A method for monitoring ocean environment based on low-orbit satellites

Abstract

The present invention discloses a marine environment monitoring method based on a low-orbit satellite, which belongs to the field of marine environment monitoring and comprises the following steps: preprocessing multi-source heterogeneous data to obtain a unified characterization vector; performing cluster analysis on the unified characterization vector to obtain a current marine environment pattern; comparing the deviation between the current marine environment pattern and a normal state pattern to determine an abnormal state; classifying the abnormal state to obtain an abnormal type, and classifying the abnormal type into predefined abnormal types to obtain a classified abnormal state; predicting an abnormal development trend of the classified abnormal state, and evaluating an abnormal diffusion risk based on the predicted abnormal development trend; generating a marine environment abnormal event report for the current sea area based on the abnormal type, abnormal development trend and abnormal diffusion risk; and dynamically monitoring the marine environment abnormal event reports for several sea areas based on a distributed computing framework to obtain a multi-level marine environment analysis ranging from short-term fluctuations in local sea areas to long-term trends on a global scale.

Description

Technical Field

[0001] The present invention belongs to the technical field of marine environment monitoring, and in particular relates to a marine environment monitoring method based on low-orbit satellites. Background Art

[0002] Marine environmental monitoring is a vital area of ​​global ecological protection and resource management, directly related to climate change, biodiversity, and the sustainable development of the marine economy. As human activities increasingly impact marine ecosystems, real-time and accurate monitoring of marine environmental parameters has become an indispensable requirement. Low-orbit satellites, with their wide coverage and high data acquisition efficiency, demonstrate irreplaceable value in this area. However, current marine environmental monitoring methods still face significant limitations. Traditional monitoring methods often rely on a single data source, such as buoys or shipborne equipment, resulting in narrow data coverage and insufficient timeliness. Existing satellite monitoring systems often lack accuracy when processing multi-source data fusion and complex environmental feature identification, making it difficult to meet diverse monitoring needs.

[0003] These limitations point to deeper technical challenges. First, the effective integration of multi-dimensional data is a core bottleneck. The types of parameters such as ocean temperature, salinity, and chlorophyll concentration transmitted back by low-orbit satellites are diverse and heterogeneous with multi-source remote sensing information, making it difficult to achieve efficient integration. Secondly, the automatic identification of marine anomalies faces difficulties. Due to the large differences in sea area characteristics, traditional algorithms are difficult to adapt adaptively, resulting in frequent misjudgments or missed judgments. Finally, the dynamic analysis capabilities of spatiotemporal data are insufficient. The existing system lacks an efficient computing framework when processing continuous time series and wide-area spatial information, which limits the real-time and comprehensiveness of monitoring. These unresolved technical factors make it difficult for monitoring systems to achieve high precision and intelligence in complex marine environments.

[0004] Therefore, how to build a multidimensional data analysis system that can efficiently process multi-source heterogeneous data, automatically identify ocean anomalies and adapt to the characteristics of different sea areas has become a key issue in improving the ocean environment monitoring capabilities of low-orbit satellites. Summary of the Invention

[0005] In order to solve the above technical problems, the present invention proposes a method for monitoring the ocean environment based on low-orbit satellites to solve the problems existing in the above-mentioned prior art.

[0006] To achieve the above objectives, the present invention provides a method for monitoring the ocean environment based on a low-orbit satellite, comprising:

[0007] Preprocess multi-source heterogeneous data to obtain a unified representation vector;

[0008] Performing cluster analysis on the unified characterization vector using a clustering algorithm to obtain a current ocean environment model;

[0009] using an anomaly detection algorithm to compare the deviation between the current ocean environment pattern and the normal state pattern to determine an abnormal state;

[0010] Using a deep learning model to classify the abnormal state to obtain an abnormal type, and classifying the abnormal type into a predefined abnormal type to obtain a classified abnormal state;

[0011] Using a time series analysis algorithm to predict the abnormal development trend of the classified abnormal state, and assessing the abnormal spread risk based on the predicted abnormal development trend;

[0012] Generate a marine environmental abnormal event report for the current sea area based on the abnormal type, abnormal development trend and abnormal spread risk;

[0013] A distributed computing framework is constructed, and based on the distributed computing framework, dynamic monitoring of reports of abnormal marine environmental events in several sea areas is carried out to obtain a multi-level marine environmental analysis ranging from short-term fluctuations in local sea areas to long-term trends at the global scale.

[0014] Optionally, the process of preprocessing multi-source heterogeneous data to obtain a unified representation vector includes:

[0015] Performing standardization processing on the multi-source heterogeneous data to obtain standardized data;

[0016] Using a principal component analysis algorithm to extract key features from the standardized data to obtain a feature set;

[0017] Performing weighted fusion on the feature set using preset weights to obtain an initial representation vector;

[0018] Determining the distribution regularity of the initial characterization data to obtain an environmental state characterization;

[0019] A unified representation vector of multi-source data is generated based on the environmental state representation.

[0020] Optionally, the process of performing cluster analysis on the unified characterization vector using a clustering algorithm to obtain the current ocean environment model includes:

[0021] Constructing a multidimensional feature space based on the unified representation vector, and grouping the data points in the multidimensional feature space using a clustering algorithm to obtain a preliminary grouping result;

[0022] Calculate the distance between classes based on the preliminary grouping result to obtain a distance calculation result;

[0023] Analyze the intra-class closeness based on the distance calculation result to obtain closeness analysis data;

[0024] Determining an optimal number of clusters based on the compactness analysis data and the distance calculation result to obtain a cluster number representation;

[0025] Adopting the cluster number representation to adjust the preliminary classification result to obtain an optimized grouping set;

[0026] Determine the ocean environment pattern according to the optimized grouping set and obtain the environmental pattern representation;

[0027] The distribution law of the multi-dimensional feature space is generated by environmental pattern representation to obtain the current ocean environment pattern.

[0028] Optionally, the process of using an anomaly detection algorithm to compare the deviation between the current ocean environment pattern and the normal state pattern to determine the abnormal state includes:

[0029] Build a normal state pattern library through historical data and obtain the pattern library representation;

[0030] The model library representation is compared with the current ocean environment model to obtain the deviation degree data;

[0031] Anomaly detection algorithms are used to process the deviation degree data to obtain anomaly detection results;

[0032] If the abnormality detection result exceeds the preset threshold, it is determined to be an abnormal state.

