A high-precision detection and identification system for atmospheric gravity waves in the South China Sea region

Abstract

The present invention relates to the field of atmospheric gravity wave detection, and particularly to a high-precision detection and identification system for atmospheric gravity waves in the South China Sea region; Technical problem: The existing atmospheric gravity wave detection and identification systems often have difficulty in accurately identifying the weak signals of atmospheric gravity waves, are easily affected by noise and other interference factors, and when performing large-scale data processing, there are often problems of low processing efficiency and long time consumption; Technical solution: A high-precision detection and identification system for atmospheric gravity waves in the South China Sea region, including a data acquisition module, an identification and analysis module, a real-time feedback module, and a visualization display module; The present invention utilizes artificial intelligence technologies such as deep learning and machine learning, can more accurately identify the characteristics of atmospheric gravity waves from complex data, reduce misjudgment and missed judgment, and through real-time online data processing, can process and analyze the data immediately after it is generated, providing timely support for forecasting and early warning.

Description

Technical Field

[0001] The present invention relates to the field of atmospheric gravity wave detection, and particularly to a high-precision detection and identification system for atmospheric gravity waves in the South China Sea region. Background Art

[0002] The high-precision detection and identification system for atmospheric gravity waves in the South China Sea region is an integrated system that combines various detection means and data processing technologies, aiming to achieve high-precision detection, identification, and analysis of gravity waves in the atmosphere of the South China Sea region. The system may include multiple components such as ground observation stations, upper-air sounding equipment (such as lidar, weather balloons, etc.), satellite remote sensing data reception and processing platforms, etc., integrating a variety of advanced detection means and data processing technologies, providing strong support for atmospheric science research, weather forecasting, climate prediction, and marine meteorological services in the South China Sea region. However, the existing atmospheric gravity wave detection and identification systems often have difficulty in accurately identifying the weak signals of atmospheric gravity waves, are easily affected by noise and other interference factors, and often have problems of low processing efficiency and long processing time when dealing with large-scale data. Summary of the Invention

[0003] In order to overcome the problems that the existing atmospheric gravity wave detection and identification systems often have difficulty in accurately identifying the weak signals of atmospheric gravity waves, are easily affected by noise and other interference factors, and often have problems of low processing efficiency and long processing time when dealing with large-scale data.

[0004] The technical solution of the present invention is as follows: A high-precision detection and identification system for atmospheric gravity waves in the South China Sea region, including:

[0005] A data acquisition module, which is responsible for receiving the original data from various detection devices and performing data cleaning, data compression, and time synchronization processing on the original data;

[0006] An identification and analysis module, which is used to extract features, identify, classify, and estimate parameters of atmospheric gravity waves from the preprocessed data by applying deep learning models and machine learning algorithms;

[0007] A real-time feedback module, which is used to realize real-time processing of data and instant feedback of results, including real-time monitoring of atmospheric gravity waves, generation and release of early warning information;

[0008] A visualization display module, which is used to display the analysis results in the form of intuitive charts and images, support multi-perspective and multi-level data exploration, and provide an interactive interface for researchers and decision-makers to use.

[0009] Preferably, the data acquisition module is responsible for receiving the raw data from various detection devices, and performing data cleaning, data compression, and time synchronization processing on the raw data; the identification and analysis module applies deep learning models and machine learning algorithms to extract features, identify, classify, and estimate parameters of the atmosphere gravity waves from the preprocessed data; the real-time feedback module realizes the real-time processing of data and the instant feedback of results, including the real-time monitoring of atmosphere gravity waves, the generation and release of early warning information; the visualization display module displays the analysis results in the form of intuitive charts and images, supports multi-perspective and multi-level data exploration, and provides an interactive interface for researchers and decision-makers to use.

