Wireless star chain real-time early warning data processing method and system for landslide

Abstract

The invention relates to the technical field of data processing, in particular to a wireless star chain real-time early warning data processing method and system for landslide. The method comprises the following steps of collecting data influencing landslide occurrence, performing time sequence mode learning by using historical data, selecting key time sequence characteristics from different time windows, finally distributing a weight for each key time sequence characteristic, and according to the real-time data influencing landslide occurrence, determining the time sequence of the landslide. And generating a prediction model in combination with the selected key time sequence features and the weights of the key time sequence features, and combining the probability with the key time sequence features by using a fuzzy inference rule to evaluate the risk level in real time. According to the wireless star chain real-time early warning data processing method and system for landslide, the most valuable features for landslide prediction can be selected more effectively, and the relevance of the features is enhanced by calculating the interaction information among the features, so that the accuracy of a landslide prediction model is improved.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a wireless Starlink real-time warning data processing method and system for landslides. Background Art

[0002] Landslides pose a serious threat to people's lives and property in earthquake-prone areas and areas with concentrated rainfall. Extreme weather events brought about by climate change, such as frequent heavy rainfall and rainstorms, have increased the moisture content of the soil, leading to reduced stability of the mountains. In earthquake-prone areas, the impact of earthquake vibrations on soil pressure is also an important factor in the occurrence of landslides. Earthquakes not only directly trigger landslides, but may also change the existing geological balance at the critical moment of landslides, further increasing the risk of landslides.

[0003] Therefore, timely and accurate landslide warning is crucial to reducing disaster losses and protecting people's lives and property. The occurrence of landslides is usually a complex process, which is affected by multiple factors, including precipitation, soil moisture, earthquake vibrations, etc., and the interaction of these factors may increase the risk of landslides. Most existing early warning systems rely on a single feature or simple feature weighting, which has certain limitations when facing complex landslide processes and cannot fully capture the interactive relationship between different features, thus affecting the accuracy and response speed of the warning. Therefore, a wireless Starlink real-time early warning data processing method and system for landslides are provided. Summary of the invention

[0004] The purpose of the present invention is to provide a wireless Starlink real-time warning data processing method and system for landslides, so as to solve the problem that most of the existing warning systems proposed in the above background technology rely on a single feature or simple feature weighting, which has certain limitations when facing complex landslide processes and cannot fully capture the interactive relationship between different features, thereby affecting the accuracy of the warning and the response speed.

[0005] To achieve the above object, the present invention aims to provide a wireless Starlink real-time warning data processing method for landslides, comprising the following steps: S1. Collect data affecting the occurrence of landslides through sensors deployed on the mountain, and transmit the collected data affecting the occurrence of landslides to a ground base station via radio waves; S2. The ground base station performs real-time preprocessing on the received data affecting the occurrence of landslides, and preliminarily determines whether each data affecting the occurrence of landslides exceeds its own safety threshold. If one data exceeds its own safety threshold, a first warning signal is generated and transmitted to the warning system via wireless starlink. Otherwise, a landslide occurrence prediction is performed. S3. Use the historical data on landslide occurrence stored in the ground base station to learn the time series pattern, select key time series features from different time windows, and introduce the mutual influence of soil moisture and precipitation and the mutual influence of earthquake magnitude and soil pressure in the process of selecting key time series features for optimization, and finally assign weights to each key time series feature; S4. Generate a prediction model based on the real-time data that affects the occurrence of landslides, combined with the selected key time series features and the weights of the key time series features. The prediction model can predict the time window and probability of landslide occurrence, and use fuzzy reasoning rules to combine the probability with the key time series features to evaluate the risk level in real time; S5. Once the probability of a landslide occurring within the predicted landslide occurrence time window exceeds the safety probability value or the risk level exceeds the safety risk level, a second warning signal is generated, and the second warning signal is transmitted to the warning system through multiple channels using wireless Starlink. The warning system issues warnings through multiple channels for the received first warning signal and second warning signal.

[0006] As a further improvement of the present technical solution, in S1, the data affecting the occurrence of landslides include geological data, meteorological data, seismic data and topographic data; the geological data include at least soil moisture and soil pressure; the meteorological data include at least precipitation and temperature; the seismic data include at least earthquake magnitude; and the topographic data include at least slope angle and slope direction.

[0007] As a further improvement of the technical solution, the historical data affecting the occurrence of landslides stored in the ground base station is used in S3 to learn the time series pattern and select key time series features from different time windows, as follows: Set time window and landslide occurrence tags ,in, Indicates that a landslide has occurred. Indicates that the landslide has not occurred. The data in S2 that affect the occurrence of landslides after processing constitutes a feature collection ; Feature Collection Includes soil moisture, soil pressure, precipitation, temperature, earthquake magnitude, slope angle and aspect; Calculate the tags for landslide occurrence With Features The mutual information between: ; in, is mutual information; is the time step Labels where landslides occurred; is the time step Previous Features is the dynamic weighting factor; is the time step Labels on landslide occurrence Information entropy of is the time step Previous Features Information entropy of is the time step The label of the upper landslide occurred with the The joint information entropy of the features; ; is the length of the time window; The calculated mutual information values ​​are sorted, several features with higher mutual information values ​​are selected as key time series features, and the key time series features are used as inputs of the prediction model.