[0033] Optionally, the process of obtaining the classified abnormal state includes:

[0034] Obtain abnormal feature sets through historical data, train them with deep learning models, and obtain type representations;

[0035] Extract feature data from the current abnormal state, apply a pattern matching algorithm to the feature data, and determine the degree of match with the type representation;

[0036] Classifying the abnormal state into a predefined abnormal type based on the matching degree to obtain a classified abnormal state;

[0037] After the abnormal state is classified, the boundary conditions of the type representation are adjusted to obtain the updated type definition;

[0038] The updated type definition is used to re-compare abnormal features in historical data to obtain optimized feature extraction rules.

[0039] Optionally, the process of assessing the risk of anomaly spread includes:

[0040] Obtain time series data through classified abnormal states;

[0041] Using a time series analysis algorithm to process the time series data to obtain a change sequence of abnormal states;

[0042] Extracting dynamic characteristics of environmental parameters based on the change sequence of the abnormal state to determine the development trend of the abnormal state;

[0043] Obtain diffusion characteristics from environmental parameters of spatially adjacent areas and determine the distribution range of diffusion characteristics;

[0044] Based on the development trend and diffusion characteristics of the abnormal state, the level of abnormal diffusion risk is obtained.

[0045] Optionally, the process of generating a marine environmental abnormal event report of the current sea area based on the abnormality type, abnormality development trend and abnormality diffusion risk includes:

[0046] By combining the anomaly type and the risk of anomaly spread, we can obtain the response strategies from the pre-set strategy library and determine the solution with the highest matching degree.

[0047] Extract dynamic features from the abnormal development trend, determine the correlation between abnormal diffusion risk and abnormal development trend, and obtain the preliminary distribution of risk levels;

[0048] Based on the preliminary distribution of risk levels and anomaly types, adjust the priority of response strategies to obtain optimized strategic plans.

[0049] Optionally, the process of dynamically monitoring reports of marine environmental anomalies in several sea areas based on the distributed computing framework to obtain a multi-level marine environmental analysis ranging from short-term fluctuations in local sea areas to long-term trends at the global scale includes:

[0050] By collecting environmental data from different sea areas, we can obtain a distribution set of sea area characteristics and environmental differences;

[0051] Obtain the distribution set of sea area characteristics and environmental differences through monitoring needs and determine data characteristics;

[0052] Clustering algorithms are used to adjust the aggregation granularity according to data characteristics to obtain optimized data aggregation results;

[0053] Obtain characteristic distributions of short-term fluctuations and long-term trends based on the optimized data aggregation results, and determine the analysis level;

[0054] If the short-term fluctuation exceeds the short-wave threshold, the analysis algorithm is adjusted through feedback to obtain the change characteristics of the local sea area;

[0055] If the long-term trend deviates from the long-wave range, the global-scale variation pattern is extracted through a convolutional neural network to determine the adjusted trend distribution;

[0056] A multi-level analysis method is used based on the changing characteristics of local sea areas and the adjusted trend distribution to obtain a dynamic monitoring summary from local to global.

[0057] Compared with the prior art, the present invention has the following advantages and technical effects:

[0058] The present invention discloses an intelligent ocean environment anomaly monitoring method based on low-orbit satellites. The method constructs a multi-dimensional ocean environment feature space to perform standardization processing and feature fusion on multi-source heterogeneous data such as temperature, salinity, and chlorophyll concentration transmitted by satellites. The present invention uses cluster analysis to determine the current ocean environment pattern, and compares it with the historical normal state pattern library to achieve anomaly detection. For the detected abnormal state, the present invention uses a deep learning model to classify it, and combines time series analysis to predict the abnormal development trend. According to the anomaly type, trend and diffusion risk, the present invention automatically generates response strategies and abnormal event reports. Through a distributed computing framework and an adaptive parameter adjustment mechanism, the present invention realizes real-time dynamic monitoring of a large range of sea areas, and can optimize the anomaly detection effect according to the characteristics of different sea areas. This method significantly improves the accuracy and efficiency of marine environmental monitoring, and provides strong support for marine ecological protection and resource management. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] The accompanying drawings, which constitute part of this application, are intended to provide a further understanding of this application. The exemplary embodiments and descriptions of this application are intended to explain this application and do not constitute an improper limitation on this application. In the accompanying drawings:

[0060] Figure 1 This is a flow chart of a method for monitoring the ocean environment based on a low-orbit satellite according to an embodiment of the present invention. DETAILED DESCRIPTION

[0061] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0062] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0063] Example 1

[0064] like Figure 1 As shown, this embodiment provides a method for monitoring the ocean environment based on a low-orbit satellite, comprising the following steps:

[0065] S101. Use the data preprocessing module to standardize the multi-source heterogeneous data such as ocean temperature, salinity, and chlorophyll concentration transmitted by low-orbit satellites. Use the feature extraction algorithm to extract the key features of each type of data, and perform feature fusion according to preset weights to obtain a unified representation vector.

[0066] Furthermore, the process of preprocessing multi-source heterogeneous data to obtain a unified representation vector includes: standardizing the multi-source heterogeneous data to obtain standardized data; using the principal component analysis algorithm to extract key features of the standardized data to obtain a feature set; using preset weights to perform weighted fusion on the feature set to obtain an initial representation vector; judging the distribution law of the initial representation data to obtain an environmental state representation; and generating a unified representation vector for multi-source data based on the environmental state representation.

[0067] Furthermore, as a specific implementation of this embodiment, standardization processing is performed on multi-source heterogeneous data such as ocean temperature, salinity data, and chlorophyll concentration through a data preprocessing module in order to eliminate differences in different data dimensions and ranges.

[0068] Ocean temperature is expressed in degrees Celsius, ranging from -2 to 30 degrees Celsius, while salinity is expressed in parts per thousand, ranging from 30 to 40, and chlorophyll concentration is expressed in micrograms per liter, ranging from 0.1 to 10. Normalization maps these data to a distribution with mean 0 and variance 1, making them easier to analyze.