[0010] Preferably, the data acquisition module includes a detection device group and a preprocessing unit; the detection device group includes lidar, satellite remote sensing devices, and environmental sensors, which are used to collect various environmental data, including the raw data of meteorological information such as atmospheric temperature, humidity, wind speed, and wind direction; the preprocessing unit uses data cleaning to remove the noise and outliers of the raw data, uses data compression technology to reduce the burden of data storage and transmission, uses time synchronization technology to ensure the time consistency of multi-source data, and the preprocessing unit integrates an adaptive filtering algorithm to dynamically adjust the filtering parameters according to the data characteristics and optimize the preprocessing effect.

[0011] Preferably, when the data acquisition module is working, it includes the following steps:

[0012] S101: The data acquisition module receives the raw data streams from lidar, satellite remote sensing devices, and environmental sensors;

[0013] S102: Perform a preliminary verification on the received data to check the integrity, format correctness, and validity of the timestamp of the data;

[0014] S103: Set the initial parameters of the filter, including the filter order and cut-off frequency; determine the parameter adjustment strategy, including the gradient descent method and the least squares method;

[0015] S104: Input the raw data to be processed into the filter;

[0016] S105: Apply the filter to denoise the data, use the parameter adjustment strategy to calculate the new filter parameters, and update the filter state. The parameters of the filter will be dynamically adjusted according to the characteristics of the input data;

[0017] S106: Output the filtered data, that is, the data after removing noise and outliers;

[0018] S107: Select the LZ4 compression algorithm to compress the cleaned data to reduce the requirements for storage space and transmission bandwidth;

[0019] S108: Use the timestamp calibration technology to align the data of different devices to the same time reference and perform time synchronization processing on the data from different detection devices;

[0020] S109: Use a distributed database to store the data.

[0021] Preferably, the recognition and analysis module uses a convolutional neural network model and a clustering analysis algorithm to automatically extract features, identify, classify, and estimate parameters of atmospheric gravity waves from the preprocessed data. Through transfer learning technology, a pre-trained model containing known atmospheric gravity wave events is used to accelerate the convergence speed of the atmospheric gravity wave recognition model.

[0022] Preferably, when the recognition and analysis module is working, it includes the following steps:

[0023] S201: Receive the multi-source meteorological data that has been cleaned, compressed, and time-synchronized from the data acquisition module;

[0024] S202: Convert the meteorological data into a format suitable for processing by the CNN model;

[0025] S203: Construct a CNN architecture according to the task requirements, including an input layer, multiple convolutional layers, a fully connected layer, and an output layer;

[0026] S204: Perform a sliding window operation on the input data through multiple convolutional kernels to extract local features; a non-linear activation function ReLU and a pooling layer follow each convolutional layer;

[0027] S205: After multiple convolutional and pooling layers, generate a feature map containing high-level abstract features;

[0028] S206: Select and further extract features from the original data or the feature map extracted by the CNN to construct a feature set for clustering;

[0029] S207: Standardize or normalize the selected features;

[0030] S208: Use known atmospheric gravity wave events to perform supervised training on the CNN model, adjust the model parameters through the backpropagation algorithm, and minimize the prediction error;

[0031] S209: Apply the clustering algorithm K-means to perform clustering analysis on the feature set in an unsupervised or weakly supervised situation to identify different atmospheric gravity wave patterns or categories;

[0032] S210: Optimize the performance of the CNN model and the clustering algorithm through cross-validation and hyperparameter tuning;

[0033] S211: Automatically extract features and classify the input data using the trained CNN model to identify atmospheric gravity wave events;

[0034] S212: For the identified atmospheric gravity wave events, estimate their wave speed, wavelength, and amplitude parameters in combination with the results of cluster analysis;

[0035] S213: Output the identification and analysis results.

[0036] Preferably, the real-time feedback module uses the stream processing technology Apache Kafka to ensure the real-time nature of data processing, adopts a parallel computing framework to improve the processing speed, and constructs a distributed processing architecture; through an intelligent early warning algorithm, combines historical data with the current trend to predict the influence range and intensity of atmospheric gravity waves, and realizes the real-time processing of data and the instant feedback of results, including real-time monitoring of atmospheric gravity waves, generation and release of early warning information.

