Intelligent meteorology-oriented airborne detection data analysis method and system

Abstract

The invention relates to the technical field of meteorological data analysis, in particular to an airborne detection data analysis method and system oriented to intelligent meteorology, airborne detection data is disassembled by using data scanning of a first dimension, and an aviation meteorological embedded state vector is generated by analyzing meteorological characteristic correlation between data units; the intrinsic law and the change trend of the meteorological data can be more comprehensively captured, and the accuracy and the reliability of prediction are improved. By adjusting the step length of a data scanning kernel and introducing data disassembly of different dimensions, multi-level and multi-angle analysis of meteorological data is realized, complexity and dynamics of the meteorological data can be more comprehensively understood, powerful support is provided for subsequent prediction and decision support, the data transmission efficiency is optimized, and the real-time performance of the meteorological data is improved. And the real-time performance and the accuracy of the information are ensured. The method is crucial to real-time monitoring and early warning in the flight process, accurate meteorological information can be provided for pilots in time, and the pilots are helped to make scientific flight decisions.

Description

Technical Field

[0001] The present application relates to the technical field of meteorological data analysis, and in particular to an airborne detection data analysis method and system for intelligent meteorology. Background Art

[0002] In existing aviation meteorological services, the prediction and monitoring of weather patterns are important links to ensure flight safety. However, traditional meteorological forecasting methods often rely on limited meteorological data and simple analysis models, which makes it difficult to accurately capture the inherent laws and complex changes of meteorological data. Especially during flight, real-time and accuracy are particularly important for obtaining meteorological information. Therefore, how to accurately identify and predict complex weather patterns has become an urgent problem to be solved in the field of aviation meteorological services. Summary of the invention

[0003] In order to improve the above problems, the present application provides an airborne detection data analysis method and system for intelligent meteorology.

[0004] The embodiment of the present application provides an airborne detection data analysis method for intelligent meteorology, which is applied to a big data system. The method includes: acquiring airborne detection data, the airborne detection data is meteorological detection data of a weather pattern to be predicted, the weather pattern is used to characterize the meteorological state change of multidimensional meteorological data in the airborne detection data, the airborne detection data is disassembled into a plurality of first data scan sets by a data scan core of a first dimension, the airborne detection data is disassembled into a plurality of data units by a second dimension, the first dimension is larger than the second dimension; according to the meteorological feature correlation between the data units in the plurality of first data scan sets, the first airborne detection data corresponding to the plurality of data units are determined; The invention relates to an air meteorological embedded state vector; adjusting the data scanning kernel of the first dimension in the airborne detection data by a set step size, and re-decomposing the airborne detection data into a plurality of second data scanning sets; determining the second aviation meteorological embedded state vectors corresponding to the plurality of data units respectively according to the meteorological characteristic correlation between the first aviation meteorological embedded state vectors in the plurality of second data scanning sets; predicting the weather pattern in the airborne detection data according to the second aviation meteorological embedded state vector, and obtaining a weather pattern prediction label corresponding to the airborne detection data, wherein the weather pattern prediction label is used to characterize the prediction result of the airborne detection data relative to the flight safety weather state.

[0005] An embodiment of the present application provides a big data system, comprising at least one processor and a memory; the memory stores computer-executable instructions; the at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the above method.

[0006] An embodiment of the present application provides a computer-readable storage medium on which a computer program is stored. The computer program implements the above method when running.

[0007] The embodiment of the present application conducts in-depth mining and analysis of a large amount of meteorological data collected by airborne detection equipment, and utilizes the inherent laws and correlations of multi-dimensional meteorological data to achieve accurate identification and prediction of complex weather patterns. Compared with traditional meteorological forecasting methods, the embodiment of the present application has the following significant advantages: First, the airborne detection data is disassembled by scanning and checking the data in the first dimension, and the aviation meteorological embedded state vector is generated by analyzing the meteorological feature correlation between data units. This method can more comprehensively capture the inherent laws and changing trends of meteorological data and improve the accuracy and reliability of prediction.

[0008] Secondly, by adjusting the step size of the data scanning kernel and introducing data decomposition in different dimensions, a multi-level and multi-angle analysis of meteorological data is achieved. This helps to more comprehensively understand the complexity and dynamics of meteorological data and provide strong support for subsequent forecasting and decision support.

[0009] In addition, the data transmission efficiency is optimized to ensure the real-time and accuracy of information, which is crucial for real-time monitoring and early warning during flight, and can provide pilots with accurate weather information in a timely manner to help them make scientific flight decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 A flowchart of an airborne detection data analysis method for intelligent meteorology provided in an embodiment of the present application.

[0011] Figure 2 A schematic diagram of the structure of a big data system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0012] In order to better understand the above technical scheme, the technical scheme of the present application is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present application and the specific features in the embodiments are detailed descriptions of the technical scheme of the present application, rather than limitations on the technical scheme of the present application. In the absence of conflict, the embodiments of the present application and the technical features in the embodiments can be combined with each other.

[0013] Figure 1 A method for analyzing airborne detection data for intelligent meteorology is shown, which is applied to a big data system. The method includes the following steps 110 to 150.

[0014] In one application scenario, a big data system is used to process and analyze real-time meteorological data from airborne detection equipment to predict weather patterns and generate corresponding flight safety prediction labels. First, the big data system receives detection data from airborne detection equipment in real time. These airborne detection data contain rich meteorological information, such as temperature, humidity, wind speed, wind direction, air pressure, etc., which are crucial for predicting weather patterns. These data are identified by the system as meteorological detection data of the weather pattern to be predicted. Next, the big data system uses a pre-set first-dimensional data scan core to disassemble the airborne detection data. This data scan core is a specific algorithm or tool for dividing data according to a specific dimension. In this step, the airborne detection data is disassembled into several first data scan sets. Each first data scan set contains part of the meteorological data, but the dimensions of these data sets are reduced compared to the original airborne detection data. Subsequently, the big data system performs meteorological feature correlation analysis on the data units in each first data scan set. This analysis process involves comparing meteorological features between different data units, such as the correlation between temperature and humidity, the correlation between wind speed and wind direction, etc. By analyzing these correlations, the big data system determines a first aviation meteorological embedded state vector for each data unit. The first aviation meteorological embedded state vector is a mathematical representation used to describe the position and state of the data unit in the meteorological feature space. After determining the first aviation meteorological embedded state vector, the big data system adjusts the step size of the data scan kernel of the first dimension and disassembles it again in the airborne detection data. The result of this disassembly produces several second data scan sets. Compared with the first data scan set, the second data scan set may have a different data distribution and feature combination. Next, the big data system uses the meteorological feature correlation analysis again to determine the correlation between each first aviation meteorological embedded state vector in the second data scan set. This analysis process is similar to the previous step, but is based on different data scan sets and embedded state vectors. By analyzing these correlations, the big data system determines a second aviation meteorological embedded state vector for each data unit. This vector may be different from the first aviation meteorological embedded state vector because it is based on a more fine-grained or different angle of data analysis. Finally, the big data system uses the second aviation meteorological embedded state vector to predict weather patterns in the airborne detection data. This process can involve complex machine learning algorithms or models, such as neural networks, decision trees, etc. By comparing and analyzing the relationship between the embedded state vector and known weather patterns, the big data system generates a weather pattern prediction label. This label is an indicator that characterizes the prediction results of the onboard detection data relative to the flight safety weather state. For example, if the prediction label shows that the current weather pattern may pose a threat to flight safety, then the airline or pilot can take corresponding measures to ensure flight safety.Throughout the process, the big data system has achieved an in-depth understanding and utilization of airborne detection data through continuous disassembly, analysis and prediction, which not only improves the accuracy and reliability of weather forecasts, but also provides strong support for flight safety.

[0015] Step 110: Acquire airborne detection data, where the airborne detection data is meteorological detection data whose weather pattern is to be predicted, and the weather pattern is used to characterize the change in meteorological state of multidimensional meteorological data in the airborne detection data. The airborne detection data is decomposed into a plurality of first data scan sets by a data scan core of a first dimension, and the airborne detection data is decomposed into a plurality of data units by a second dimension, and the first dimension is larger than the second dimension.

[0016] In step 110, the airborne detection data refers to the real-time meteorological information collected by various meteorological detection equipment installed on the aircraft. These data include but are not limited to various meteorological parameters such as temperature, humidity, wind speed, wind direction, and air pressure. After being acquired by sensors and processed, they become an important basis for flight safety assessment and weather pattern prediction. For example, during flight, the airborne radar can detect precipitation clouds ahead in real time, and these data are part of the airborne detection data.

[0017] Weather patterns refer to the combination and change patterns of various meteorological elements in the atmosphere (such as temperature, humidity, air pressure, wind speed, etc.) within a specific time and space range. Weather patterns can characterize the state and evolution trend of weather systems and are key elements in weather forecasting and flight safety assessment. For example, a tropical cyclone (typhoon) is a typical weather pattern with specific meteorological characteristics and movement patterns.

[0018] Meteorological state changes refer to the changes of various meteorological elements in the atmosphere over time and space. These changes may be small fluctuations or significant mutations, which together constitute complex and changeable weather phenomena. During flight, changes in meteorological state directly affect flight safety and comfort. For example, sudden strong wind shear during flight may cause aircraft turbulence or even crash.

[0019] The data scanning kernel is a specific algorithm or tool used for data decomposition and analysis in big data systems. It can decompose complex airborne detection data into multiple data subsets (i.e., data scanning sets) according to preset dimensions and rules for more in-depth analysis and prediction. The design and selection of data scanning kernels are crucial to the accuracy and efficiency of data analysis. For example, in meteorological data analysis, the data scanning kernel can decompose data into different subsets such as temperature, humidity, and wind speed according to the correlation of meteorological elements.

[0020] The first data scan set is a data subset obtained by the big data system after disassembling the airborne detection data through the first dimension data scan. These subsets contain some meteorological data, but their dimensions are reduced compared to the original airborne detection data, which is more convenient for subsequent analysis and prediction. The number and specific content of the first data scan set depends on the setting of the data scan core and the characteristics of the airborne detection data.