[0069] Principal component analysis (PCA) is used to extract key features from standardized data. The goal is to reduce redundant information and focus on the main components that explain data variability. The standardized ocean temperature, salinity, and chlorophyll concentration dataset contains 1,000 samples, each with three variables. PCA reveals that the first principal component explains 70% of the variation, likely due to a high correlation between temperature and salinity, while the second principal component explains 20% and is related to chlorophyll concentration. This results in a feature set composed of principal components. The advantage of this method is that it can extract key information from multidimensional data while reducing computational complexity. The feature set is weighted and fused according to preset weights to produce an initial representation vector, which reflects the varying importance of different features to the final result.

[0070] In a possible implementation, assuming that the weight of the first principal component is set to 0.6 and the weight of the second principal component is set to 0.4, a comprehensive vector is obtained after weighting.

[0071] For a sample with a first principal component value of 2 and a second principal component value of 1, the initial representation vector value is 2 × 0.6 + 1 × 0.4 = 1.6. This weighted fusion method can highlight features with greater impact on the marine environment and improve the targeted nature of the representation. If the dimensionality of the initial representation vector exceeds a preset threshold, dimensionality reduction is used to adjust the dimensionality and obtain an optimized representation vector.

[0072] In this specific implementation, the initial vector dimension is 10, and the threshold is 5. The vector is then adjusted by retaining the first five principal components or using other dimensionality reduction methods such as t-SNE. This dimensionality reduction not only reduces complexity but also preserves key information, improving the efficiency of subsequent processing. A clustering algorithm is then used to determine the distribution patterns among the data for the optimized representation vectors, yielding classification results.

[0073] Using K-means clustering, the data was divided into three categories: high temperature and high salinity, low temperature and low salinity, and intermediate conditions. Assuming a sample's optimized representation vector value is 1.5, it would be classified as high temperature and high salinity after clustering. This classification reveals underlying patterns in ocean data and provides a basis for environmental state analysis. By matching the classification results with a pre-defined ocean environment model, a representation of the environmental state is obtained.

[0074] In one embodiment, the pre-set model includes three states: "eutrophic," "normal," and "oligotrophic." If high temperature and high salinity match eutrophication, the area is characterized as eutrophic. This matching intuitively reflects the characteristics of the marine environment and facilitates ecological monitoring. Based on the environmental state representation, a unified representation vector for multi-source data is generated.

[0075] Preferably, the state label can be combined with an optimized representation vector, such as the vector [1.5, eutrophication]. This unified representation integrates information from multiple sources and helps downstream tasks such as prediction or visualization.

[0076] For example, when monitoring algal blooms, this characterization can provide comprehensive data support and improve the accuracy of early warnings.

[0077] S102. Construct a multidimensional ocean environment feature space based on the unified characterization vector, perform cluster analysis on the data points in the feature space using a clustering algorithm, determine the optimal number of clusters by calculating the inter-cluster distance and intra-cluster closeness, and obtain the current ocean environment model.

[0078] Furthermore, the process of using a clustering algorithm to perform cluster analysis on the unified characterization vector to obtain the current ocean environment pattern includes: constructing a multidimensional feature space based on the unified characterization vector, using a clustering algorithm to group the data points in the multidimensional feature space to obtain a preliminary grouping result; calculating the inter-class distance based on the preliminary grouping result to obtain a distance calculation result; analyzing the intra-class closeness based on the distance calculation result to obtain closeness analysis data; determining the optimal number of clusters based on the closeness analysis data and the distance calculation result to obtain a cluster number representation; using the cluster number representation to adjust the preliminary classification result to obtain an optimized grouping set; determining the ocean environment pattern based on the optimized grouping set to obtain an environmental pattern representation; generating the distribution law of the multidimensional feature space through the environmental pattern representation to obtain the current ocean environment pattern.

[0079] Furthermore, as a specific implementation of this embodiment, when constructing a multidimensional feature space by unifying the representation vector, it can be understood as mapping multi-source data such as ocean temperature, salinity, and chlorophyll concentration into a higher-dimensional coordinate system.

[0080] This embodiment has three types of data: temperature range of 15-25 degrees Celsius, salinity between 30-35, and chlorophyll concentration between 0.1-0.5 mg / m3. After normalization, they are mapped into a three-dimensional space, and each data point corresponds to a coordinate.

[0081] In a possible implementation, when using a clustering algorithm to group data points, a K-means algorithm may be selected.

[0082] In this example, the initial number of clusters was set to 3, representing high-temperature, high-salinity areas, low-temperature, low-salinity areas, and chlorophyll-rich areas. After running the algorithm, the data points were assigned to different groups. The preliminary grouping results showed that the high-temperature, high-salinity areas were concentrated in one corner of the space.

[0083] It should be noted that the initial grouping may not be accurate enough and therefore needs further optimization.

[0084] When calculating the inter-class distance based on the preliminary grouping result, this embodiment uses the Euclidean distance as a measure.

[0085] Specifically, the coordinates of the center point of the high-temperature and high-salinity area are (20, 34, 0.2), and the center point of the low-temperature and low-salinity area is (16, 31, 0.3). The distance between the two points is calculated to reflect the separation between the groups.

[0086] Preferably, the average distance from the data points in each group to the center can be counted.

[0087] In one embodiment, the average distance from data points to the center in high-temperature, high-salinity areas is 0.8, while in low-temperature, low-salinity areas it is 1.2. If the preset threshold is 1, the high-temperature, high-salinity areas have a good compactness. This step helps determine whether the grouping is reasonable and avoids overly loose classification. If the compactness analysis data meets the preset threshold, the optimal number of clusters is determined based on the distance calculation results and the compactness analysis data.

[0088] In this example, the initial value was set to 3. However, analysis revealed that the inter-cluster distance was small and the compactness was poor, so an adjustment of 2 was more appropriate, resulting in a cluster number representation. This adjustment better reflects the true distribution patterns of the marine environment. Using the cluster number representation to adjust the initial grouping results can be understood as rerunning the clustering algorithm.

[0089] In this example, by adjusting the number of groups from 3 to 2, the data points are reclassified into high-temperature and low-temperature zones, and the optimized grouping sets better align with actual environmental characteristics. This optimization can improve the reliability of subsequent analysis. When determining the ocean environment model based on the optimized grouping sets.