[0037] Preferably, when the real-time feedback module is working, it includes the following steps:

[0038] S301: Use Kafka as the message queue, and the data receives real-time data streams from various detection devices through the Apache Kafka stream processing platform;

[0039] S302: Load the pre-trained deep learning model, use GPU acceleration to perform forward propagation on the input data to obtain the identification results and parameter estimation values of atmospheric gravity waves; adopt a parallel processing algorithm to parallel process data using multi-GPU or multi-CPU resources;

[0040] S303: Integrate the output data on atmospheric gravity waves with the relevant data in the historical database;

[0041] S304: Write Python scripts or use data preprocessing libraries to extract and construct features; among them, use the groupby and transform methods of Pandas to calculate seasonal change features, and use the array operations of NumPy to calculate similar events, seasonal changes, and geographical location features in historical data;

[0042] S305: Extract similar event features, seasonal change features, and geographical location features in historical data that are useful for predicting the future trend of atmospheric gravity waves from the integrated data;

[0043] S306: Select the GBDT implementation library GradientBoostingRegressor, set its initial parameters, and use the fit method of the model to train using the training set data;

[0044] S307: Call the predict method of the model using the test set data for prediction, and calculate the prediction error; adjust the model parameters according to the evaluation results, increase the number of trees, and adjust the learning rate;

[0045] S308: Use the GBDT model to analyze the influence of features on the prediction results;

[0046] S309: Set the warning thresholds for atmospheric gravity waves based on the predicted values of amplitude, frequency, and propagation speed parameters. When the prediction results exceed these thresholds, the system will trigger the warning mechanism;

[0047] S310: When the prediction results of the GBDT model trigger the warning conditions, the system generates warning information with the prediction results of the GBDT model and the warning thresholds as inputs according to the preset format and template;

[0048] S311: Send the warning information to the target users through the communication interface.

[0049] Preferably, the visualization display module includes a visualization unit and an interaction unit; the visualization unit is used to display the analysis results in the form of intuitive charts and images, supporting multi-perspective and multi-level data exploration; the interaction unit provides an interactive interface for researchers and decision-makers to use; augmented reality and virtual reality technologies are adopted to provide researchers with a data analysis experience; through intelligent question answering, natural language processing technology is used to respond to user queries for information retrieval and knowledge transfer.

[0050] Advantages of the present invention:

[0051] Compared with the existing atmospheric gravity wave detection and identification systems, it is often difficult to accurately identify the weak signals of atmospheric gravity waves, is easily affected by noise and other interference factors, and when dealing with large-scale data, there are often problems of low processing efficiency and long time consumption; this system uses artificial intelligence technologies such as deep learning and machine learning, can more accurately identify the characteristics of atmospheric gravity waves from complex data, reduce misjudgment and missed judgment, and through real-time online data processing, can process and analyze data immediately after it is generated, providing timely support for forecasting and warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 Shown is a schematic diagram of the working process of the high-precision detection and identification system for atmospheric gravity waves in the South China Sea region of the present invention;

[0053] Figure 2 Shown is a schematic diagram of the working steps of the identification and analysis module of the high-precision detection and identification system for atmospheric gravity waves in the South China Sea region of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0054] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0055] Please refer to Figure 1-2 , the present invention provides an embodiment: a high-precision detection and identification system for atmospheric gravity waves in the South China Sea region, including:

[0056] A data acquisition module, which is responsible for receiving the raw data from various detection devices, and performing data cleaning, data compression, and time synchronization processing on the raw data;

[0057] An identification and analysis module, which is used to extract features, identify, classify, and estimate parameters of atmospheric gravity waves from the preprocessed data by applying deep learning models and machine learning algorithms;

[0058] A real-time feedback module, which is used to realize real-time processing of data and instant feedback of results, including real-time monitoring of atmospheric gravity waves, generation and release of early warning information;

[0059] A visualization display module, which is used to display the analysis results in the form of intuitive charts and images, support multi-perspective and multi-level data exploration, and provide an interactive interface for researchers and decision-makers to use.