[0021] The data unit is the smallest data unit in the airborne detection data, which contains the numerical information of a single meteorological element at a certain moment. In the big data system, the data unit is the basis for analyzing the correlation of meteorological characteristics and constructing the embedded state vector. By comparing the correlation of meteorological characteristics between different data units, the big data system can reveal the internal connection and change law between meteorological elements.

[0022] In detail, the big data system receives detection data from the airborne detection equipment in real time based on step 110. These data are meteorological detection data for weather patterns to be predicted, and contain meteorological state change information of multi-dimensional meteorological data. In order to better perform subsequent analysis and prediction, the big data system uses the data scan of the first dimension to disassemble the airborne detection data and divide it into several first data scan sets. At the same time, the airborne detection data is also disassembled into several data units in the second dimension, where the first dimension is larger than the second dimension, so as to perform data analysis and processing at different levels.

[0023] Step 120: Determine first aviation meteorological embedded state vectors corresponding to the plurality of data units, respectively, based on meteorological characteristic correlations between the plurality of data units in the plurality of first data scan sets.

[0024] In step 120, meteorological feature correlation refers to the degree of association between different meteorological features (such as temperature, humidity, wind speed, etc.) in airborne detection data. This correlation can be measured by statistical analysis methods (such as correlation coefficient, covariance, etc.). For example, in airborne detection data, there may be a high correlation between temperature and humidity, because in some cases, when the temperature increases, the humidity will also increase accordingly. This correlation is crucial to understanding the inherent laws of meteorological data and predicting weather patterns. For example, airborne detection data contains data on temperature and humidity, and the correlation between them can be measured by calculating the correlation coefficient between the two. If the correlation coefficient is close to 1 (or -1), it means that there is a high positive correlation (or negative correlation) between the two; if the correlation coefficient is close to 0, it means that the correlation between the two is weak.

[0025] The first aviation meteorological embedded state vector is a mathematical representation used to describe the position and state of each data unit in the airborne detection data in the meteorological feature space. This vector can be composed of multiple values, each of which represents the performance of the data unit on a specific meteorological feature. These values ​​can be extracted from the original meteorological data through various algorithms (such as principal component analysis, neural networks, etc.). For example, each data unit in the airborne detection data is composed of three meteorological features: temperature, humidity, and wind speed. Through a certain algorithm (such as a deep learning model), a three-dimensional embedded state vector can be generated for each data unit, such as [0.5, 0.8, 0.3]. The three values ​​in this vector represent the performance of the data unit on the three features of temperature, humidity, and wind speed.

[0026] Based on step 120, the big data system first analyzes the meteorological feature correlation between the data units in each first data scan set. This includes calculating the correlation coefficient, covariance and other statistics between different meteorological features to reveal the degree of association between them. Then, the big data system uses this correlation information and combines it with a specific algorithm (such as a deep learning model, principal component analysis, etc.) to generate a first aviation meteorological embedded state vector for each data unit. This vector is a multi-dimensional numerical representation that integrates the performance of the data unit on various meteorological features and can accurately locate the position and state of the data unit in the meteorological feature space. Specifically, the big data system can first pre-process the original meteorological data, such as data cleaning and normalization, to ensure the quality and consistency of the data. Then, it will select a suitable algorithm (such as a deep learning model) and set the corresponding parameters and training strategies. Through the training model, the big data system can learn the meteorological feature correlation between the data units and generate the corresponding embedded state vector. This embedded state vector not only contains the numerical information of the data unit on each meteorological feature, but also contains the relative relationship and spatial structure between the data units. Therefore, it can provide strong support for subsequent weather pattern prediction. In this way, the big data system can generate an accurate and comprehensive first aviation meteorological embedded state vector for each data unit, laying a solid foundation for subsequent weather pattern predictions.

[0027] Step 130: adjust the data scan core of the first dimension by setting a step length in the airborne detection data, and re-decompose the airborne detection data into a plurality of second data scan sets.

[0028] In the application scenarios of meteorological data analysis and weather pattern prediction, the second data scan set refers to the data set obtained by re-disassembling the airborne detection data after the data scan core of the first dimension is adjusted by the set step size. The "first dimension" here can refer to the initial dimension or granularity of the data scan core applied to the original data, and "setting the step size adjustment" refers to adjusting the movement or change step size of the data scan core on the data according to certain rules or algorithms, so as to observe and analyze the data at a finer granularity or from different angles.

[0029] In the second data scan set, the data is disassembled in a similar way to the first data scan set, but because the step size of the data scan kernel has changed, each second data scan set can contain different data distributions and feature combinations. This change helps the big data system extract and analyze the inherent laws and correlations of meteorological data from multiple angles and levels, thereby more accurately predicting weather patterns.

[0030] For example, airborne detection data contains meteorological data such as temperature, humidity, wind speed, etc. of a region, and these data are arranged in chronological order. In the first round of data scanning, the data scanning core decomposes the data with a larger step size (such as every hour) to generate several first data scanning sets, each of which contains meteorological data at adjacent time points. Then, in the second round of data scanning, the step size of the data scanning core is adjusted to a smaller value (such as every 15 minutes), and the airborne detection data is re-decomposed to generate several second data scanning sets. These second data scanning sets contain finer-grained meteorological data due to the reduction in step size, allowing the big data system to observe and analyze the changes and correlations of meteorological characteristics more finely.

[0031] It can be understood that in the process of meteorological data analysis and weather pattern prediction, step 130 is a key step, which involves adjusting the step size of the data scan kernel and re-disassembling the airborne detection data accordingly. In this step, the big data system will first determine the specific value and direction of adjusting the step size of the data scan kernel according to certain rules or algorithms. The adjustment of this step size is based on the results of the first round of data scanning (generating the first data scan set) and the in-depth understanding of the inherent laws and correlations of meteorological data. By adjusting the step size, the big data system can observe and analyze meteorological data at a finer granularity or different angles, thereby extracting more useful information. Then, the big data system uses the adjusted data scan to re-disassemble the airborne detection data. This process is similar to the first round of data scanning, but due to the change in step size, each second data scan set will contain different data distributions and feature combinations. These second data scan sets will serve as the basis for subsequent meteorological feature correlation analysis and weather pattern prediction. In the process of re-disassembly, the big data system needs to ensure the integrity and accuracy of the data, and also pay attention to handling possible outliers and noise data. In addition, since the adjustment of the step size can lead to an increase or decrease in the amount of data, the big data system also needs to adjust and optimize the data processing and storage capabilities accordingly. In this way, the big data system can obtain a more detailed and comprehensive meteorological data set (i.e., the second data scan set), providing strong support for subsequent weather pattern predictions.

[0032] Step 140: Determine the second aviation meteorological embedded state vectors corresponding to the plurality of data units respectively according to the meteorological characteristic correlation between the first aviation meteorological embedded state vectors in the plurality of second data scan sets.

[0033] In the application scenarios of meteorological data analysis and flight safety prediction, the second aviation meteorological embedded state vector is a new meteorological characteristic state representation generated by the big data system for each data unit based on data analysis at a finer granularity or from different angles. It is different from the first aviation meteorological embedded state vector because the second aviation meteorological embedded state vector is determined based on the meteorological characteristic correlation between the first aviation meteorological embedded state vectors in the second data scan set after adjusting the step size of the data scan kernel.

[0034] Specifically, after the big data system re-disassembles the airborne detection data into several second data scan sets by adjusting the set step size, it will further analyze the meteorological feature correlations between the original data units (or data points) in these newly generated data sets. These correlations can include the interactions and associations between various meteorological features such as temperature and humidity, wind speed and wind direction. Based on these new correlation analyses, the big data system recalculates and generates a second aviation meteorological embedded state vector for each data unit. This new embedded state vector not only contains the meteorological feature information of the data unit at a finer granularity or at different angles, but also reflects the new position and state of the data unit in the meteorological feature space. It is a more in-depth description and representation of the meteorological feature state of the data unit, which helps the big data system to more accurately understand and predict weather patterns, thereby providing more reliable prediction results for flight safety.

[0035] It can be understood that in the process of meteorological data analysis and flight safety prediction, step 140 is an important step, which involves determining the second aviation meteorological embedded state vector for each data unit based on data analysis at a finer granularity or from different angles. In this step, the big data system first acquires and analyzes the first aviation meteorological embedded state vectors in the second data scan set. These embedded state vectors are generated on the basis of the first round of data scanning and meteorological feature correlation analysis, and they have made a preliminary description of the position and state of the data unit in the meteorological feature space. Then, the big data system further analyzes the meteorological feature correlation between these first aviation meteorological embedded state vectors. This analysis process may include calculating the similarity, distance or other correlation indicators between vectors to reveal the mutual relationship and change trend of different data units in meteorological features. Based on these new correlation analyses, the big data system recalculates and generates a second aviation meteorological embedded state vector for each data unit. This new embedded state vector will contain meteorological feature information at a finer granularity or from different angles, and more accurately reflect the position and state of the data unit in the meteorological feature space. In the process of calculating the second aviation meteorological embedded state vector, the big data system can adopt various advanced algorithms and technologies, such as neural networks, deep learning, etc. These algorithms and technologies can effectively process and analyze large amounts of meteorological data, and extract useful information and features to provide strong support for subsequent weather pattern forecasts. Through this step, the big data system can generate a more accurate and comprehensive second aviation meteorological embedded state vector for each data unit, providing a more reliable basis for subsequent flight safety forecasts. This will help improve the accuracy and reliability of flight safety forecasts and provide more valuable decision support for airlines and pilots.

[0036] Step 150: Predict the weather pattern in the airborne detection data based on the second aviation meteorological embedded state vector to obtain a weather pattern prediction label corresponding to the airborne detection data, wherein the weather pattern prediction label is used to characterize the prediction result of the airborne detection data relative to the flight safety weather state.

