to be repaired, a repair suitability assessment model is established by combining weather information including wave height, ocean current speed and typhoon path forecast. The prediction result of the best repair time window is given by a multi-objective optimization algorithm.

[0071] For the grid to be repaired, a repair suitability assessment model is established by combining weather information, including wave height, ocean current speed, and typhoon path forecast. The optimal repair time window is predicted by a multi-objective optimization algorithm.

[0072] Meteorological and marine condition data sources include: wave forecast products from the National Marine Environmental Forecasting Center, providing daily wave height forecasts for the next 30 days with a spatial resolution of 0.1°×0.1°; ocean current velocity forecasts provided by ocean numerical models, with a temporal resolution of 3 hours and a spatial resolution of 0.05°×0.05°; and typhoon track probability forecasts issued by the China Meteorological Administration, including the typhoon center location, intensity level, movement speed, and forecast error circle.

[0073] Optimal repair time window [t] s ,t e Determined through a remediation suitability assessment model;

[0074] The model is represented by the objective function as follows:

[0075] The constraints are:

[0076]

[0077] Where Ω represents the set of all graticles to be repaired, and S min =0.7 is the minimum requirement for repair suitability, S I (i,j,t) represents the repair suitability of the raster (i,j) at time t, where T min The shortest time required to complete a repair cycle; P t (t) represents the probability of a typhoon occurring at time t; H w (t) represents the wave height at time t, V c(i,j,t) represents the ocean current velocity at time t. Solve for the objective function and maximize its value under the constraints. The corresponding time window is the optimal time window.

[0078] Repair suitability S of grid (i,j) at time t I The formula for calculating (i,j,t) is:

[0079]

[0080] The shortest time T required to complete a repair cycle min The time allotted for restoration work is determined based on the type of restoration work: 14 days for coral transplantation and 21 days for artificial base construction. Coral transplantation includes coral fragment collection (3 days), transportation and cultivation (2 days), on-site transplantation (7 days), and initial monitoring (2 days), totaling 14 days. Artificial base construction includes material preparation (5 days), base installation (10 days), coral attachment (4 days), and stability observation (2 days), totaling 21 days.

[0081] Taking a 1km×1km coral reef area as an example, the method of this invention was applied to identify areas to be repaired and predict repair time windows. The area was divided into 10,000 grids. An autonomous underwater vehicle completed data collection for the entire area in 15 days, identifying 1,247 grids to be repaired, with a total area of ​​12.47 hectares.

[0082] Health index calculation results show that the regional average health index is 0.52, with 657 grid cells having a health index below 0.4 (severe degradation) and 590 grid cells having a health index between 0.4 and 0.6 (moderate degradation). Degradation rate analysis revealed that 82% of the grid cells requiring restoration showed a continuous degradation trend, with a monthly average degradation rate of 0.015. Prioritization determined 200 grid cells to be prioritized for restoration, mainly distributed in gentle slope areas with water depths of 8-15 meters and good connectivity.

[0083] The predicted repair window spans from March to May 2024, taking into account historical typhoon data and spring sea state characteristics. The calculations show that April 15-30 is the optimal repair window, during which the typhoon probability is only 3%, the average wave height is 0.8 meters, the current velocity is 0.12 m / s, and the water temperature remains stable between 25.8-26.2℃, resulting in a repair suitability index of 0.85.

[0084] Actual restoration work verification: Coral transplantation and restoration were carried out according to the predicted time window, with a total of 2,000 staghorn coral fragments transplanted. The transplantation success rate reached 87%, which is 22 percentage points higher than the historical average success rate (65%). A follow-up survey three months later showed that the coral coverage in the restored area increased from 15% to 42%, and the biodiversity index increased from 0.6 to 1.2, verifying the effectiveness of the method of this invention.

[0085] Example 2

[0086] like Figure 2 The diagram shown is a schematic representation of the composition of an intelligent marine environmental data acquisition and analysis system according to the present invention. The system includes: a data acquisition module, a grid identification module to be repaired, and a repair window prediction module.