[0060] Preferably, the data acquisition module is responsible for receiving the raw data from various detection devices, and performing data cleaning, data compression, and time synchronization processing on the raw data; the identification and analysis module applies deep learning models and machine learning algorithms to extract features, identify, classify, and estimate parameters of atmospheric gravity waves from the preprocessed data; the real-time feedback module realizes real-time processing of data and instant feedback of results, including real-time monitoring of atmospheric gravity waves, generation and release of early warning information; the visualization display module displays the analysis results in the form of intuitive charts and images, support multi-perspective and multi-level data exploration, and provide an interactive interface for researchers and decision-makers to use.

[0061] Preferably, the data acquisition module includes a detection device group and a preprocessing unit; the detection device group includes lidar, satellite remote sensing equipment, and environmental sensors, and the lidar, satellite remote sensing equipment, and environmental sensors are used to collect various environmental data, including the raw data of atmospheric temperature, humidity, wind speed, and wind direction meteorological information; the preprocessing unit uses data cleaning to remove the noise and outliers of the raw data, uses data compression technology to reduce the burden of data storage and transmission, uses time synchronization technology to ensure the time consistency of multi-source data, and the preprocessing unit integrates an adaptive filtering algorithm to dynamically adjust the filtering parameters according to the data characteristics to optimize the preprocessing effect.

[0062] Preferably, when the data acquisition module is working, it includes the following steps:

[0063] S101: The data acquisition module receives the original data streams from lidars, satellite remote sensing devices, and environmental sensors;

[0064] S102: Conduct a preliminary verification on the received data to check the integrity of the data, the correctness of the format, and the validity of the timestamp;

[0065] S103: Set the initial parameters of the filter, including the filter order and the cut-off frequency; determine the parameter adjustment strategy, including the gradient descent method and the least squares method;

[0066] S104: Input the original data to be processed into the filter;

[0067] S105: Apply the filter to denoise the data, calculate the new filter parameters using the parameter adjustment strategy, and update the filter state. The parameters of the filter will be dynamically adjusted according to the characteristics of the input data;

[0068] S106: Output the filtered data, that is, the data with noise and outliers removed;

[0069] S107: Select the LZ4 compression algorithm to compress the cleaned data to reduce the requirements for storage space and transmission bandwidth;

[0070] S108: Use the timestamp calibration technology to align the data from different devices to the same time reference and perform time synchronization processing on the data from different detection devices;

[0071] S109: Store the data using a distributed database.

[0072] Preferably, the identification and analysis module uses a convolutional neural network model and a clustering analysis algorithm to automatically extract features, identify, classify, and estimate parameters of atmospheric gravity waves from the preprocessed data. Through transfer learning technology, a pre-trained model containing known atmospheric gravity wave events is used to accelerate the convergence speed of the atmospheric gravity wave identification model.

[0073] Preferably, when the identification and analysis module is working, it includes the following steps:

[0074] S201: Receive the multi-source meteorological data that has been cleaned, compressed, and time-synchronized from the data acquisition module;

[0075] S202: Convert the meteorological data into a format suitable for processing by the CNN model;

[0076] S203: Construct a CNN architecture according to the task requirements, including an input layer, multiple convolutional layers, a fully connected layer, and an output layer;

[0077] S204: Perform a sliding window operation on the input data through multiple convolutional kernels to extract local features; after each convolutional layer, a non-linear activation function ReLU and a pooling layer are followed;

[0078] S205: After passing through multiple convolutional and pooling layers, generate a feature map containing high-level abstract features;

[0079] S206: Select and further extract features from the original data or the feature map extracted by the CNN to construct a feature set for clustering;

[0080] S207: Perform standardization or normalization processing on the selected features;

[0081] S208: Use known atmospheric gravity wave events to perform supervised training on the CNN model, adjust the model parameters through the backpropagation algorithm, and minimize the prediction error;

[0082] S209: In an unsupervised or weakly supervised situation, apply the clustering algorithm K-means to perform clustering analysis on the feature set to identify different atmospheric gravity wave patterns or categories;

[0083] S210: Optimize the performance of the CNN model and the clustering algorithm through cross-validation and hyperparameter tuning;

[0084] S211: Use the trained CNN model to perform automatic feature extraction and classification on the input data to identify atmospheric gravity wave events;

[0085] S212: For the identified atmospheric gravity wave events, combine the results of the clustering analysis to estimate their wave speed, wavelength, and amplitude parameters;

[0086] S213: Output the identification and analysis results.