[0037] In step 150, the weather pattern prediction label refers to the identification or label of the possibility of future weather patterns obtained after analyzing the airborne detection data through big data and machine learning algorithms. These labels can be based on complex algorithms and models, through comprehensive analysis of multiple features such as temperature, humidity, wind speed, wind direction in meteorological data, to predict future weather patterns, such as sunny, cloudy, rainy, stormy, etc. These labels not only contain basic information about the weather pattern, but also reflect the possible impact of the weather pattern on flight safety.

[0038] The prediction result of airborne detection data relative to flight safety weather conditions refers to the evaluation result of the possible impact of current or future weather conditions on flight safety obtained based on airborne detection data through a series of analysis and prediction algorithms. This prediction result can be based on weather pattern prediction tags, combined with multiple factors such as aircraft performance parameters, flight routes, and flight altitudes, to comprehensively evaluate the impact of weather on flight safety. For example, if a strong storm is predicted in the future, then this prediction result can prompt the pilot to change the flight route or lower the flight altitude to ensure flight safety.

[0039] In the process of meteorological data analysis and flight safety prediction, step 150 is a key step, which involves using the second aviation meteorological embedded state vector to predict the weather pattern in the airborne detection data and generate a weather pattern prediction label based on it. First, the big data system uses the calculated second aviation meteorological embedded state vectors, which contain meteorological feature information at a finer granularity or different angles, and can more accurately reflect the position and state of the data unit in the meteorological feature space. Then, the big data system inputs these second aviation meteorological embedded state vectors into the pre-trained weather pattern prediction model. This model may be built based on deep learning, neural networks or other machine learning algorithms, and can automatically identify and predict different weather patterns through analysis and learning of meteorological data. During the model operation, it will calculate the most likely weather pattern prediction result based on the input second aviation meteorological embedded state vector, combined with the parameters and algorithms inside the model. This result can be expressed in the form of a label, such as "sunny", "cloudy", "rainy", etc. Finally, this weather pattern prediction label will be used to characterize the prediction result of the airborne detection data relative to the flight safety weather state. In other words, this label not only contains basic information about the weather pattern, but also reflects the possible impact of the weather pattern on flight safety. Pilots and airlines can evaluate the current flight conditions based on this label and make corresponding decisions and adjustments to ensure flight safety. Through this step, the big data system can provide pilots and airlines with timely and accurate weather pattern forecast information, helping them better understand the current flight conditions and take corresponding measures to ensure flight safety.

[0040] The embodiment of the present application uses a series of advanced data processing technologies and intelligent algorithms to conduct in-depth mining and analysis of a large amount of meteorological data collected by airborne detection equipment, thereby achieving accurate identification and prediction of complex weather patterns. Specifically, the embodiment of the present application uses a first-dimensional data scan to disassemble the airborne detection data, and generates a first aviation meteorological embedded state vector by analyzing the meteorological feature correlation between data units. Subsequently, the airborne detection data is re-disassembled by adjusting the step size of the data scan kernel, and a more accurate second aviation meteorological embedded state vector is generated based on the meteorological feature correlation in the second data scan set. Finally, based on these embedded state vectors, the embodiment of the present application can predict the weather pattern corresponding to the airborne detection data, and output the prediction results in the form of weather pattern prediction labels.

[0041] First, through the deep mining of multi-dimensional meteorological data and the application of intelligent algorithms, the accuracy and reliability of weather pattern predictions have been significantly improved. This is of great significance to aviation meteorological support and can effectively improve the level of flight safety. Secondly, by adjusting the step size of the data scanning kernel and introducing data decomposition in different dimensions, multi-level and multi-angle analysis of meteorological data is achieved, thereby more comprehensively capturing the inherent laws and changing trends of meteorological data. This not only improves the efficiency and accuracy of data analysis, but also provides strong support for subsequent meteorological forecasts and decision support. In addition, the efficiency of data transmission is optimized to ensure the real-time and accuracy of information. This is crucial for real-time monitoring and early warning during flight, and can provide pilots with accurate meteorological information in a timely manner to help them make scientific flight decisions.

[0042] In summary, through the application of advanced data processing technology and intelligent algorithms, we have achieved accurate identification and prediction of complex weather patterns, significantly improving the reliability and safety of aviation meteorological services. This is of great significance for improving flight safety and optimizing meteorological services.

[0043] It should be noted that the adjustment of the step size in the airborne detection data for the data scanning core of the first dimension is one of the core inventions of the embodiment of the present application. The following is a detailed introduction to it.

[0044] 1. The Importance of Setting Step Size Adjustment In aviation meteorological monitoring and forecasting, the step size setting of the data scanning kernel has a decisive influence on the accurate extraction and analysis of meteorological data. The size of the step size determines the accuracy and efficiency of data scanning, which in turn affects the accuracy of weather pattern prediction. Therefore, adjusting the step size of the data scanning kernel in the first dimension is a key step in optimizing data processing and forecasting effects.

[0045] 2. Specific operation of setting step length adjustment Initial step size setting: Before starting data scanning, first set an initial data scanning kernel step size based on the characteristics of airborne detection data and prediction requirements. This step size can be determined based on historical data and empirical values ​​to ensure that the main characteristics and changes of meteorological data can be covered.

[0046] Dynamic adjustment mechanism: During the data scanning process, the step size of the data scanning kernel is dynamically adjusted according to the scanning results and real-time weather conditions. This dynamic adjustment can be automatic adjustment based on machine learning algorithms or manual adjustment under human intervention.

[0047] Automatic adjustment based on machine learning algorithms: By training a machine learning model, the model can automatically adjust the step size according to the quality and complexity of real-time data. When the data changes are complex and the weather patterns are changeable, the model will automatically reduce the step size to improve scanning accuracy; when the data changes are stable and the weather patterns are stable, the model will appropriately increase the step size to improve scanning efficiency.

[0048] Manual adjustment under human intervention: In some cases, such as encountering special weather events or flight mission requirements, the operator can manually adjust the step size according to the actual situation. This adjustment requires the operator to have rich meteorological knowledge and experience to ensure that the adjusted step size can more accurately reflect the actual situation of meteorological data.

[0049] Multi-step combined scanning: In order to extract more comprehensive information from meteorological data, multi-step combined scanning can be used. That is, in the same data scanning process, multiple scans are performed using different step sizes, and then the results of multiple scans are fused and analyzed. This method can cover a wider range of data and feature changes, and improve the accuracy of prediction.

[0050] 3. Effect of setting step size adjustment By adjusting the step size of the data scanning kernel in the first dimension, the following effects can be achieved: Improve the accuracy of data scanning: By dynamically adjusting the step size, key information and characteristic changes in meteorological data can be extracted more accurately, reducing information loss and errors.

[0051] Improve data processing efficiency: Under the premise of ensuring scanning accuracy, appropriately increasing the step size can improve the speed and efficiency of data scanning and reduce computing resources and time costs.

[0052] Improve the accuracy of weather pattern predictions: By more accurately extracting and analyzing meteorological data, more accurate predictions and identification of complex weather patterns can be achieved, improving the reliability and safety of aviation meteorological services.

[0053] To sum up, adjusting the step size of the data scanning kernel of the first dimension is one of the core inventive points of the embodiment of the present application. It realizes more efficient and accurate extraction and analysis of meteorological data by dynamically adjusting the scanning step size and optimizing the data processing method, providing strong support for aviation meteorological monitoring and forecasting.

[0054] In some optional embodiments, the first aviation meteorological embedded state vectors corresponding to the several data units are determined based on the meteorological feature correlation between the data units in the several first data scan sets, including: determining the first aviation meteorological focus features corresponding to the data units in the first data scan set based on a linkage feature focusing strategy; performing interval numerical mapping on the first aviation meteorological focus features corresponding to the several data units to obtain first aviation meteorological interval quantitative features; and performing residual connection processing on the first aviation meteorological interval quantitative features through a deep residual network to obtain the first aviation meteorological embedded state vector.

[0055] In this embodiment, the big data system first obtains airborne detection data, which are meteorological detection data for weather patterns to be predicted, and are used to characterize meteorological state changes of multidimensional meteorological data. Next, the big data system processes these airborne detection data to extract and express their inherent meteorological characteristics.

[0056] In the first step of processing, the big data system uses a specific data scan to disassemble the airborne detection data to form a number of first data scan sets. Each data scan set contains a portion of data units in the airborne detection data, and these data units represent different meteorological characteristics or conditions.

[0057] Subsequently, the big data system determines the first aviation meteorological focus feature corresponding to each data unit in the first data scan set based on the linkage feature focusing strategy. The linkage feature focusing strategy is a feature selection method that analyzes the correlation between meteorological features between data units to find out the key features, namely the focus features, that have an important impact on weather pattern prediction. These focus features can accurately reflect the main changes in meteorological conditions and are the basis for subsequent forecast analysis. After obtaining the first aviation meteorological focus features, the big data system will perform interval numerical mapping on these features. Interval numerical mapping is a feature conversion technology that maps the original feature value to a specific numerical interval for subsequent quantitative analysis and processing. Through interval numerical mapping, the big data system obtains the first aviation meteorological interval quantitative features, which have better numerical expression capabilities and computability.

[0058] Next, the big data system uses the deep residual network to perform residual connection processing on the first aviation meteorological interval quantitative features. The deep residual network is a deep learning model that optimizes the network structure by introducing residual connections (shortcut connections) to improve the training effect and generalization ability of the model. During the residual connection processing, the big data system inputs the interval quantitative features into the deep residual network, and obtains the first aviation meteorological embedded state vector corresponding to each data unit through network learning and calculation. These embedded state vectors contain rich meteorological information and can fully reflect the position and state of the data unit in the meteorological feature space.

[0059] After obtaining the first aviation meteorological embedded state vector, the big data system will make further adjustments and optimizations. It first adjusts the data scanning kernel of the first dimension in the airborne detection data with a set step size to capture the changes in meteorological data more carefully. Then, based on the adjusted data scanning kernel, the big data system re-disassembles the airborne detection data into several second data scanning sets.

[0060] Based on the second data scan set, the big data system will again use the linkage feature focusing strategy and deep residual network to determine the second aviation meteorological embedded state vector corresponding to each data unit. These second aviation meteorological embedded state vectors can more accurately express the meteorological state of the data unit after considering more dimensions of meteorological features.