[0087] The data acquisition module divides the coral reef monitoring area into 10m×10m rectangular grids, and constructs the acquisition path of the autonomous underwater vehicle using the center coordinates of each grid as the annotation point.

[0088] The data acquisition module is responsible for the spatial division and data acquisition of the coral reef monitoring area. It divides the monitoring area into 10m × 10m rectangular grids and constructs the data acquisition path for the autonomous underwater vehicle (AUV) using the center coordinates of each grid as markers. This module includes a path planning subsystem, a navigation control subsystem, and a data recording subsystem. The path planning subsystem uses GIS technology to automatically generate a grid coordinate system and navigation path based on the geographic coordinates and water depth / topographic data of the monitoring area. The navigation control subsystem combines GPS positioning and inertial navigation to achieve precise positioning and path tracking of the AUV, with a positioning accuracy of ±1 meter.

[0089] The autonomous underwater vehicle (AUV) adopts a modular design, with a streamlined hull measuring 1.2 meters in length, 0.8 meters in width, and 0.6 meters in height. It is equipped with a lithium battery pack, providing an endurance of 8 hours and a maximum operating depth of 100 meters. The propulsion system includes four vector thrusters, providing forward, backward, surfacing, diving, and steering functions, with a maximum speed of 1.5 m / s and a hovering accuracy of ±0.5 meters. Sensor payloads include a 4K camera (Sony α7R IV, 42.4 megapixels), an LED lighting system (6000 lumens), a multi-parameter water quality probe (YSI EXO2), and an acoustic positioning system (USBL).

[0090] Environmental data acquisition accuracy is strictly controlled: the water temperature sensor uses a platinum resistance thermometer, with a measurement range of -5℃ to 50℃, an accuracy of ±0.01℃, and a response time of 1 second; the salinity sensor uses the conductivity method, with a measurement range of 0-70 psu and an accuracy of ±0.1 psu; the pH sensor uses the ion-selective electrode method, with a measurement range of 6-9, an accuracy of ±0.01, and a calibration cycle of 24 hours; the dissolved oxygen sensor uses the fluorescence method, with a measurement range of 0-50 mg / L, an accuracy of ±0.1 mg / L, and requires no electrolyte consumption. All sensor data is transmitted digitally at a sampling frequency of 10Hz and stored on a waterproof solid-state drive.

[0091] The raster recognition module is responsible for image processing, health assessment, and restoration requirements analysis. It utilizes a high-performance image processing workstation equipped with an NVIDIA RTX 4090 graphics card and 64GB of memory, supporting parallel processing of a large number of high-resolution images. The image recognition algorithm is developed based on the deep learning framework TensorFlow 2.0, using a pre-trained ResNet-101 backbone network and a DeepLab V3+ segmentation head, and undergoes fine-tuning training specifically for coral reef scenes.

[0092] The training dataset contains 50,000 labeled images from different sea areas, including the South China Sea, the Red Sea, and the Great Barrier Reef in Australia, covering 30 common coral species and 10 substrate types. Data annotation uses polygon segmentation, achieving pixel-level accuracy. Model training employs a transfer learning strategy, adapting to the domain based on ImageNet pre-trained weights. The training run consists of 200 epochs with a learning rate of 0.001, a batch size of 16, and utilizes the AdamW optimizer.

[0093] Image preprocessing includes geometric correction, color equalization, and noise reduction. Geometric correction eliminates lens distortion and the effects of shooting angle, using a checkerboard calibration board to obtain camera intrinsic parameters and distortion parameters. Color equalization compensates for uneven underwater lighting and color attenuation, based on a white balance algorithm and gamma correction. Noise reduction employs a bilateral filter to smooth texture noise while maintaining edge sharpness.

[0094] Coral coverage calculation is based on pixel statistics. First, the segmentation results are used to generate mask images according to categories. Then, the number of pixels in each category is counted, and the pixel area is converted into the actual area according to the camera calibration parameters. The bleaching rate detection combines shape and color features: normal corals are brown, green, etc., with RGB mean values ​​in the range of (80-180, 100-200, 60-150); bleached corals are white or pale yellow, with RGB mean values ​​in the range of (200-255, 200-255, 180-255) and color saturation below 0.3.