[0087] Preferably, the real-time feedback module uses the stream processing technology Apache Kafka to ensure the real-time nature of data processing, adopts a parallel computing framework to improve the processing speed, and constructs a distributed processing architecture; through an intelligent early warning algorithm, combined with historical data and current trends, predict the influence range and intensity of atmospheric gravity waves, and achieve real-time data processing and instant result feedback, including real-time monitoring of atmospheric gravity waves, generation and release of early warning information.

[0088] Preferably, when the real-time feedback module is working, it includes the following steps:

[0089] S301: Use Kafka as the message queue, and the data receives real-time data streams from various detection devices through the Apache Kafka stream processing platform;

[0090] S302: Load the pre-trained deep learning model, use GPU acceleration to perform forward propagation on the input data to obtain the recognition results of atmospheric gravity waves and parameter estimation values; use a parallel processing algorithm to parallel process data using multi-GPU or multi-CPU resources;

[0091] S303: Integrate the output data on atmospheric gravity waves with the relevant data in the historical database;

[0092] S304: Write Python scripts or use data preprocessing libraries to extract and construct features; among them, use the groupby and transform methods of Pandas to calculate seasonal change features, and use the array operations of NumPy to calculate similar events, seasonal changes, and geographical location features in historical data;

[0093] S305: Extract similar event features, seasonal change features, and geographical location features in historical data that are useful for predicting the future trend of atmospheric gravity waves from the integrated data;

[0094] S306: Select the GBDT implementation library GradientBoostingRegressor, set its initial parameters, and use the training set data to call the fit method of the model for training;

[0095] S307: Use the test set data to call the predict method of the model for prediction and calculate the prediction error; adjust the model parameters according to the evaluation results, increase the number of trees, and adjust the learning rate;

[0096] S308: Use the GBDT model to analyze the influence of features on the prediction results;

[0097] S309: Set the warning thresholds for atmospheric gravity waves based on the predicted values of amplitude, frequency, and propagation speed parameters. When the prediction results exceed these thresholds, the system will trigger the warning mechanism;

[0098] S310: When the prediction results of the GBDT model trigger the warning conditions, the system uses the predicted results of the GBDT model and the warning thresholds as inputs according to the preset format and template to generate warning information;

[0099] S311: Use a communication interface to send the warning information to the target users.

[0100] Preferably, the visualization display module includes a visualization unit and an interaction unit; the visualization unit is used to display the analysis results in the form of intuitive charts and images, supporting multi-perspective and multi-level data exploration; the interaction unit provides an interactive interface for researchers and decision-makers to use; augmented reality and virtual reality technologies are adopted to provide researchers with a data analysis experience; through intelligent Q&A, natural language processing technology is used to respond to user queries for information retrieval and knowledge transfer.

[0101] Preferably, when the visualization display module is working, it includes the following steps:

[0102] S401: Receive the data processed by the intelligent recognition and analysis module, including the recognition results and parameter estimation of atmospheric gravity waves;

[0103] S402: Check whether the received data is complete and whether the format is correct, and convert the received data into a format suitable for visualization display, such as JSON, CSV;

[0104] S404: Design a visualization scheme according to the data characteristics, and select a suitable chart type and layout;

[0105] S405: Analyze the statistical characteristics of the data, such as the maximum value, minimum value, average value, etc., to determine the details of the visualization scheme;

[0106] S406: According to the data type and display requirements, select a suitable chart type such as line chart, scatter plot, heat map;

[0107] S407: Use the data visualization library D3.js to encode and implement the designed visualization scheme;

[0108] S408: Initialize the visualization canvas according to the browser window size or the specified container size;

[0109] S409: Bind the processed data to chart elements, such as coordinate axes, data points;

[0110] S410: Add interactive elements such as buttons, sliders, selection boxes, and write event handling functions to respond to user operations.