[0061] Finally, the big data system predicts the weather patterns in the airborne detection data based on the second aviation meteorological embedded state vector and generates corresponding weather pattern prediction labels. These prediction labels are used to characterize the prediction results of the airborne detection data relative to the flight safety weather conditions, providing strong support for aviation meteorological support.

[0062] In this way, the big data system can achieve deep mining and efficient processing of airborne detection data, and accurately identify and predict complex weather patterns. At the same time, by optimizing data transmission efficiency and improving prediction accuracy, the embodiment of the present application provides a more reliable and secure guarantee for aviation meteorological services. It not only improves the efficiency and quality of aviation meteorological services, but also provides important technical support for flight safety.

[0063] In the next step, the first aviation meteorological focus feature corresponding to the data unit in the first data scan set is determined based on the linkage feature focusing strategy, including: extracting the local aviation meteorological embedded state vectors corresponding to the several data units in the airborne detection data; determining the local aviation meteorological state correlation between the u-th local aviation meteorological embedded state vector corresponding to the u-th data unit in the first data scan set and the v-th local aviation meteorological embedded state vector corresponding to the v-th data unit, where u and v are positive integers; for the u-th data unit in the first data scan set, the local aviation meteorological state correlation corresponding to the u-th data unit in the first data scan set is used as a reinforcement coefficient, and the local aviation meteorological embedded state vector of the data unit in the first data scan set is feature-enhanced to obtain the first aviation meteorological focus feature corresponding to the u-th data unit.

[0064] Based on this embodiment, first, after the big data system obtains the airborne detection data, it will use a specific data scan to check the data and disassemble it to form a number of first data scan sets. Each data scan set contains a part of the data units in the airborne detection data, and these data units represent different meteorological characteristics or conditions.

[0065] Next, the big data system begins to implement the linkage feature focus strategy. The core of this strategy is to find out the key features, namely the focus features, by analyzing the meteorological feature correlations between data units. In this step, the big data system will first extract the local aviation meteorological embedded state vector corresponding to each data unit. The local aviation meteorological embedded state vector is obtained by encoding and embedding the meteorological features of the data unit, which reflects the local state of the data unit in the meteorological feature space.

[0066] Then, the big data system determines the local aviation meteorological state correlation between different data units in the first data scan set. Specifically, the system calculates the correlation between the u-th local aviation meteorological embedded state vector corresponding to the u-th data unit and the v-th local aviation meteorological embedded state vector corresponding to the v-th data unit. This correlation can be obtained by calculating the similarity, distance or other metric between the two vectors. u and v are positive integers, which represent different data units in the first data scan set respectively.

[0067] After determining the correlation of local aviation meteorological status, the big data system will perform feature enhancement on the local aviation meteorological embedded state vector of each data unit. The purpose of this step is to enhance the expression of key features based on the correlation between data units to make them more prominent. Specifically, for the u-th data unit in the first data scan set, the big data system will use the local aviation meteorological status correlation between it and other data units as the enhancement coefficient, and perform weighted processing on the local aviation meteorological embedded state vector corresponding to the u-th data unit. In this way, features that are strongly correlated with the u-th data unit will receive greater weights, thereby playing a more important role in subsequent analysis.

[0068] After feature enhancement processing, the big data system obtains the first aviation meteorological focus features corresponding to each data unit. These focus features not only contain the meteorological information of the data unit itself, but also incorporate the correlation information between other data units, so they can more accurately reflect the status and role of the data unit in the meteorological system.

[0069] In this way, the big data system can achieve deep mining and efficient processing of airborne detection data, thereby more accurately identifying and predicting complex weather patterns. Specifically, by extracting local aviation meteorological embedded state vectors, calculating local aviation meteorological state correlations, and performing feature enhancement processing, the big data system can identify key features that have an important impact on weather pattern predictions, providing strong support for subsequent forecast analysis. This not only improves the accuracy and reliability of aviation meteorological services, but also provides important technical guarantees for flight safety.

[0070] In some preferred technical schemes, the second aviation meteorological embedded state vectors corresponding to the several data units are determined based on the meteorological feature correlation between the first aviation meteorological embedded state vectors in the several second data scan sets, including: determining the second aviation meteorological focus features corresponding to the first aviation meteorological embedded state vectors in the second data scan set according to the linkage feature focusing strategy; performing interval numerical mapping on the second aviation meteorological focus features corresponding to the several data units to obtain second aviation meteorological interval quantitative features; performing residual connection processing on the second aviation meteorological interval quantitative features through a deep residual network to obtain the second aviation meteorological embedded state vector.

[0071] In the above technical solutions, the linkage feature focus strategy is a strategy used to identify key features or focus in data analysis. It analyzes the correlation between data to find the features that have the greatest impact on the results. Interval value mapping is a method of mapping raw data to a specific numerical interval for quantitative analysis and comparison. The deep residual network is a deep learning network structure that introduces residual connections to solve the gradient vanishing problem in deep network training and improve network performance.

[0072] Now, taking a specific application scenario as an example, the following explains in detail how the big data system determines the second aviation meteorological embedded state vectors corresponding to several data units based on the meteorological feature correlation between the first aviation meteorological embedded state vectors in several second data scan sets.

[0073] When an aircraft is in flight, the onboard detection equipment collects a large amount of meteorological data in real time, including temperature, humidity, wind speed and other meteorological characteristics. The big data system first uses the first dimension data scan to disassemble these onboard detection data and generate several first data scan sets. Then, the system analyzes the meteorological feature correlation between the data units in each first data scan set and generates a first aviation meteorological embedded state vector for each data unit.

[0074] Then, the big data system adjusts the step size of the data scan core of the first dimension according to the requirements of the flight mission and the accuracy requirements of the data analysis. This adjustment may include reducing the step size to obtain finer-grained data, or changing the direction of the step size to observe data changes at different angles. The adjusted data scan core re-disassembles the airborne detection data to generate several second data scan sets.

[0075] After obtaining the second data scan set, the big data system starts to implement the linkage feature focus strategy. This strategy determines the second aviation meteorological focus features corresponding to each data unit by analyzing the meteorological feature correlation between the first aviation meteorological embedded state vectors in the second data scan set. These focus features are key meteorological features that affect the weather pattern prediction results. They may be a combination of one or more features of temperature, humidity, and wind speed.

[0076] Next, the big data system performs interval numerical mapping on the second aviation meteorological focus features corresponding to each data unit. This process converts the original meteorological data into quantitative features with a uniform numerical interval, which is convenient for subsequent mathematical analysis and calculation. Through interval numerical mapping, the big data system obtains a set of second aviation meteorological interval quantitative features.

[0077] Finally, the big data system uses the deep residual network to perform residual connection processing on the quantitative features of the second aviation meteorological interval. By introducing residual connections, the deep residual network effectively solves the gradient vanishing problem that may occur during the training process of the deep network and improves the performance of the network. During the residual connection processing, the network can learn more complex meteorological feature relationships between data units and generate a more accurate second aviation meteorological embedded state vector based on this.

[0078] Through this series of data processing and analysis steps, the big data system generates a second aviation meteorological embedded state vector for each data unit. These embedded state vectors not only contain the location and state information of the data unit in the meteorological feature space, but also contain the relative relationship and spatial structure between the data units. Based on these embedded state vectors, the big data system can more accurately predict weather patterns and provide more reliable prediction results for flight safety.

[0079] In this way, through the disassembly of data scanning kernels, analysis of meteorological feature correlations, application of linkage feature focusing strategies, interval value mapping, and residual connection processing of deep residual networks, in-depth analysis and processing of meteorological data in airborne detection data are achieved. This technical solution improves the accuracy and reliability of weather pattern prediction and provides a strong guarantee for flight safety. At the same time, this technical solution also demonstrates the application potential and value of big data and machine learning technologies in the field of meteorological data analysis.

[0080] In some other preferred technical schemes, the airborne detection data is decomposed into x first data scan sets by the data scanning core of the first dimension, where x is a positive integer; the second aviation meteorological focus feature corresponding to the first aviation meteorological embedded state vector in the second data scan set is determined according to the linkage feature focusing strategy, including: determining that the number of the second data scan sets obtained by decomposing the airborne detection data after the data scanning core of the first dimension wanders is y, where y is a positive integer and y>x; cross-matching the data units in the second data scan set based on time series features to obtain x third data scan sets; and determining the second aviation meteorological focus feature corresponding to the first aviation meteorological embedded state vector in the third data scan set according to the linkage feature focusing strategy.

[0081] When explaining the implementation example of this technical solution in detail, a specific application scenario will be used to illustrate how the big data system determines the second aviation meteorological focus feature based on the linkage feature focusing strategy.

[0082] For example, when an aircraft is performing a long-distance flight mission, the onboard detection equipment continuously collects meteorological data, including temperature, humidity, wind speed and other meteorological characteristics. These data change in real time and are critical to flight safety. The big data system first uses the first-dimensional data scan to check the onboard detection data and generates x first data scan sets. Each data scan set contains a series of meteorological data units within a specific time period.

[0083] However, in order to further refine the analysis and capture subtle changes in meteorological data, the big data system decided to adjust the step size of the data scan kernel and re-disassemble the airborne detection data. The adjusted data scan kernel performs a wandering disassembly of the airborne detection data and obtains y second data scan sets, where y is a positive integer and y>x. This means that the new disassembly method provides more data subsets, each of which reflects changes in meteorological data at a finer granularity or at different angles.

[0084] After obtaining the second data scan set, the big data system did not directly analyze these new data subsets, but adopted a cross-matching method based on time series characteristics. Since meteorological data has obvious time series, that is, data changes over time, the big data system believes that by cross-matching data units at different time points, the internal connection and change pattern between meteorological characteristics can be better revealed.

[0085] Therefore, the big data system cross-matches the data units in the second data scan set based on time series features. Specifically, the system matches each data unit with similar data units at other time points based on the timestamp, forming a new set of data subsets, namely x third data scan sets. These third data scan sets not only contain the meteorological feature information in the original data, but also introduce time series correlation information through cross-matching.