[0090] We can see that data points in the high-temperature region are mostly distributed in the upper half of the space, while those in the low-temperature region are in the lower half. The final model clearly depicts the data distribution characteristics. This distribution pattern will be helpful for subsequent environmental prediction and monitoring.

[0091] S103: Extract the normal state pattern library of the sea area from historical data, use anomaly detection algorithm to compare the deviation between the current ocean environment pattern and the normal state pattern, and determine it as an abnormal state if the deviation exceeds a preset threshold.

[0092] Furthermore, an anomaly detection algorithm is used to compare the deviation between the current ocean environment pattern and the normal state pattern. The process of determining the abnormal state includes: constructing a normal state pattern library through historical data to obtain the pattern library representation; using the pattern library representation to compare with the current ocean environment pattern to obtain the deviation degree data; using the anomaly detection algorithm to process the deviation degree data to obtain the anomaly detection result; if the anomaly detection result exceeds the preset threshold, it is determined to be an abnormal state.

[0093] Furthermore, as a specific implementation of this embodiment, when constructing a normal state pattern library through historical data, it can be understood as extracting typical features from long-term ocean environment observations.

[0094] This embodiment collects temperature and salinity data from the past five years to form a stable baseline set.

[0095] In one possible implementation, assuming the temperature is typically between 18 and 22 degrees Celsius and the salinity is between 32 and 34 degrees Celsius, a pattern library representation is generated by statistically analyzing the average and range of values ​​to reflect the distribution patterns under normal conditions. This approach facilitates subsequent comparisons. When comparing the pattern library representation with the current model,

[0096] Specifically, the observation data for the current week can be analyzed for differences with the benchmarks in the pattern library.

[0097] The current temperature is 23 degrees Celsius and the salinity is 35, both exceeding the normal range. When calculating the degree of deviation, we can simply express it as a difference, such as a temperature deviation of 1-5 degrees Celsius or a salinity deviation of 1-3. This deviation data intuitively reflects the current state of deviation. When applying anomaly detection algorithms to the degree of deviation data,

[0098] Preferably, deviations exceeding two standard deviations from the mean are flagged as potential anomalies.

[0099] For example, if a temperature deviation of 5 degrees Celsius exceeds the normal fluctuation range, the algorithm will identify it as an anomaly. This method is simple and efficient, and can quickly screen for abnormal conditions. If the anomaly detection result exceeds the preset threshold, it is considered an abnormal condition.

[0100] It should be noted that the threshold can be set according to business needs.

[0101] In one embodiment, a temperature deviation greater than 3 degrees Celsius is considered abnormal. Based on the above example, the current state is determined to be abnormal, and a state representation is generated. This representation provides a basis for subsequent adjustments. Adjusting the grouping criteria for the current mode based on the state representation can be understood as redefining the classification criteria.

[0102] In this example, the original grouping based on fixed temperature and salinity ranges has been adjusted to dynamic ranges based on abnormal data. For example, the upper limit of the high-temperature group has been increased from 22 degrees Celsius to 24 degrees Celsius. This adjusted pattern set better reflects current environmental changes. This adjusted pattern set can be used to analyze the distribution patterns of environmental patterns.

[0103] Specifically, we can observe how the data points cluster under the new grouping.

[0104] In one possible implementation, high-temperature data points are concentrated in the upper ocean, while low-temperature data points are biased toward the deeper layers, generating a distribution representation. This distribution pattern reveals the spatial characteristics of the environment. This distribution representation is used to update the normal state of the pattern library.

[0105] This embodiment incorporates the new high temperature range into the pattern library, replacing the old reference range.

[0106] S104: Classify the detected abnormal state, use a deep learning model to learn the characteristic patterns of different types of abnormalities from historical abnormal data, and classify the current abnormality into a predefined abnormality type through a pattern matching algorithm.

[0107] Furthermore, the process of obtaining the classified abnormal state includes: obtaining an abnormal feature set through historical data, training the abnormal feature set using a deep learning model, and obtaining a type representation; extracting feature data from the current abnormal state, applying a pattern matching algorithm to the feature data, and determining the degree of matching with the type representation; classifying the abnormal state into a predefined abnormal type based on the degree of matching to obtain a classified abnormal state; adjusting the boundary conditions of the type representation of the classified abnormal state to obtain an updated type definition; and using the updated type definition to re-match the abnormal features in the historical data to obtain an optimized feature extraction rule.

[0108] Furthermore, as a specific implementation of this embodiment, when obtaining an abnormal feature set through historical data, it can be understood as screening out sample data that deviates from the normal state from long-term marine environment records.

[0109] Assuming that there have been multiple instances of sudden temperature increases or salinity drops in the past decade, these anomalies are grouped into a feature set.

[0110] Specifically, a record shows a temperature reaching 25 degrees Celsius and a salinity drop to 30, significantly deviating from the normal range. These data points form the basis of anomaly features. When using a deep learning model to train anomaly features, one possible implementation is to learn the inherent patterns of anomalies through a neural network.

[0111] In this embodiment, temperature, salinity, and time series are input into the model, and after training, a type representation is generated, which may be a "high temperature and low salinity type" or a "transient fluctuation type".

[0112] In one embodiment, the model identifies that high temperature anomalies are often accompanied by changes in surface water flow, forming a unique type of characterization. When extracting feature data from the current anomaly.

[0113] Preferably, this embodiment focuses on key indicators in real-time monitoring. For example, if the current observed temperature is 24 degrees Celsius and the salinity is 31, these values ​​are extracted as feature data. Applying a pattern matching algorithm to this feature data can be understood as comparing the current data with a trained type representation.

[0114] This embodiment calculates the similarity between the current feature and the "high temperature and low salt type" and obtains a matching degree of 85%. If the matching degree exceeds a preset threshold, such as 80%, the current anomaly is determined to belong to the corresponding type.

[0115] It's important to note that this classification provides a clear direction for subsequent processing. For example, classifying the current anomaly as "high temperature, low salinity" helps quickly locate the source of the problem. This is especially true when adjusting the boundary conditions for the type representation based on the classification results.

[0116] In one embodiment, if multiple high-temperature anomalies exceed 24 degrees Celsius, the original boundary of 23 degrees Celsius can be adjusted to 25 degrees Celsius, generating an updated type definition. This adjustment makes the type more closely aligned with actual changes. The updated type definition is then used to re-compare anomaly features in historical data.