[0111] Preferably, when the visualization display module is working, it further includes the following steps:

[0112] S411: Use Web front-end technologies HTML5, CSS3 and JavaScript to develop a user interactive interface, and integrate the visualization chart into the interface;

[0113] S412: Build the overall layout of the page, including the navigation bar, sidebar, and content area;

[0114] S413: Integrate the visualization chart as a component into the page content area;

[0115] S414: Write the jump logic between pages and the interaction logic between components to ensure smooth user operations;

[0116] S415: Use the performance testing tool LoadRunner to conduct stress testing on the interaction unit;

[0117] S416: Analyze the performance bottlenecks based on the test results, such as network latency and data processing speed;

[0118] S417: Collect user feedback and iteratively optimize the visualization display module based on the user feedback.

[0119] Embodiment

[0120] Optionally, when performing data recognition and analysis, it is preset to receive a new set of meteorological data in the South China Sea region. These data have been preprocessed and converted into a format suitable for input to the CNN model; the input data is as follows:

[0121] Time range: One week;

[0122] Spatial range: A specific sea area in the South China Sea;

[0123] Data dimension: Four-dimensional tensor (time, height, longitude, latitude), with temperature, humidity, wind speed (u-component, v-component), and pressure parameters included in each dimension;

[0124] Send the data into the trained CNN model, and the model automatically performs multi-layer convolution and pooling operations to extract features related to atmospheric gravity waves. The features may include but are not limited to: areas with significant temperature gradient changes, stratifications with sudden increases in wind speed, and pressure fluctuation patterns;

[0125] The fully connected layer and the Softmax layer convert the extracted features into a probability distribution, outputting the probability that each data point belongs to different atmospheric gravity wave categories. Set a threshold, and when the probability of a certain category exceeds the threshold, it is considered that the data point belongs to that category of atmospheric gravity wave event;

[0126] Select from the features extracted by the CNN model the periodic change features of the temperature gradient and wind speed of a specific stratification related to parameters such as wave speed, wavelength, and amplitude; according to the results of K-means clustering analysis, group the identified atmospheric gravity wave events, and each group represents atmospheric gravity waves generated by different sources or with different propagation characteristics;

[0127] Combined with wind speed data and gravity wave propagation direction information, the wave speed (V) is calculated by measuring the distance that an atmospheric gravity wave propagates in a specific direction within a fixed time period; the wavelength (λ) is estimated using the distance between adjacent wave crests or wave troughs in space; the amplitude (A) is estimated by calculating half of the difference between the maximum and minimum values of the wave parameters; for each identified atmospheric gravity wave event, its wave speed, wavelength, and amplitude parameters are estimated, and the association between these parameters and the clustering results is recorded;

[0128] The recognition and analysis results are output in a structured form, including: the type of atmospheric gravity wave (such as terrain wave, convective wave, etc., based on the clustering results), the occurrence location (latitude, longitude, altitude), the occurrence time (specific time point or time period), the estimated wave speed, wavelength, and amplitude parameters.

[0129] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments, and various changes can be made without departing from the spirit of the present invention within the knowledge scope of those skilled in the art.