[0086] Next, the big data system determines the second aviation meteorological focus feature corresponding to the first aviation meteorological embedded state vector in the third data scan set based on the linkage feature focusing strategy. The linkage feature focusing strategy is a data analysis method that analyzes the correlation between different features in the data set to find the key features that have the greatest impact on the results. In this scenario, the big data system identifies the feature combination that has the greatest impact on the weather pattern forecast results, namely the second aviation meteorological focus feature, by analyzing the correlation between the meteorological features in the third data scan set.

[0087] Through this process, the big data system not only obtains a more fine-grained subset of meteorological data, but also determines the key meteorological features that affect weather pattern prediction results through cross-matching of time series features and linkage feature focus strategies. These key features will provide strong support for subsequent weather pattern predictions, making the prediction results more accurate and reliable.

[0088] It can be seen that by adjusting the step size of the data scanning kernel, cross-matching based on time series features, and applying the linkage feature focusing strategy, in-depth analysis and processing of airborne detection data is achieved. This technical solution not only improves the accuracy and efficiency of meteorological data analysis, but also provides more accurate key feature information for weather pattern prediction. This is of great significance for ensuring flight safety and improving the accuracy of weather forecasts. At the same time, this technical solution also demonstrates the strong potential and application value of big data and machine learning technologies in the field of meteorological data analysis.

[0089] Under other possible design ideas, the method of determining the second aviation meteorological focus feature corresponding to the first aviation meteorological embedded state vector in the third data scan set based on the linkage feature focusing strategy includes: obtaining the timing cross-matching information corresponding to the pth third data scan set, the timing cross-matching information is used to characterize the timing continuity of the second data scan set after cross-matching, p is a positive integer, p≤x; obtaining the aviation state annotation information corresponding to the pth third data scan set based on the timing cross-matching information; determining the embedded vector commonality value between the first aviation meteorological embedded state vectors in the third data scan set based on the linkage feature focusing strategy; weighting the embedded vector commonality value and the aviation state annotation information to obtain the aviation state weighted commonality value; and performing feature enhancement on the first aviation meteorological embedded state vector of the data unit in the third data scan set based on the aviation state weighted commonality value to obtain the second aviation meteorological focus feature.

[0090] When further explaining in detail how the big data system determines the second aviation meteorological focus feature corresponding to the first aviation meteorological embedded state vector in the third data scan set based on the linkage feature focusing strategy, it will be explained through a specific application scenario.

[0091] For example, when an aircraft flies over a complex meteorological area, the onboard detection equipment continuously collects meteorological data. The big data system has obtained x third data scan sets through the adjustment of the data scan kernel and the cross-matching of the time series features. Each third data scan set contains meteorological data units after cross-matching, and these units maintain a certain continuity in time series.

[0092] First, the big data system starts to process the pth third data scan set, where p is a positive integer and p≤x. The system obtains the time series cross-matching information corresponding to the data scan set. Time series cross-matching information refers to the information about the continuity and correlation of data units in time series recorded by the system during the cross-matching process. This information is crucial for understanding the changing trends and internal laws of meteorological data.

[0093] Next, the big data system obtains the aviation status annotation information corresponding to the pth third data scan set based on the time series cross-matching information. Aviation status annotation information refers to the information that annotates the impact of meteorological data on flight safety based on the time series cross-matching results. For example, the system may mark certain meteorological data as "potentially dangerous areas" to alert pilots.

[0094] Then, the big data system determines the embedding vector commonality value between the first aviation meteorological embedded state vectors in the third data scan set based on the linkage feature focusing strategy. The embedding vector commonality value refers to the similarity or common features of different embedded state vectors in the meteorological feature space. By calculating the embedding vector commonality value, the system can find the embedding state vector combination that has the greatest impact on the weather pattern prediction results.

[0095] Next, the big data system performs weighted processing on the embedding vector commonality values ​​and aviation status annotation information. Weighted processing means weighting the embedding vector commonality values ​​according to the importance of aviation status annotation information. Through this step, the system can further highlight the role of key meteorological features in weather pattern prediction.

[0096] Finally, the big data system performs feature enhancement on the first aviation meteorological embedded state vector of the data unit in the third data scan set based on the weighted commonality value of the aviation state. Feature enhancement means increasing the weight of key meteorological features or adjusting their numerical values ​​so that these features play a greater role in subsequent weather pattern forecasts. The embedded state vector after feature enhancement, namely the second aviation meteorological focus feature, will serve as an important input for weather pattern forecasts.

[0097] Through this embodiment, the big data system not only considers the continuity and correlation of meteorological data in time series, but also combines aviation status annotation information and embedded vector commonality values ​​to enhance the key meteorological features. This enables the system to more accurately identify the combination of meteorological features that has the greatest impact on flight safety, thereby improving the accuracy and reliability of weather pattern prediction. This technical solution is of great significance in ensuring flight safety and improving the quality of weather forecasts.

[0098] In some other embodiments, the step size of setting the data scanning core of the first dimension in the airborne detection data is adjusted to re-disassemble the airborne detection data into a plurality of second data scanning sets, including: wandering the data scanning core of the first dimension at intervals of w data units on a first wandering path corresponding to the airborne detection data; wandering the data scanning core of the first dimension at intervals of r data units on a second wandering path corresponding to the airborne detection data to determine the plurality of second data scanning sets; wherein w and r are positive integers.

[0099] For example, when an aircraft flies over a changing weather area, the onboard detection equipment collects a large amount of meteorological data in real time. These data cover multiple dimensions such as temperature, humidity, wind speed, and wind direction, which are crucial for flight safety. The big data system has used the first-dimensional data scan core to preliminarily disassemble these airborne detection data and generate several first data scan sets. However, in order to more comprehensively capture the subtle changes and inherent laws of meteorological data, the system decided to adjust the step size of the first-dimensional data scan core and re-disassemble the data.

[0100] In this embodiment, the big data system uses two different walking paths to adjust the step size of the data scanning core. First, on the first walking path corresponding to the airborne detection data, the system walks the data scanning core of the first dimension at intervals of w data units. This means that each time the data is disassembled, the data scanning core will skip w data units to obtain a new data subset. By adjusting the size of w, the system can observe the changing trend of meteorological data at different data granularities.

[0101] Next, on the second wandering path corresponding to the airborne detection data, the system wanders the data scanning core of the first dimension at intervals of r data units. The r here is different from w on the first wandering path, and it represents another granularity of data decomposition. By changing the value of r, the system can capture the characteristics of meteorological data from another perspective. It is worth noting that the first wandering path and the second wandering path may be based on different data features or time series in order to analyze meteorological data from multiple dimensions.

[0102] During the walking process, the big data system records the data subsets obtained from each disassembly and combines them into several second data scan sets. These second data scan sets contain meteorological data of different granularities and angles because they use different walking paths and step sizes. They provide a rich data foundation for the subsequent linkage feature focusing strategy, embedding vector commonality value calculation and feature enhancement processing.

[0103] Through this embodiment, the big data system can flexibly adjust the step size of the data scanning kernel and disassemble the airborne detection data from multiple angles and granularities. This not only helps to fully capture the subtle changes and inherent laws of meteorological data, but also improves the accuracy and reliability of weather pattern prediction. This technical solution has important application value in the fields of flight safety prediction, meteorological data analysis, etc.

[0104] In this way, by adjusting the step size of the data scanning kernel and disassembling the airborne detection data by using multiple walk paths, the big data system can obtain richer and more multi-dimensional meteorological data information. This information provides strong support for subsequent feature analysis, pattern recognition and prediction, making weather pattern prediction results more accurate and reliable.

[0105] In an optional embodiment, predicting the weather pattern in the airborne detection data based on the second aviation meteorological embedded state vector to obtain a weather pattern prediction label corresponding to the airborne detection data includes: obtaining an aviation mission description feature, the aviation mission description feature is used to assist in predicting the weather pattern corresponding to the airborne detection data; inputting the aviation mission description feature and the second aviation meteorological embedded state vector into a pre-debugged aviation meteorological prediction network to obtain the weather pattern prediction label.

[0106] In a more refined embodiment, when the big data system predicts the weather pattern in the airborne detection data based on the second aviation meteorological embedded state vector, it not only considers the meteorological data itself, but also introduces the aviation mission description features as auxiliary information to further improve the accuracy of the prediction. The specific steps and beneficial effects of this embodiment will be explained in detail below.

[0107] For example, an airplane is performing a long-distance cargo mission, and its flight route passes through multiple areas with complex meteorological conditions. In order to ensure flight safety, the big data system needs to accurately predict the weather patterns along the flight path. First, the system has acquired the airborne detection data through the previous steps, and generated a second aviation meteorological embedded state vector for each data unit through methods such as disassembly of the data scanning core and linkage feature focusing strategy.

[0108] However, relying solely on meteorological data for prediction may not be comprehensive enough. This is because different aviation missions have different requirements for weather conditions. For example, cargo missions may pay more attention to wind speed and direction, while passenger missions may pay more attention to precipitation and visibility. Therefore, in this embodiment, the big data system introduces aviation mission description features as auxiliary prediction information.

[0109] Aviation mission description features can include multiple dimensions such as the type of flight mission (such as cargo, passenger, military, etc.), flight route, estimated flight time, flight altitude, aircraft model, etc. These features can be obtained through channels such as airline mission planning and flight management systems. The big data system associates these features with the second aviation meteorological embedded state vector to more fully understand the requirements and restrictions of the current flight mission on weather conditions.

[0110] Next, the big data system inputs the aviation mission description features and the second aviation weather embedded state vector into the pre-debugged aviation weather forecast network. This forecast network is a deep learning model that has been trained and optimized with a large amount of historical data. It can automatically learn and identify the relationship and rules between different weather patterns and flight missions based on the input meteorological data and mission characteristics.