[0095] Species identification employs a hierarchical classification strategy: first, hard corals are distinguished from soft corals, and then further subdivided into specific species such as staghorn corals, brain corals, plate corals, surface-porous corals, and cup corals. The identification algorithm classifies based on morphological features (branched, spherical, plate-like, tubular, etc.) and texture features (smooth, rough, striped, etc.), achieving an accuracy rate of 92%.

[0096] The health index calculation module implements parallel computing, supporting the simultaneous processing of health assessments for 1000 grid cells. The degradation rate calculation employs time series analysis, establishing a linear regression model of health index changes and fitting the trend line using the least squares method. Repair priority ranking utilizes multi-criteria decision analysis, comprehensively considering factors such as health status, degradation rate, spatial connectivity, and repair feasibility.

[0097] The autonomous underwater vehicle is used to collect image information and environmental data from various grids, including water temperature, salinity, pH value, and dissolved oxygen.

[0098] The grid identification module is used to identify the grids to be repaired based on the image information and environmental data of each grid, calculate the coral reef health index and degradation rate of each grid, obtain the repair priority of each grid, and mark the grids to be repaired.

[0099] The repaired window forecast module integrates meteorological and marine state data interfaces to obtain real-time marine environmental forecast information, including multiple data sources such as the China Meteorological Administration, the European Centre for Medium-Range Weather Forecasts (ECMWF), and the U.S. National Hurricane Center (NHC). Data acquisition utilizes a RESTful API interface, supporting JSON and XML formats.

[0100] Typhoon track prediction employs an ensemble forecasting method, integrating forecasts from multiple numerical models, including the China Typhoon Model (CMA-TYM), the Global Forecast System (GFS), and the European Medium-Range Weather Forecasting Model (ECMWF). Each typhoon track forecast includes 24-hour, 48-hour, 72-hour, 96-hour, and 120-hour location probability distributions, and the Monte Carlo method is used to calculate the probability of typhoon impact.

[0101] Wave forecasting is based on the third-generation wave numerical model SWAN. Input wind field data includes wind speed and direction at a height of 10 meters, with a spatial resolution of 0.1°×0.1° and a temporal resolution of 1 hour. Ocean current forecasting uses the ocean circulation numerical model HYCOM, providing a three-dimensional ocean current velocity field with 40 vertical layers and a surface resolution of 5 meters.

[0102] The repair window prediction module is used to establish a repair suitability assessment model for the grid to be repaired, combining weather information including wave height, ocean current speed and typhoon path forecast, and to give the prediction result of the best repair time window through a multi-objective optimization algorithm.

[0103] The remediation suitability assessment employs a multi-objective optimization algorithm, with objective functions including maximizing suitability, maximizing remediation efficiency, and minimizing risk. Constraints include environmental, resource, time, and safety constraints. The optimization algorithm utilizes a hybrid strategy combining genetic algorithm (GA) and particle swarm optimization (PSO), with a population size of 100, 200 generations, a crossover probability of 0.8, and a mutation probability of 0.1.

[0104] The search time window is set to cover the next 90 days, with a time resolution of 1 day, resulting in 90 candidate time windows. For each candidate window, the objective function value and constraint satisfaction are calculated, and the optimal time window is determined using the Pareto optimal solution set. The algorithm takes approximately 10 minutes to run and provides 3-5 alternative time windows for decision-making reference.

[0105] The restoration effect prediction is based on a machine learning model using historical restoration data. The training data includes information such as environmental conditions, restoration methods, restoration scale, and success rate of 500 historical restoration projects. The prediction model uses a random forest algorithm, which contains 100 decision trees. The feature variables include 15 factors such as average water temperature, ocean current speed, wave height, typhoon frequency, and restoration area during the restoration period. The model's prediction accuracy reaches 82%, providing a quantitative assessment of the success probability for restoration decisions.