Claims

1. A high-precision detection and identification system for atmospheric gravity waves in the South China Sea region, characterized by: Included are: The data acquisition module is responsible for receiving the raw data from various detection devices and performing data cleaning, data compression and time synchronization processing on the raw data; The recognition and analysis module is used to apply deep learning models and machine learning algorithms to perform feature extraction, atmospheric gravity wave recognition, classification, and parameter estimation on preprocessed data; The recognition and analysis module uses a convolutional neural network model and cluster analysis algorithm to automatically extract features, identify atmospheric gravity waves, classify and estimate parameters of pre-processed data. Through transfer learning technology, a pre-trained model containing known atmospheric gravity wave events is used to accelerate the convergence speed of the atmospheric gravity wave recognition model. Real-time feedback module, used to realize real-time data processing and instant feedback of results, including real-time monitoring of atmospheric gravity waves, generation and release of early warning information; The real-time feedback module uses stream processing technology Apache Kafka to ensure real-time data processing, uses a parallel computing framework to increase processing speed, and builds a distributed processing architecture; Through intelligent early warning algorithms, historical data and current trends are combined to predict the impact range and intensity of atmospheric gravity waves, and to achieve real-time data processing and instant feedback of results, including real-time monitoring of atmospheric gravity waves, generation and release of early warning information; The visualization display module is used to display the analysis results in the form of intuitive charts and images, supports multi-perspective and multi-level data exploration, and provides an interactive interface for researchers and decision makers.

2. The high-precision detection and identification system for atmospheric gravity waves in the South China Sea region according to claim 1 is characterized in that: The data acquisition module includes a detection equipment group and a preprocessing unit; the detection equipment group includes lidar, satellite remote sensing equipment and environmental sensors, which are used to collect various environmental data, including raw data of atmospheric temperature, humidity, wind speed and wind direction meteorological information; the preprocessing unit uses data cleaning to remove noise and outliers from the raw data, uses data compression technology to reduce the burden of data storage and transmission, and uses time synchronization technology to ensure the time consistency of multi-source data. The preprocessing unit integrates an adaptive filtering algorithm, dynamically adjusts the filtering parameters according to the data characteristics, and optimizes the preprocessing effect.

3. The high-precision detection and identification system for atmospheric gravity waves in the South China Sea region according to claim 2 is characterized in that: When the data acquisition module is working, it includes the following steps: S101: The data acquisition module receives raw data streams from laser radar, satellite remote sensing equipment and environmental sensors; S102: Perform preliminary verification on the received data to check the integrity, format correctness and validity of the timestamp of the data; S103: setting initial parameters of the filter, including filter order and cutoff frequency; determining parameter adjustment strategy, including gradient descent method and least square method; S104: inputting the raw data to be processed into the filter; S105: Apply the filter to denoise the data, use the parameter adjustment strategy to calculate new filter parameters, and update the filter state. The filter parameters will be dynamically adjusted according to the characteristics of the input data; S106: Output the filtered data; S107: Select LZ4 compression algorithm to compress the cleaned data; S108: using a timestamp calibration technology to align data from different devices to the same time reference, and performing time synchronization processing on data from different detection devices; S109: Use a distributed database to store data.

4. The high-precision detection and identification system for atmospheric gravity waves in the South China Sea region according to claim 3 is characterized by: When the recognition and analysis module is working, the following steps are included: S201: receiving cleaned, compressed and time-synchronized multivariate meteorological data from a data acquisition module; S202: Convert meteorological data into a format suitable for CNN model processing; S203: Construct a CNN architecture according to task requirements, including an input layer, multiple convolutional layers, a fully connected layer, and an output layer; S204: Perform sliding window operation on the input data through multiple convolution kernels to extract local features; each convolution layer is followed by a nonlinear activation function ReLU and a pooling layer; S205: After multiple convolution and pooling layers, a feature map containing high-level abstract features is generated; S206: Select and further extract features from the original data or the feature map extracted by CNN to construct a feature set for clustering; S207: Standardize or normalize the selected features; S208: Use known atmospheric gravity wave events to conduct supervised training of the CNN model, and adjust the model parameters through the back-propagation algorithm to minimize the prediction error; S209: Under unsupervised or weakly supervised conditions, apply the clustering algorithm K-means to perform cluster analysis on the feature set to identify different atmospheric gravity wave patterns or categories; S210: Optimizing the performance of CNN models and clustering algorithms through cross-validation and hyperparameter tuning; S211: Automatically extract and classify input data using the trained CNN model to identify atmospheric gravity wave events; S212: For the identified atmospheric gravity wave events, estimate their wave velocity, wavelength, and amplitude parameters based on the results of cluster analysis; S213: Output recognition and analysis results.