[0111] Through the calculation of the aviation weather forecast network, the big data system finally obtains the weather pattern forecast label. This label not only contains the weather pattern information in the current or future period (such as sunny, cloudy, rainy, stormy, etc.), but also combines the characteristics of the aviation mission description to comprehensively evaluate the impact of weather patterns on flight missions. For example, if it is predicted that there will be a strong storm in a certain period of time in the future, and the storm may have a serious impact on the flight safety of the cargo mission, then the forecast label can specifically mark this point to remind pilots and airlines to take corresponding measures.

[0112] Through this implementation, the big data system not only utilizes the information of the meteorological data itself, but also combines the characteristics of aviation mission description to achieve a more comprehensive and accurate prediction of weather patterns. This prediction method not only helps to improve the level of flight safety, but also provides airlines with more scientific decision support and optimizes flight plans and resource allocation. At the same time, this technical solution also demonstrates the broad application prospects of big data and deep learning technology in the fields of meteorological forecasting and aviation safety.

[0113] In the next step, the aviation mission description feature and the second aviation meteorological embedded state vector are input into the pre-debugged aviation meteorological forecast network to obtain the weather pattern prediction label, including: inputting the aviation mission description feature and the second aviation meteorological embedded state vector into the pre-debugged aviation meteorological forecast network to obtain the meteorological state bias prediction information of the aviation meteorological forecast network; obtaining a set label index indication, the label index indication is used to match and output prediction keywords that meet the target aviation weather requirements; and updating the meteorological state bias prediction information through the label index indication to obtain the weather pattern prediction label.

[0114] When further explaining the steps of how the big data system utilizes the aviation mission description features and the second aviation meteorological embedded state vector to predict weather patterns, it will be elaborated on how to generate weather pattern prediction labels through the pre-debugged aviation meteorological prediction network and how to more accurately meet specific aviation weather needs through label index indication.

[0115] For example, an airplane is carrying out an important passenger mission, and its flight route passes through multiple regions with variable climates. In order to ensure the safety and comfort of passengers, the big data system needs to accurately predict the weather patterns along the flight path, and pay special attention to those weather conditions that may have an adverse effect on the flight.

[0116] First, the big data system has acquired the airborne detection data through the previous steps and generated the second aviation meteorological embedded state vector. At the same time, the system also obtains descriptive features related to the current aviation mission, such as the type of flight mission (passenger transport), flight route, and estimated flight time.

[0117] Next, the big data system inputs the aviation mission description features and the second aviation weather embedded state vector into the pre-debugged aviation weather forecast network. This network is a deep learning model that has been trained and optimized with a large amount of historical meteorological data and flight mission data. It can automatically learn and identify the relationship and rules between different weather patterns and flight missions based on the input feature vector.

[0118] When the input feature vector enters the aviation weather forecast network, the network will perform a series of complex calculations and processing, and finally output the weather state bias forecast information. This information is a quantitative description of the weather pattern in the future, including the change trend and possible range of multiple meteorological elements such as temperature, humidity, wind speed, wind direction, etc.

[0119] However, relying solely on meteorological state bias forecast information may not be enough to meet specific aviation weather needs. For example, in passenger missions, passengers may have high requirements for comfort, so special attention needs to be paid to weather conditions that may cause turbulence or discomfort. In order to meet these needs more accurately, the big data system introduces the tool of label index indication.

[0120] The label index indication is a set of preset rules or mapping tables used to match and output forecast keywords that meet the target aviation weather requirements. For example, the label index indication may stipulate that when the wind speed is predicted to exceed a certain threshold, the label of "strong wind weather" is output; when the visibility is predicted to be lower than a certain standard, the label of "low visibility weather" is output, etc. These labels not only describe the basic information of weather patterns, but also combine the specific needs of aviation missions, providing more intuitive and targeted guidance for pilots and airlines.

[0121] The big data system updates the weather state bias forecast information through label index indication. Specifically, the system first filters and matches the forecast information according to the rules indicated by the label index to find keywords or labels that meet specific weather requirements. Then, the system combines these keywords or labels with the original weather state bias forecast information to generate the final weather pattern forecast label.

[0122] Through this embodiment, the big data system not only uses the aviation mission description features and the second aviation meteorological embedded state vector to predict weather patterns, but also performs more refined processing of the prediction results through label index indication. This processing method not only improves the accuracy of the prediction, but also makes the prediction results more in line with the specific needs of the aviation mission. This is of great significance for ensuring flight safety and improving flight comfort. At the same time, this technical solution also demonstrates the strong potential and application value of big data and deep learning technology in the fields of meteorological forecasting and aviation safety.

[0123] In some further preferred embodiments, the weather pattern prediction label includes the meteorological state jump type, meteorological state distribution characteristics and meteorological state risk characteristics of the weather pattern. After predicting the weather pattern in the airborne detection data based on the second aviation meteorological embedded state vector to obtain the weather pattern prediction label corresponding to the airborne detection data, it also includes: based on the meteorological state distribution characteristics indicated by the weather pattern prediction label, decomposing the airborne detection data into local airborne detection results corresponding to several weather patterns; predicting the state jump characteristics within the local airborne detection results based on the meteorological state jump type indicated by the weather pattern prediction label, and generating state jump mapping data corresponding to the local airborne detection results; judging the state jump mapping data corresponding to the several weather patterns based on the meteorological state risk characteristics indicated by the weather pattern prediction label to generate an aviation risk judgment opinion.

[0124] When further exploring how big data systems can implement weather pattern prediction and conduct in-depth analysis based on the prediction results. For example, an aircraft is performing a cross-border flight mission, and the flight path passes through a variety of different climate regions. In order to ensure flight safety, the big data system needs to accurately predict the weather patterns in the airborne detection data and generate detailed local weather analysis and risk judgment opinions based on the prediction results.

[0125] First, the big data system uses the second aviation meteorological embedded state vector through the previous steps to predict the weather pattern in the airborne detection data and generate the corresponding weather pattern prediction label. In this embodiment, the weather pattern prediction label not only contains the basic information of the weather pattern, but is also further subdivided into three parts: meteorological state jump type, meteorological state distribution characteristics and meteorological state risk characteristics. The meteorological state jump type describes the type of change that may occur in various meteorological elements (such as temperature, wind speed, etc.) in the weather pattern; the meteorological state distribution characteristics reflect the distribution of different weather patterns in time and space; the meteorological state risk characteristics evaluate the degree of risk that each weather pattern may bring to flight safety.

[0126] After obtaining the weather pattern forecast label, the big data system begins to disassemble the airborne detection data based on the meteorological state distribution characteristics in the label. Specifically, the system will divide the airborne detection data into several local areas according to the distribution of different weather patterns indicated by the label, and each area corresponds to one or more weather patterns. This disassembly helps the system to more accurately analyze the specific impact of different weather patterns on flight safety.

[0127] Next, the big data system predicts the state jump characteristics in each local airborne detection result based on the type of meteorological state jump in the weather pattern prediction label. The state jump characteristics describe the possible change trend and speed of various meteorological elements under a specific weather pattern. By generating state jump mapping data, the system can more intuitively display these changes and provide timely warnings and guidance to pilots.

[0128] Finally, the big data system determines the state transition mapping data corresponding to each local airborne detection result based on the meteorological state risk characteristics in the weather pattern forecast label. This step aims to evaluate the potential risks of different weather patterns to flight safety and generate corresponding aviation risk judgment opinions. These opinions not only include the evaluation of risk levels, but may also include specific risk sources, impact ranges, and response measures, providing comprehensive decision support for pilots and airlines.

[0129] Through this embodiment, the big data system not only realizes the accurate prediction of weather patterns in airborne detection data, but also conducts in-depth analysis and judgment based on the prediction results. This comprehensive analysis method not only improves the level of flight safety protection, but also provides airlines with a more scientific and comprehensive basis for decision-making. Specifically, the system comprehensively reveals the impact mechanism of different weather patterns on flight safety through multiple steps such as disassembling airborne detection data, predicting state jump characteristics, and judging aviation risks, providing pilots with more accurate and timely guidance information. At the same time, this comprehensive analysis method can also provide strong support for airlines to optimize flight plans, improve service quality, etc., further promoting the safe and efficient development of the aviation industry.

[0130] In an optional embodiment, the method is implemented by a pre-debugged weather pattern prediction algorithm, which is debugged by a prediction algorithm to be debugged, and the debugging process of the prediction algorithm to be debugged includes: obtaining airborne detection data samples and a weather pattern certification sample set corresponding to the airborne detection data samples, the weather pattern certification sample set being used to characterize the weather pattern training annotation chain corresponding to the airborne detection data samples; inputting the airborne detection data samples into the prediction algorithm to be debugged to generate a weather pattern prediction training result; and based on the difference between the weather pattern prediction training result and the weather pattern certification sample set, loop debugging the prediction algorithm to be debugged to obtain the weather pattern prediction algorithm.

[0131] In a more refined and systematic embodiment, the big data system predicts the weather pattern in the airborne detection data through a pre-debugged weather pattern prediction algorithm. This pre-debugged weather pattern prediction algorithm is not directly applied to the actual prediction task, but is obtained through careful debugging and optimization to ensure its accuracy and reliability. The following will explain in detail the debugging process of the prediction algorithm to be debugged and how to obtain the pre-debugged weather pattern prediction algorithm through this process.

[0132] First, in order to debug and optimize the prediction algorithm to be debugged, the big data system needs to obtain two sets of key data: airborne detection data samples and weather pattern certification sample sets corresponding to the airborne detection data samples. The airborne detection data samples are a series of representative meteorological detection data in actual flight processes, which contain meteorological data characteristics under various weather modes. The weather pattern certification sample set is the weather pattern annotation information corresponding to these airborne detection data samples, which has been certified by experts or verified by historical data. This certification sample set actually constitutes a weather pattern training annotation chain, which provides clear learning objectives and evaluation criteria for the prediction algorithm to be debugged.

[0133] After obtaining these two sets of data, the big data system inputs the airborne detection data samples into the prediction algorithm to be debugged, allowing the algorithm to generate weather pattern prediction training results based on the input data. This process simulates the scenarios in the actual prediction task, allowing the prediction algorithm to be debugged to initially demonstrate its prediction ability.