[0106] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. 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. An intelligent collection and analysis method of marine environment data, used for judging a coral reef area to be repaired and determining a repair time window, characterized in that, The method comprises the following steps: Step 1, the coral reef monitoring area is divided into 10m 10m rectangular grid, and the center coordinates of each grid are taken as the marking points to construct the collection path of the autonomous underwater vehicle. The autonomous underwater vehicle is used for collecting picture information and environmental data of each grid, the environmental data including water temperature, salinity, pH value and dissolved oxygen; the collection path of the autonomous underwater vehicle is planned according to the grid row and column order, the navigation time between adjacent grids is not less than 2 minutes, and the collection time at the center of each grid is not less than 3 minutes; the picture information is used for identifying the coral coverage, whitening rate and species diversity index; In step S2, according to the picture information and environmental data of each grid, the grid to be repaired is identified, the coral reef health index and degradation rate of each grid are calculated, the repair priority of each grid is obtained, and the grid to be repaired is marked; Coral coverage rate is extracted from image information using image recognition technology. Whitening rate and species diversity index ,in, For the first Line 1 The coordinates of the grid's annotation points are denoted as grid. ; where is the area of the grid covered by coral , is the area of a single grid cell; where is the area of the grid covered by white coral , where, is the relative abundance of the th coral species, is the number of coral species identified within the grid. In step S3, for the grid to be repaired, a repair suitability evaluation model is established by combining weather information including sea wave height, sea current speed and typhoon path forecast, and a prediction result of the best repair time window is given through a multi-objective optimization algorithm.

2. The method of claim 1, wherein, grid coral health index is calculated as ； wherein, and are the temperature suitability index and the chemical environment suitability index, respectively, of the grid of cells ； ； wherein, is a grid at time temperature of water; , , are respectively a grid at time pH value suitability, saturation dissolved oxygen concentration suitability, and salinity suitability; ； ； ; , , are grid pH, saturation dissolved oxygen concentration and salinity at time .

3. The method of claim 2, wherein, Grid degradation rate Health index change calculated by monitoring for 30 consecutive days: ; wherein, days.

4. The method of claim 3, wherein, Repair priority is: wherein the area factor : , is the number of grid cells to be repaired in communication with the grid cell. When the grid satisfies and the grid is marked as to be repaired.

5. The method of intelligent acquisition and analysis of marine environment data according to claim 4, characterized in that, Optimal repair time window determined by a repair appropriateness assessment model; The model is expressed in terms of an objective function as: ; The constraint condition is: ； wherein, is the set of all grids to be repaired, is the minimum requirement of repair suitability, is the grid is the repair suitability at time is the minimum time required to complete a repair cycle; is the probability of typhoon occurrence at time is the sea wave height at time is the sea current velocity at time is the sea current velocity at time is the sea wave height at time is the sea current velocity at time is the sea current velocity at time 6. The method of intelligent acquisition and analysis of marine environment data according to claim 5, characterized in that, grid at time repair suitability the calculation formula is: 。 7. The method of intelligent acquisition and analysis of marine environment data according to claim 6, characterized in that, The minimum time required to complete a repair cycle Depending on the type of repair construction, 14 days for coral transplantation repair and 21 days for artificial substrate construction were determined.

8. An intelligent acquisition and analysis system for marine environment data, configured to perform the method of any one of claims 1-7, characterized in that, The system comprises a data collection module, a grid to be repaired identification module and a repair window prediction module; The data acquisition module divides the coral reef monitoring area into 10m 10m rectangular grid, and constructs the acquisition path of the autonomous underwater vehicle with the center coordinates of each grid as the marking points. The autonomous underwater vehicle is used for collecting picture information and environmental data of each grid, the environmental data including water temperature, salinity, pH value and dissolved oxygen; The grid to be repaired identification module is used for identifying the grid to be repaired according to the picture information and environmental data of each grid, calculating the coral reef health index and degradation rate of each grid, obtaining the repair priority of each grid and marking the grid to be repaired; The repair window prediction module is used for establishing a repair suitability evaluation model by combining weather information including sea wave height, sea current speed and typhoon path forecast for the grid to be repaired, and giving a prediction result of the best repair time window through a multi-objective optimization algorithm.

Images