[0117] This embodiment re-examines past data and finds that some unidentified anomalies now conform to the new boundaries, so the optimized feature extraction rules are more accurate.

[0118] In one embodiment, the new rule prioritizes the time window of temperature mutation to select more effective features. When the optimized feature extraction rule is used to process the current anomaly, it can be determined whether the algorithm application strategy needs to be adjusted.

[0119] S105. For the classified abnormal state, extract the recent environmental parameter change sequence of the sea area from the time dimension, use the time series analysis algorithm to predict the abnormal development trend, and evaluate the abnormal spread risk in combination with the environmental status of the spatially adjacent areas.

[0120] Furthermore, the process of assessing the risk of abnormal diffusion includes: obtaining time series data through classified abnormal states; processing the time series data using a time series analysis algorithm to obtain a change sequence of the abnormal state; extracting the dynamic characteristics of environmental parameters based on the change sequence of the abnormal state to determine the development trend of the abnormal state; obtaining diffusion characteristics from environmental parameters in spatially adjacent areas to determine the distribution range of the diffusion characteristics; and obtaining the level of abnormal diffusion risk based on the development trend and diffusion characteristics of the abnormal state.

[0121] Furthermore, as a specific implementation method of this embodiment, time series data is obtained through the classified abnormal state, and a time series analysis algorithm is used for processing to obtain a change sequence of the abnormal state. The dynamic characteristics of the environmental parameters are extracted based on the change sequence to determine the development trend of the abnormal state. The diffusion characteristics are obtained from the environmental parameters of the spatially adjacent areas, and the distribution range of the diffusion characteristics is determined. A cluster analysis algorithm is used for processing based on the development trend and diffusion characteristics to obtain the diffusion risk level of the abnormal state. If the diffusion risk level exceeds the preset threshold, the boundary conditions of the time series are updated through the regional environmental data to obtain an adjusted change sequence. The sequence analysis is re-run based on the adjusted change sequence to obtain an optimized development trend. The classification rules of the abnormal state are updated through the optimized development trend and diffusion characteristics to obtain an improved state classification result.

[0122] S106. Based on the anomaly type, development trend, and diffusion risk, call the pre-set response strategy library, select the response plan with the highest matching degree, and generate an abnormal event report containing anomaly description, impact assessment, and recommended measures.

[0123] Furthermore, the process of generating a report on abnormal marine environmental events in the current sea area based on the abnormality type, abnormal development trend and abnormal diffusion risk includes: obtaining response strategies in a pre-set strategy library through a combination of abnormality type and abnormal diffusion risk, and determining the plan with the highest matching degree; extracting dynamic features from the abnormal development trend, judging the correlation between the abnormal diffusion risk and the abnormal development trend, and obtaining a preliminary distribution of risk levels; adjusting the priority of the response strategy according to the preliminary distribution of risk levels and the abnormality type, and obtaining an optimized strategy plan.

[0124] Furthermore, as a specific implementation of this embodiment, when obtaining a response strategy through a combination of anomaly type and diffusion risk, it can be understood as extracting a matching solution from a pre-designed strategy library.

[0125] Assuming the anomaly type is "water pollution", the diffusion risk is divided into three levels: "low, medium, and high". The strategy library includes options such as "local isolation", "regional monitoring", and "comprehensive governance".

[0126] If the anomaly type is "water pollution" and the diffusion risk is "high," the system prioritizes the "comprehensive treatment" solution because its coverage and intensity match the high risk. The degree of matching can be determined based on the success rate of historical cases. For example, when treating pollution in a certain sea area, comprehensive treatment achieved an 80% efficiency in restoring water quality in similar scenarios, making it the optimal solution. This is when extracting dynamic features from development trends.

[0127] As a specific implementation of this example, in a monitoring of a certain sea area, the pollution concentration increased from 10 mg / L to 15 mg / L, with a rate of change increasing by 0.5 mg / L per day. This trend indicates that the anomaly is intensifying. To determine the correlation between diffusion risk and development trends, a time window analysis can be used. For example, the pollution range has expanded from 1 square kilometer to 3 square kilometers in the past 7 days, indicating that the diffusion risk is highly correlated with the rising concentration trend. The initial distribution of risk levels is therefore determined to be "high," providing a basis for subsequent strategy adjustments. When adjusting the priority of response strategies.

[0128] As a specific implementation method of this embodiment, if the risk level is "high" and the abnormality type involves "heavy metal pollution", the priority is adjusted to "emergency isolation" rather than "long-term monitoring" because the former can curb the spread faster. After integrating the abnormality description (such as "heavy metal exceeds the standard") and the impact assessment (such as "affecting a 30% decrease in fishery production"), the generated event summary includes the dynamic feature "diffusion rate of 0.2 square kilometers / day", and the scope of application of the recommended measures is limited to "the core area of ​​pollution and the surrounding 5 kilometers." When using the support vector machine algorithm to process development trends and diffusion risks.

[0129] Preferably, the concentration change rate and diffusion range can be used as input features to predict the distribution of abnormal events.

[0130] In this example, after inputting concentration data and range data over a period of 7 days, the algorithm predicts that pollution will cover an additional 2 square kilometers within 3 days. This predicted distribution provides a basis for determining whether the risk level exceeds the threshold. If the preset threshold is exceeded (for example, the range increases to 5 square kilometers), the policy library will update the response measures, such as upgrading from "local isolation" to "joint governance." The adjusted recommended measures include "increasing the number of monitoring points to 10 and starting purification equipment." When a complete event summary is generated based on the predicted distribution and the optimized policy plan.

[0131] In one example, the anomaly description might be "Heavy metal pollution in a certain sea area, concentration reaching 20 mg / L," and the recommended action might be "Deploy an isolation zone and activate purification equipment, monitoring concentration changes daily." This approach not only allows for rapid response to anomalies but also reduces the long-term impact of spread through dynamic adjustments.

[0132] After the purification equipment was operational, the concentration dropped to 12 mg / L within five days, and the diffusion range stabilized within three square kilometers. This approach, combining prediction and strategy optimization, ensures targeted and efficient response measures, providing reliable support for abnormal management.