[0134] However, it is not enough to just generate forecast results. The big data system also needs to evaluate the accuracy of the forecast results. To this end, the big data system compares the forecast training results with the weather pattern certification sample set to find out the differences and differences between the two. These differences may be reflected in the type of weather pattern predicted, the trend of meteorological elements, the assessment of risk level, and many other aspects.

[0135] Based on these comparison results, the big data system begins to debug the prediction algorithm to be debugged in a loop. The debugging process can include modifying the internal parameters of the algorithm, adjusting the model structure, adding new learning features, and other means. Through continuous trial and error, the big data system gradually optimizes the performance of the prediction algorithm to be debugged, so that it can better fit the actual meteorological data characteristics and improve the accuracy of the prediction.

[0136] After multiple rounds of cyclic debugging and optimization, the performance of the prediction algorithm to be debugged has been significantly improved, and the big data system has identified it as the pre-debugged weather pattern prediction algorithm. This pre-debugged algorithm not only has a high prediction accuracy, but also can maintain stable performance under different weather modes. Therefore, when the big data system faces actual airborne detection data, it can quickly and accurately generate the corresponding weather pattern prediction labels, providing strong protection for flight safety.

[0137] Through this example, the big data system demonstrates a systematic and refined debugging process for the weather pattern prediction algorithm. This process not only improves the accuracy and reliability of the prediction algorithm, but also ensures the stability and usability of the algorithm in practical applications. This optimization method based on data-driven and loop debugging is not only applicable to the field of weather pattern prediction, but can also be extended to other scenarios that require high-precision predictions, and has broad application prospects and important practical value.

[0138] In an independent embodiment, the obtaining of a weather pattern authentication sample set corresponding to the airborne detection data sample includes: (1) obtaining meteorological environment information of the airborne detection data sample, wherein the meteorological environment information includes training and learning annotations of the weather pattern in the airborne detection data sample; (2) adjusting the airborne detection data sample into a meteorological monitoring relationship network, wherein the meteorological monitoring relationship network includes a meteorological status relationship subnet and an abnormal warning relationship subnet; (3) updating the meteorological monitoring relationship network based on the weather pattern, disassembling the meteorological monitoring relationship network according to the connection relationship between the entity update information based on the weather pattern, and obtaining a plurality of first knowledge relationship spectra, wherein the knowledge relationship feature set corresponding to the first knowledge relationship spectra includes pre-order meteorological knowledge derivation information and reverse order meteorological state tracing information; (4) performing entity updates based on the spatiotemporal modality on the plurality of first knowledge relationship spectra, respectively, and disassembling the first knowledge relationship spectra according to the connection relationship between the entity update information based on the spatiotemporal modality to obtain a plurality of second knowledge relationship spectra, wherein the knowledge relationship feature sets corresponding to the second knowledge relationship spectra include time-domain meteorological knowledge features and space-domain meteorological knowledge features; (5) determining at least one meteorological knowledge coverage area according to the preceding meteorological knowledge derivation information and the reverse meteorological state tracing information corresponding to the first knowledge relationship spectrum, and the time domain meteorological knowledge features and the space domain meteorological knowledge features corresponding to the second knowledge relationship spectrum; (6) performing iterative decomposition based on the entity update information based on the weather pattern and the entity update information based on the spatiotemporal modality in the meteorological knowledge coverage area; (7) On the basis that there is no new knowledge relationship spectrum under the entity update information based on the weather pattern, or the number of knowledge relationship spectra under the entity update information based on the spatiotemporal modality is not greater than a set number, the weather pattern in the meteorological environment information is judged based on the decomposition order of the knowledge relationship spectrum to obtain the weather pattern authentication sample set.

[0139] When explaining this independent embodiment in detail, a specific scenario will be used to illustrate how the big data system obtains the weather pattern certification sample set corresponding to the airborne detection data sample. This process not only involves in-depth analysis of the airborne detection data, but also integrates the multi-dimensional association and iterative decomposition of meteorological knowledge to ensure the accuracy and comprehensiveness of the weather pattern certification sample set.

[0140] First, the big data system extracts meteorological environment information from airborne detection data. This meteorological environment information contains rich meteorological data, such as temperature, humidity, wind speed, wind direction, etc., as well as training and learning annotations of corresponding weather patterns. These annotations can be given by meteorological experts based on historical data and professional knowledge, providing an important reference for subsequent weather pattern predictions.

[0141] Next, the big data system adjusts the airborne detection data samples into a meteorological monitoring relationship network. This relationship network is a complex meteorological data association network, which includes a meteorological state relationship subnet and an abnormal warning relationship subnet. The meteorological state relationship subnet mainly describes the relationship and action mechanism between different meteorological elements, while the abnormal warning relationship subnet focuses on identifying possible meteorological anomalies and potential risks.

[0142] Then, the big data system updates the entities of the meteorological monitoring network based on the weather pattern. This process actually updates the entities in the network based on the latest meteorological data to reflect the changing trend of the current weather pattern. New connection relationships are formed between the updated entity information. The big data system disassembles the meteorological monitoring network based on these connection relationships to obtain several first knowledge relationship spectra. Each first knowledge relationship spectrum corresponds to a knowledge relationship feature set, which contains the previous meteorological knowledge derivation information and the reverse meteorological state tracing information. The previous meteorological knowledge derivation information refers to the causes and processes that lead to the current meteorological state, while the reverse meteorological state tracing information describes the future meteorological state that may evolve from the current state.

[0143] Subsequently, the big data system updates each first knowledge relationship spectrum based on the spatiotemporal modality. The spatiotemporal modality refers to the distribution and change characteristics of meteorological data in time and space. Through entity updates based on the spatiotemporal modality, the big data system can more accurately capture the dynamic changes of meteorological data in time and space. The updated entity information is connected again, and the big data system disassembles the first knowledge relationship spectrum according to these connection relationships to obtain several second knowledge relationship spectra. The knowledge relationship feature set corresponding to each second knowledge relationship spectrum includes time domain meteorological knowledge features and space domain meteorological knowledge features. The time domain meteorological knowledge features describe the changing laws of meteorological data in the time dimension, while the space domain meteorological knowledge features reflect the characteristics of meteorological data in spatial distribution.

[0144] After obtaining the first knowledge relationship spectrum and the second knowledge relationship spectrum, the big data system determines at least one meteorological knowledge coverage area based on their corresponding knowledge relationship feature sets. These coverage areas represent the scope of influence of the current weather pattern in the meteorological knowledge space. The big data system performs iterative decomposition within the meteorological knowledge coverage area, and gradually refines the knowledge relationship spectrum based on the entity update information of the weather pattern and the entity update information based on the spatiotemporal modality. This process will continue until no new knowledge relationship spectrum is generated under the entity update information based on the weather pattern, or the number of knowledge relationship spectra under the entity update information based on the spatiotemporal modality is not greater than the set number.

[0145] Finally, after the disassembly is completed, the big data system identifies the weather patterns in the meteorological environment information based on the disassembly order of the knowledge relationship spectrum. This process comprehensively considers information from multiple dimensions, including the derivation of previous meteorological knowledge, the tracing of reverse meteorological status, the characteristics of meteorological knowledge in the time domain, and the characteristics of meteorological knowledge in the spatial domain, to ensure the accuracy and comprehensiveness of weather pattern authentication. After the identification, the big data system generates a weather pattern authentication sample set, which provides important training data and evaluation criteria for subsequent weather pattern prediction algorithms.

[0146] Through this implementation, the big data system not only realizes the in-depth analysis and processing of airborne detection data, but also ensures the accuracy and comprehensiveness of the weather pattern certification sample set through multi-dimensional meteorological knowledge association and iterative disassembly. This technical solution not only improves the accuracy and reliability of weather pattern prediction, but also provides a new idea and method for the field of meteorological data analysis, which has important application value and promotion significance.

[0147] In other independent embodiments, the meteorological monitoring relationship network is updated based on the weather pattern to obtain entity update information based on the weather pattern, including: obtaining past airborne detection data, the past airborne detection data including the meteorological parameters and status information of the current flight area; matching the past airborne detection data with the entities in the meteorological monitoring relationship network to determine the key entities directly related to the current weather pattern; dynamically updating the status of the key entities based on the past airborne detection data and the preset weather pattern model to reflect the meteorological characteristics of the current weather pattern; during the entity update process, recording the meteorological parameters and status information related to the key entity status changes to form entity update information based on the weather pattern; the weather pattern-based entity update information includes the update status of the key entities, the update time, the associated meteorological parameter changes and the weather pattern feature description, which are used for the subsequent meteorological knowledge relationship disassembly and weather pattern discrimination.

[0148] In this embodiment, the big data system first obtains past airborne detection data, which contains meteorological parameters and status information of the current flight area. These meteorological parameters may include temperature, humidity, wind speed, wind direction, etc., while the status information reflects the status of different meteorological elements at a specific time point. Next, the big data system matches the past airborne detection data with entities in the meteorological monitoring relationship network. The meteorological monitoring relationship network is a complex meteorological data association network that contains multiple meteorological-related entities, such as pressure systems, wind fields, precipitation clouds, etc. Through the matching process, the system can determine the key entities directly related to the current weather pattern. After determining the key entities, the big data system dynamically updates the status of these key entities based on the past airborne detection data and the preset weather pattern model. The weather pattern model is a trained and optimized mathematical model that can predict the changing trend of the weather pattern based on meteorological data. By applying this model, the system can update the status of key entities based on past airborne detection data to reflect the meteorological characteristics of the current weather pattern. During the entity update process, the big data system records the meteorological parameters and status information related to the state changes of the key entities. This information includes the update status of key entities, update time, associated meteorological parameter changes, and weather pattern feature descriptions. This information is crucial for the subsequent decomposition of meteorological knowledge relationships and identification of weather patterns, as it provides important clues about the formation and evolution of weather patterns.