5. The high-precision detection and identification system for atmospheric gravity waves in the South China Sea region according to claim 4 is characterized in that: When the real-time feedback module is working, it includes the following steps: S301: Using Kafka as a message queue, data is received through the Apache Kafka stream processing platform in real time from various detection devices. S302: Load the pre-trained deep learning model, use GPU acceleration to forward propagate the input data, and obtain the recognition results and parameter estimation values ​​of atmospheric gravity waves; use parallel processing algorithms to process data in parallel using multiple GPUs or multiple CPU resources; S303: Integrate the output data on atmospheric gravity waves with relevant data in the historical database; S304: Write Python scripts or use data preprocessing libraries to extract and construct features; use Pandas' groupby and transform methods to calculate seasonal change features, and use NumPy's array operations to calculate similar events, seasonal changes, and geographic location features in historical data; S305: extracting similar event features, seasonal variation features, and geographic location features in historical data useful for predicting future trends of atmospheric gravity waves from the integrated data; S306: Select the GBDT implementation library GradientBoostingRegressor, set its initial parameters, and use the training set data to call the model's fit method for training; S307: Use the test set data to call the predict method of the model for prediction and calculate the prediction error; adjust the model parameters according to the evaluation results, increase the number of trees, and adjust the learning rate; S308: Using the GBDT model to analyze the influence of features on the prediction results; S309: Setting warning thresholds for atmospheric gravity waves based on predicted values ​​of amplitude, frequency, and propagation speed parameters. When the predicted results exceed these thresholds, the system will trigger a warning mechanism. S310: When the prediction result of the GBDT model triggers the warning condition, the system generates warning information using the prediction result of the GBDT model and the warning threshold as input according to the preset format and template; S311: Use the communication interface to send the warning information to the target user.

6. The high-precision detection and identification system for atmospheric gravity waves in the South China Sea region according to claim 5 is characterized by: The visualization display module includes a visualization unit and an interaction unit; the visualization unit is used to display the analysis results in the form of intuitive charts and images, supporting multi-perspective and multi-level data exploration; the interaction unit provides an interactive interface for researchers and decision makers; it uses augmented reality and virtual reality technologies to provide researchers with data analysis experience; it responds to user queries through intelligent question and answer through natural language processing technology, and conducts information retrieval and knowledge transfer.

7. The high-precision detection and identification system for atmospheric gravity waves in the South China Sea region according to claim 6 is characterized by: When the visualization module is working, it includes the following steps: S401: receiving data processed by the recognition and analysis module, including recognition results and parameter estimation of atmospheric gravity waves; S402: Check whether the received data is complete and in a correct format, and convert the received data into a JSON or CSV format suitable for visual display; S404: Design a visualization scheme based on data characteristics and select appropriate chart types and layouts; S405: Analyze the statistical characteristics of the data, including maximum value, minimum value and average value; S406: Select a chart type, such as line chart, scatter chart, or heat map, according to the data type and display requirements; S407: Use the data visualization library D3.js to encode the designed visualization solution; S408: Initialize the visualization canvas according to the browser window size or the specified container size; S409: Binding the processed data to the coordinate axes and data points on the chart elements; S410: Add interactive elements, including any one of buttons, sliders, and selection boxes, and write event processing functions to respond to user operations.

8. The high-precision detection and identification system for atmospheric gravity waves in the South China Sea region according to claim 7 is characterized in that: When the visualization module is working, the following steps are also included: S411: Use web front-end technologies HTML5, CSS3 and JavaScript to develop user interactive interfaces and integrate visual charts into the interfaces; S412: Build the overall layout of the page, including the navigation bar, sidebar, and content area; S413: Integrate the visualization chart as a component into the page content area; S414: Write the jump logic between pages and the interaction logic between components to ensure smooth user operation; S415: Use the performance testing tool LoadRunner to perform stress testing on the interaction unit; S416: Analyze network delay and data processing speed according to the test results; S417: Collect user feedback, and iteratively optimize the visualization display module based on the user feedback.

Images