[0133] S107. Use a distributed computing framework to process the continuously input marine environment monitoring data stream, improve computing efficiency through data sharding and parallel processing, and realize real-time dynamic monitoring of a large sea area.

[0134] A distributed computing framework is used to capture the marine environment monitoring data stream. Data sharding technology is used to segment the continuously input data, resulting in a sharded data set. Parallel processing tasks are assigned to the sharded data set, and the intermediate results processed by each node are determined. Real-time dynamic features are integrated with these intermediate results to determine the changing trends of the marine environment and obtain a preliminary distribution for dynamic monitoring. A random forest algorithm is used to process this preliminary distribution of dynamic monitoring, identifying abnormal areas within the sea area and determining the range of the abnormal distribution. If the range of the abnormal distribution exceeds a preset threshold, the processing framework adjusts the parallel processing tasks to obtain an optimized intermediate result. Based on the optimized intermediate result, real-time dynamic features are updated, a monitoring adjustment plan for a large sea area is obtained, and the final distribution is determined. This final distribution generates a dynamic monitoring summary for the marine environment, resulting in a complete event processing process.

[0135] As a specific implementation of this embodiment, the distributed computing framework plays a key role in marine environment monitoring, which can be understood as processing massive data streams through multi-node collaboration.

[0136] As a specific implementation of this example, the monitoring system continuously collects data such as temperature, salinity, and ocean current velocity from marine sensors, with a data volume of up to 10GB per second. Using data sharding technology, this data can be segmented by time period, for example, every five minutes, generating multiple data sets of approximately 1GB. This approach facilitates subsequent parallel processing and avoids overloading a single node.

[0137] Preferably, each shard can be assigned to a different computing node, such as one node processing temperature data and another processing ocean current data, to ensure task balance. When obtaining parallel processing task assignments for the sharded data set.

[0138] Specifically, it can be dynamically adjusted according to node performance.

[0139] For example, high-performance nodes process shards with large fluctuations in ocean current speed, while low-performance nodes handle stable temperature data. Intermediate results are calculated by each node, representing local trends, such as a 0.5°C / hour increase in temperature in a particular sea area. By integrating these intermediate results, real-time dynamic features can be extracted. For example, determining whether accelerated ocean currents in a particular area are associated with temperature anomalies can be determined. Preliminary analysis indicates that 10% of the sea area exhibits abnormal trends. This distributed approach significantly improves processing efficiency.

[0140] In one embodiment, when the random forest algorithm processes the preliminary distribution, it can train a model based on historical data to identify abnormal areas.

[0141] As a specific implementation of this example, the model predicts an outlier in a certain sea area when the temperature exceeds 28 degrees Celsius and the ocean current speed increases by 20%. The anomaly distribution covers 50 square kilometers. If it exceeds a preset threshold, such as 30 square kilometers, the processing framework will increase node resources, for example, from 5 nodes to 8, and optimize intermediate results. The updated dynamic features show that the anomaly area has expanded to 60 square kilometers, and the monitoring adjustment plan will then cover a larger area of ​​the sea.

[0142] It should be noted that the spatial and temporal dimensions can be combined when the final distribution generates a dynamic monitoring summary.

[0143] As a specific implementation of this example, consider an abnormal area in a certain sea area spreading from east to west at a speed of 2 km / h. The summary will record this trend and recommend dispatching additional monitoring vessels to the western waters. The complete event handling process can also reveal the cause of the anomaly, such as hydrothermal activity. This approach not only improves monitoring accuracy but also provides a basis for emergency decision-making.

[0144] As a specific implementation of this example, if a monitoring session reveals a sudden drop in salinity of 5% in an abnormal area, multi-faceted analysis can infer freshwater influx. Distributed computing confirms the drop is within a 20 square kilometer area, while random forest predicts its spread to 30 square kilometers. The adjustment plan recommends deploying buoys to track the source of the freshwater. Multiple examples support each other, such as temperature, salinity, and ocean current data all pointing to the same abnormal event, ensuring analytical consistency. This process enables rapid response to changes in the marine environment and ensures comprehensive monitoring.

[0145] S108. In view of the differences in environmental characteristics of different sea areas, an adaptive parameter adjustment mechanism is constructed to continuously optimize the anomaly detection threshold, feature extraction weights and prediction model parameters through feedback learning, thereby improving the system's adaptability to complex marine environments.

[0146] By collecting environmental data from different sea areas, a distribution set of sea area characteristics and environmental differences is obtained. Based on this distribution set, a feedback learning method is used to adjust the weights of feature extraction to obtain an optimized feature set. For this optimized feature set, intermediate parameters of the prediction model are obtained to determine preliminary environmental change trends. If the preliminary trend indicates that anomaly detection exceeds the preset detection threshold, the model parameters are updated through feedback learning to obtain adjusted prediction results. Based on the adjusted prediction results, the distribution range of anomalies in complex environments is determined to determine whether further optimization is needed. Based on the anomaly distribution range, an adaptive parameter adjustment mechanism is used to update the detection threshold to obtain the final environmental adaptation plan. Based on the final plan, a dynamic monitoring summary corresponding to the sea area characteristics is obtained, and a complete event handling process is determined.

[0147] As a specific implementation of this embodiment, by collecting environmental data of different sea areas, a distribution set of sea area characteristics and environmental differences is obtained.

[0148] For example, data on temperature, salinity, and ocean current speeds were collected from multiple sea areas, including the East China Sea and the South China Sea, to form a preliminary distribution set encompassing multiple environmental variables. These data show that surface water temperatures in the East China Sea fluctuate around 25°C, while those in the South China Sea are generally higher, reaching over 28°C. Salinity also varies depending on rainfall.

[0149] As a specific implementation of this embodiment, this distribution set can intuitively reflect the characteristic differences of different sea areas and provide a basis for subsequent analysis. According to the distribution set, the feedback learning method is used to adjust the weight of feature extraction to obtain an optimized feature set.

[0150] Specifically, feedback learning can dynamically adjust the weight of each variable by comparing historical data with current data. For example, if the initial model assumes that temperature accounts for 40% of environmental changes, but multiple feedback loops reveal that salinity changes are more representative of abnormal trends, the weight of salinity can be increased to 50% and temperature reduced to 30%.