[0149] In this way, the big data system realizes the entity update of the meteorological monitoring relationship network based on weather patterns and spatiotemporal modalities, and generates corresponding entity update information. These entity update information not only provides detailed information about the current weather pattern and past meteorological data, but also provides strong data support for the subsequent meteorological knowledge relationship disassembly and weather pattern prediction. In detail, by updating the entity status in the meteorological monitoring relationship network in real time, the system can more accurately reflect the changing trend of the current weather pattern and meteorological status, thereby improving the accuracy of weather pattern prediction. The entity update information based on weather patterns and spatiotemporal modalities provides pilots and airlines with rich meteorological data support, helping them make more scientific flight decisions. By deeply mining the knowledge relationships and weather pattern characteristics in meteorological data, the system can provide more comprehensive and accurate information support for meteorological services and optimize the quality and efficiency of meteorological services.

[0150] Further, Figure 2 This is a schematic diagram of the structure of a big data system 200 provided in an embodiment of the present application. Figure 2 The big data system 200 shown includes a processor 210, which can call and run a computer program from a memory to implement the method in the embodiment of the present application. Figure 2As shown, the big data system 200 may further include a memory 230. The processor 210 may call and run a computer program from the memory 230 to implement the method in the embodiment of the present application. The memory 230 may be a separate device independent of the processor 210, or may be integrated in the processor 210. Optionally, as Figure 2 As shown, the big data system 200 may also include a transceiver 220, and the processor 210 may control the transceiver 220 to interact with other devices, specifically, to send information or data to other devices, or to receive information or data sent by other devices. Optionally, the big data system 200 may implement the corresponding processes corresponding to the storage engine or the components (such as processing modules) in the storage engine or the device deployed with the storage engine in each method of the embodiment of the present application, which will not be described here for the sake of brevity.

[0151] Based on the above, a computer-readable storage medium is provided, on which a computer program is stored, and the computer program implements the above method when running.

[0152] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations.

Claims

1. A method for analyzing airborne detection data for intelligent meteorology, characterized in that: Applied to a big data system, the method comprises: Acquire airborne detection data, the airborne detection data being meteorological detection data whose weather pattern is to be predicted, the weather pattern being used to characterize meteorological state changes of multidimensional meteorological data in the airborne detection data, the airborne detection data being decomposed into a plurality of first data scan sets by a data scan core of a first dimension, the airborne detection data being decomposed into a plurality of data units by a second dimension, the first dimension being greater than the second dimension; Determining first aviation meteorological embedded state vectors corresponding to the plurality of data units respectively according to meteorological characteristic correlations between the plurality of data units in the plurality of first data scan sets; The data scanning core of the first dimension is adjusted by setting a step length in the airborne detection data, and the airborne detection data is re-disassembled into a plurality of second data scanning sets; Determining second aviation meteorological embedded state vectors corresponding to the plurality of data units respectively according to meteorological feature correlations between the first aviation meteorological embedded state vectors in the plurality of second data scan sets; The weather pattern in the airborne detection data is predicted according to the second aviation meteorological embedded state vector to obtain a weather pattern prediction label corresponding to the airborne detection data, wherein the weather pattern prediction label is used to characterize the prediction result of the airborne detection data relative to the flight safety weather state.

2. The method according to claim 1, characterized in that The step of determining first aviation meteorological embedded state vectors corresponding to the plurality of data units respectively based on meteorological characteristic correlations between the plurality of data units in the plurality of first data scan sets includes: Determining a first aviation meteorological focus feature corresponding to a data unit in the first data scan set according to a linkage feature focusing strategy; Performing interval numerical mapping on the first aviation meteorological focus features respectively corresponding to the plurality of data units to obtain first aviation meteorological interval quantitative features; The first aviation meteorological interval quantization features are subjected to residual connection processing through a deep residual network to obtain the first aviation meteorological embedded state vector.

3. The method according to claim 2, characterized in that The determining, according to the linkage feature focusing strategy, the first aviation meteorological focus feature corresponding to the data unit in the first data scan set includes: Extracting local aviation meteorological embedded state vectors corresponding to the plurality of data units in the airborne detection data; Determine the local aviation meteorological state correlation between the u-th local aviation meteorological embedded state vector corresponding to the u-th data unit in the first data scan set and the v-th local aviation meteorological embedded state vector corresponding to the v-th data unit, where u and v are positive integers; For the u-th data unit in the first data scan set, the correlation between the local aviation meteorological state corresponding to the u-th data unit in the first data scan set is used as the reinforcement coefficient, and the local aviation meteorological embedded state vector of the data unit in the first data scan set is feature enhanced to obtain the first aviation meteorological focus feature corresponding to the u-th data unit.

4. The method according to any one of claims 1 to 3, characterized in that: The determining, based on the meteorological characteristic correlation between the first aviation meteorological embedded state vectors in the plurality of second data scan sets, the second aviation meteorological embedded state vectors corresponding to the plurality of data units respectively comprises: Determining a second aviation meteorological focus feature corresponding to the first aviation meteorological embedded state vector in the second data scan set according to a linkage feature focusing strategy; Performing interval numerical mapping on the second aviation meteorological focus features respectively corresponding to the plurality of data units to obtain second aviation meteorological interval quantitative features; The second aviation meteorological interval quantization features are subjected to residual connection processing through a deep residual network to obtain the second aviation meteorological embedded state vector.

5. The method according to claim 4, characterized in that The airborne detection data is decomposed into x first data scan sets by the data scan core of the first dimension, where x is a positive integer; The determining, according to the linkage feature focusing strategy, the second aviation meteorological focus feature corresponding to the first aviation meteorological embedded state vector in the second data scan set includes: Determine that the number of the second data scan sets obtained by disassembling the airborne detection data after the data scan core of the first dimension wanders is y, where y is a positive integer and y>x; Performing cross-matching based on time series features on the data units in the second data scan set to obtain x third data scan sets; Determining the second aviation meteorological focus feature corresponding to the first aviation meteorological embedded state vector in the third data scan set according to the linkage feature focusing strategy; The determining, according to the linkage feature focusing strategy, the second aviation meteorological focus feature corresponding to the first aviation meteorological embedded state vector in the third data scan set includes: Acquire the timing cross-matching information corresponding to the p-th third data scan set, where the timing cross-matching information is used to characterize the timing continuity of the second data scan set after cross-matching, where p is a positive integer, p≤x; Acquire aviation status labeling information corresponding to the p-th third data scan set according to the time series cross-matching information; Determining the embedding vector commonality value between the first aviation meteorological embedded state vectors in the third data scan set according to the linkage feature focusing strategy; Weighting the embedding vector commonality value and the aviation state annotation information to obtain an aviation state weighted commonality value; Based on the aviation state weighted commonality value, feature enhancement is performed on the first aviation meteorological embedded state vector of the data unit in the third data scan set to obtain a second aviation meteorological focus feature.

6. The method according to any one of claims 1 to 3, characterized in that: The step size of the data scanning core of the first dimension is adjusted in the airborne detection data, and the airborne detection data is re-disassembled into a plurality of second data scanning sets, including: Walking the data scanning core of the first dimension at intervals of w data units on a first walking path corresponding to the airborne detection data; Walk the data scanning core of the first dimension at intervals of r data units on the second walking path corresponding to the airborne detection data to determine the plurality of second data scanning sets; Wherein, w and r are positive integers.

7. The method according to any one of claims 1 to 3, characterized in that: The step of predicting the weather pattern in the airborne detection data according to the second aviation meteorological embedded state vector to obtain a weather pattern prediction label corresponding to the airborne detection data includes: Acquire an aviation mission description feature, where the aviation mission description feature is used to assist in predicting a weather pattern corresponding to the airborne detection data; Inputting the aviation mission description feature and the second aviation weather embedded state vector into a pre-debugged aviation weather forecast network to obtain the weather pattern forecast label; The step of inputting the aviation mission description feature and the second aviation meteorological embedded state vector into a pre-debugged aviation meteorological forecast network to obtain the weather pattern forecast label comprises: Inputting the aviation mission description feature and the second aviation meteorological embedded state vector into a pre-debugged aviation meteorological prediction network to obtain meteorological state bias prediction information of the aviation meteorological prediction network; Obtaining a set label index indication, where the label index indication is used to match and output forecast keywords that meet target aviation weather requirements; The meteorological state bias prediction information is updated according to the label index indication to obtain the weather pattern prediction label.

8. The method according to any one of claims 1 to 3, characterized in that: The weather mode prediction label includes the meteorological state jump type, meteorological state distribution characteristics and meteorological state risk characteristics of the weather mode. After predicting the weather mode in the airborne detection data according to the second aviation meteorological embedded state vector and obtaining the weather mode prediction label corresponding to the airborne detection data, the method further includes: Based on the meteorological state distribution characteristics indicated by the weather pattern prediction tag, the airborne detection data is disassembled into local airborne detection results corresponding to a plurality of weather patterns respectively; According to the meteorological state jump type indicated by the weather pattern prediction tag, predict the state jump characteristics in the local airborne detection result, and generate state jump mapping data corresponding to the local airborne detection result; According to the meteorological state risk characteristics indicated by the weather mode forecast label, the state jump mapping data corresponding to the several weather modes are judged to generate an aviation risk judgment viewpoint.

9. The method according to any one of claims 1 to 3, characterized in that: The method is implemented by a pre-debugged weather pattern prediction algorithm, which is debugged by a prediction algorithm to be debugged, and the debugging process of the prediction algorithm to be debugged includes: Acquire an airborne detection data sample and a weather pattern certification sample set corresponding to the airborne detection data sample, wherein the weather pattern certification sample set is used to characterize a weather pattern training annotation chain corresponding to the airborne detection data sample; Inputting the airborne detection data samples into the prediction algorithm to be debugged to generate weather pattern prediction training results; According to the difference between the weather pattern prediction training result and the weather pattern authentication sample set, the prediction algorithm to be debugged is debugged in a loop to obtain the weather pattern prediction algorithm.

10. A big data system, characterized in that: The method comprises at least one processor and a memory; the memory stores computer-executable instructions; the at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the method according to any one of claims 1 to 9.

Images