[0151] In one possible implementation, a region in the East China Sea is assigned a higher weight due to a sudden change in salinity. The optimized feature set can better highlight salinity-driven environmental differences. Intermediate parameters of the prediction model are obtained for this optimized feature set, and preliminary environmental change trends are determined.

[0152] Preferably, the prediction model is based on time series analysis and combines the data of the previous three months to infer the temperature decline trend of a certain sea area in the next 15 days.

[0153] As a specific implementation of this embodiment, due to the influence of cold air, the temperature in a certain area of ​​the East China Sea drops from 25 degrees to 22 degrees, and the intermediate parameters will record this change rate.

[0154] It is understandable that these parameters provide a basis for subsequent anomaly detection. If the initial trend shows that the anomaly detection exceeds the preset detection threshold, the model parameters are updated through feedback learning to obtain the adjusted prediction results.

[0155] It should be noted that, assuming the threshold is set to a temperature change of more than 3 degrees per week, an update will be triggered if the temperature in a certain area of ​​the East China Sea drops by 4 degrees.

[0156] In one embodiment, the model reanalyzes ocean current and wind speed data and finds that weakening ocean currents are the primary cause, adjusting parameters to make the prediction more realistic. Based on the adjusted prediction results, the distribution range of anomalies in complex environments is determined.

[0157] For example, a typhoon caused both temperature and salinity anomalies in a region of the South China Sea, covering 200 square kilometers. Through multi-faceted analysis, such as combining data on wind speeds of 20 meters per second and rainfall of 50 millimeters per day, the anomaly was determined to be driven by short-term weather rather than a long-term trend. This multi-dimensional support ensured the accuracy of the distribution range. Based on the anomaly distribution range, an adaptive parameter adjustment mechanism was used to update the detection threshold and ultimately determine the environmental adaptation plan.

[0158] In one embodiment, if the abnormal range continues to expand, the threshold can be adjusted from 3 degrees / week to 4 degrees / week to avoid frequent false alarms.

[0159] As a specific implementation of this embodiment, the adjusted solution can better adapt to the high volatility environment during typhoon season and improve the robustness of monitoring. According to the final solution, a dynamic monitoring summary corresponding to the sea area characteristics is obtained and a complete event processing process is determined.

[0160] As a specific implementation of this embodiment, the monitoring summary of a certain area in the East China Sea shows that "the temperature decline trend has slowed down, and salinity anomalies require continued attention." The process includes data collection, feature optimization, anomaly confirmation, and other links.

[0161] Preferably, this approach can capture environmental changes in a timely manner and provide support for response measures.

[0162] S109. Design a multi-scale data analysis module to automatically select the appropriate data aggregation granularity and analysis algorithm for monitoring needs in different spatial ranges and time spans, and realize multi-level marine environmental analysis from short-term fluctuations in local sea areas to long-term trends at the global scale.

[0163] Based on monitoring requirements, a distribution set of spatial scope and time span is obtained to determine data characteristics. Clustering algorithms are used to adjust the aggregation granularity based on the data characteristics to obtain optimized data aggregation results. Based on the optimized data aggregation results, the characteristic distribution of short-term fluctuations and long-term trends is obtained to determine the analysis level. If short-term fluctuations exceed the preset threshold, the analysis algorithm is adjusted through feedback to obtain the change characteristics of the local sea area. If the long-term trend deviates from the preset range, a convolutional neural network is used to extract the global-scale change pattern and determine the adjusted trend distribution. Based on the change characteristics and trend distribution, a multi-level analysis method is used to obtain a dynamic monitoring summary from the local to the global level. The aggregation granularity and analysis algorithm are updated based on the dynamic monitoring summary to determine the final multi-scale analysis results.

[0164] As a specific implementation of this embodiment, a distribution set of spatial range and time span is obtained by monitoring demand to determine data characteristics.

[0165] It is understandable that the spatial range covers tens of kilometers in local sea areas to thousands of kilometers on a global scale, and the time span ranges from short-term monitoring at the hourly level to long-term analysis at the annual level.

[0166] As a specific implementation of this example, in a region of the East China Sea, monitoring requirements focused on ocean current changes within a 50-kilometer radius over a 72-hour period. By collecting temperature, salinity, and current velocity data, it was determined that the data characteristics exhibited a coexistence of high-frequency fluctuations and low-frequency trends. Based on these data characteristics, a clustering algorithm was used to adjust the aggregation granularity, resulting in optimized data aggregation results.

[0167] As a specific implementation of this example, K-means clustering can be used to group similar data points. In the aforementioned East China Sea scenario, if temperature data is collected every 10 minutes, the 72-hour data can be clustered by hour to reduce noise interference and adjust the aggregation granularity from minute to hour. This allows for a clearer capture of ocean current patterns. Based on the optimized data aggregation results, the characteristic distribution of short-term fluctuations and long-term trends is obtained to determine the analysis level.

[0168] Specifically, short-term fluctuations can be captured by calculating the standard deviation of hourly data. For example, a temperature fluctuation exceeding 0.5 degrees Celsius is considered an anomaly. Long-term trends are extracted using a moving average method, such as a gradual decrease in salinity by 0.2 units over the past 30 days. The analysis is therefore divided into two dimensions: local anomaly detection and global trend assessment. If short-term fluctuations exceed a preset threshold, feedback is used to adjust the analysis algorithm, revealing the characteristics of changes in the local ocean area.

[0169] In one possible implementation, if the hourly temperature standard deviation reaches 0.8 degrees Celsius, exceeding a threshold of 0.5, an adaptive filter can be introduced to remove incidental noise and recalculate the data, yielding the signature of localized ocean current acceleration. This adjustment can improve the accuracy of anomaly identification. If the long-term trend deviates from a preset range, a convolutional neural network is used to extract global-scale patterns of change and determine the adjusted trend distribution.

[0170] As a specific implementation of this example, the salinity of the East China Sea dropped by 0.1 units beyond the expected range. A convolutional neural network was used to analyze global sea temperature and salinity image data, extracting the changing patterns under the influence of El Niño and adjusting trend forecasts to be more realistic. Based on the changing characteristics and trend distribution, a multi-level analysis method was used to obtain a dynamic monitoring overview from the local to the global level.

[0171] Optimally, local features indicating accelerated ocean currents and global trends reflecting the influence of climate patterns are combined to generate a dynamic summary, such as "localized ocean current anomalies in the East China Sea are driven by global warming." This provides a comprehensive perspective for multi-scale monitoring. The dynamic monitoring summary updates the aggregation granularity and analysis algorithms, determining the final multi-scale analysis results.

[0172] In one embodiment, if the summary shows frequent short-term anomalies, the aggregation granularity can be shortened from hourly to half-hourly, while algorithm parameters are optimized to improve responsiveness to rapid changes. This adaptive update can significantly improve the real-time and reliability of monitoring.

[0173] The above are merely preferred embodiments of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A method for monitoring the ocean environment based on a low-orbit satellite, characterized in that: The following steps are involved: Preprocessing multi-source heterogeneous data to obtain a unified representation vector; the process of preprocessing multi-source heterogeneous data to obtain a unified representation vector includes: standardizing the multi-source heterogeneous data to obtain standardized data; extracting key features from the standardized data using a principal component analysis algorithm to obtain a feature set; weightedly fusing the feature set using preset weights to obtain an initial representation vector; determining the distribution law of the initial representation vector to obtain an environmental state representation; and generating a unified representation vector for multi-source data based on the environmental state representation; A clustering algorithm is used to perform cluster analysis on the unified characterization vector to obtain a current ocean environment model; the process of obtaining the current ocean environment model includes: constructing a multidimensional feature space based on the unified characterization vector, grouping data points in the multidimensional feature space using a clustering algorithm to obtain a preliminary grouping result; calculating the inter-class distance based on the preliminary grouping result to obtain a distance calculation result; analyzing the intra-class closeness based on the distance calculation result to obtain closeness analysis data; determining the optimal number of clusters based on the closeness analysis data and the distance calculation result to obtain a cluster number representation; adjusting the preliminary classification result using the cluster number representation to obtain an optimized grouping set; determining the ocean environment model based on the optimized grouping set to obtain an environmental model representation; generating a distribution law of the multidimensional feature space through the environmental model representation to obtain the current ocean environment model; using an anomaly detection algorithm to compare the deviation between the current ocean environment pattern and the normal state pattern to determine an abnormal state; A deep learning model is used to classify the abnormal state to obtain an abnormal type, and the abnormal type is classified into a predefined abnormal type to obtain a classified abnormal state; the process of obtaining the classified abnormal state includes: obtaining an abnormal feature set through historical data, training the abnormal feature set using a deep learning model to obtain a type representation; extracting feature data from the current abnormal state, applying a pattern matching algorithm to the feature data, and determining the degree of matching with the type representation; classifying the abnormal state into a predefined abnormal type based on the degree of matching to obtain a classified abnormal state; adjusting the boundary conditions of the type representation of the classified abnormal state to obtain an updated type definition; and re-comparing the abnormal features in the historical data using the updated type definition to obtain an optimized feature extraction rule; A time series analysis algorithm is used to predict the abnormal development trend of the classified abnormal state, and the abnormal diffusion risk is assessed based on the predicted abnormal development trend. The process of assessing the abnormal diffusion risk includes: obtaining time series data through the classified abnormal state; processing the time series data using a time series analysis algorithm to obtain a change sequence of the abnormal state; extracting dynamic characteristics of environmental parameters based on the change sequence of the abnormal state to determine the development trend of the abnormal state; obtaining diffusion characteristics from environmental parameters of spatially adjacent areas to determine the distribution range of the diffusion characteristics; and obtaining the level of abnormal diffusion risk based on the development trend and diffusion characteristics of the abnormal state. Generate a marine environmental abnormal event report for the current sea area based on the abnormal type, abnormal development trend and abnormal spread risk; A distributed computing framework is constructed, and based on the distributed computing framework, dynamic monitoring of reports of abnormal marine environmental events in several sea areas is carried out to obtain a multi-level marine environmental analysis ranging from short-term fluctuations in local sea areas to long-term trends at the global scale.

2. The method for monitoring the ocean environment based on a low-orbit satellite according to claim 1, wherein: The process of comparing the deviation between the current ocean environment pattern and the normal state pattern using an anomaly detection algorithm and determining the abnormal state includes: Build a normal state pattern library through historical data and obtain the pattern library representation; The model library representation is compared with the current ocean environment model to obtain the deviation degree data; Anomaly detection algorithms are used to process the deviation degree data to obtain anomaly detection results; If the abnormality detection result exceeds the preset threshold, it is determined to be an abnormal state.

3. The method for monitoring the ocean environment based on a low-orbit satellite according to claim 1, wherein: The process of generating a marine environmental abnormal event report for the current sea area based on the abnormality type, abnormal development trend and abnormal diffusion risk includes: By combining the anomaly type and the risk of anomaly spread, we can obtain the response strategies from the pre-set strategy library and determine the solution with the highest matching degree. Extract dynamic features from the abnormal development trend, determine the correlation between abnormal diffusion risk and abnormal development trend, and obtain the preliminary distribution of risk levels; Based on the preliminary distribution of risk levels and anomaly types, adjust the priority of response strategies to obtain optimized strategic plans.

4. The method for monitoring the ocean environment based on a low-orbit satellite according to claim 3, wherein: The process of dynamically monitoring reports of abnormal marine environmental events in several sea areas based on the distributed computing framework to obtain multi-level marine environmental analysis from short-term fluctuations in local sea areas to long-term trends on a global scale includes: By collecting environmental data from different sea areas, we can obtain a distribution set of sea area characteristics and environmental differences; Obtain the distribution set of sea area characteristics and environmental differences through monitoring needs and determine data characteristics; Clustering algorithms are used to adjust the aggregation granularity according to data characteristics to obtain optimized data aggregation results; Obtain characteristic distributions of short-term fluctuations and long-term trends based on the optimized data aggregation results, and determine the analysis level; If the short-term fluctuation exceeds the short-wave threshold, the analysis algorithm is adjusted through feedback to obtain the change characteristics of the local sea area; If the long-term trend deviates from the long-wave range, the global-scale variation pattern is extracted through a convolutional neural network to determine the adjusted trend distribution; A multi-level analysis method is used based on the changing characteristics of local sea areas and the adjusted trend distribution to obtain a dynamic monitoring summary from local to global.

Images