[0008] As a further improvement of the technical solution, in S3, the mutual influence of soil moisture and precipitation and the mutual influence of earthquake magnitude and soil pressure are introduced in the process of selecting key time series features for optimization, and the optimization is as follows: Joint information entropy of soil moisture and precipitation: ; in, Soil moisture and precipitation at time The joint information entropy of For in time The soil moisture characteristic value on ; For in time Characteristic value of precipitation on ; For in time The joint probability distribution of soil moisture and precipitation; Mutual information between soil moisture and precipitation: ; in, is the mutual information between soil moisture and precipitation; is the individual information entropy of soil moisture; is the individual information entropy of precipitation; is the joint information entropy of soil moisture and precipitation; Joint information entropy of earthquake vibration and soil pressure: ; in, is the earthquake magnitude and soil pressure at time The joint information entropy of For in time The characteristic value of earthquake magnitude on ; For in time Characteristic value of soil pressure on ; For in time The joint probability distribution of earthquake magnitude and soil pressure; Mutual information of earthquake vibrations and soil pressure: ; in, is the mutual information between earthquake magnitude and soil pressure; is the individual information entropy of the earthquake magnitude; is the individual information entropy of soil pressure; is the joint information entropy of earthquake magnitude and soil pressure; Key timing feature selection: ; in, is the optimized mutual information; is the characteristic value of soil moisture; is the characteristic value of precipitation; is the characteristic value of earthquake magnitude; is the characteristic value of soil pressure; is the mutual information between the label of landslide occurrence and soil moisture and precipitation; is the mutual information between the label of landslide occurrence and earthquake magnitude and soil pressure; The optimized mutual information value is introduced into the sorting list, several features with top rankings in the optimized sorting list are selected as key time series features, and the several key time series features are used as inputs of the prediction model.

[0009] As a further improvement of the technical solution, in S3, a weight is assigned to each key timing feature, specifically: ; in, For the The weight of the key timing features; Features and Features The interaction between them.

[0010] As a further improvement of the technical solution, in S4, according to the real-time data affecting the occurrence of landslides, combined with the selected key time series features and the weights of the key time series features, a prediction model is generated, and the prediction model can predict the time window and probability of occurrence of landslides. The specific steps are as follows: Prediction of landslide occurrence time window: ; in, is the predicted landslide occurrence time window; is the number of key timing features; ; Label for landslide occurrence With Features In time Mutual information on ; ; in, is the weighted sum of all features; is a nonlinear transformation function; The predicted landslide occurrence time window Probability of occurrence; , is the total number of key time series features in the predicted landslide occurrence time window A collection of .

[0011] As a further improvement of the technical solution, in S4, fuzzy inference rules are used to combine probability with key time series features to evaluate the risk level in real time, as follows: S41. Standardize the key features and predicted probabilities that affect landslides; S42, select key features and predicted probabilities related to landslide occurrence as input variables of fuzzy reasoning and transform them into fuzzy sets; S43, formulate a fuzzy rule base to describe the fuzzy relationship between input variables; S44, by fuzzifying the input data and applying fuzzy rules to perform reasoning, an intermediate result is obtained, and then the result is converted into an output value of landslide risk by defuzzification; S45. Based on the output of fuzzy reasoning, assess the risk level of landslide occurrence.

[0012] As a further improvement of the present technical solution, in S5, the early warning system issues an immediate early warning for the first early warning signal, issues early warnings of varying degrees according to the risk level of the second early warning signal, and generates an emergency strategy.

[0013] As a further improvement of the present technical solution, in S5, the multiple channels include mobile phone APP, text messages, emails and social platforms.

[0014] On the other hand, the present invention provides a wireless Starlink real-time warning data processing system for landslides, including a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the computer program to implement the above-mentioned wireless Starlink real-time warning data processing method for landslides.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. In the wireless Starlink real-time warning data processing method and system for landslides, the interactive effects between soil moisture and precipitation, seismic vibration and soil pressure are introduced to optimize the process of time series feature selection. Through the comprehensive evaluation of the relationship between multiple features, the most valuable features for landslide prediction can be more effectively selected. By calculating the interactive information between these features, the correlation of features is strengthened, thereby improving the accuracy of the landslide prediction model.

[0016] 2. In the wireless Starlink real-time warning data processing method and system for landslides, the response capability of the warning system is effectively improved through a two-level warning mechanism. When the first warning signal exceeds the safety threshold, an alarm can be quickly issued to ensure timely handling of potential landslide risks. The generation of the second warning signal is based on the prediction of the probability and risk level of landslides, providing a hierarchical and dynamically adjusted warning mechanism that can take corresponding emergency response measures at different risk levels, thereby improving the accuracy of the system and the efficiency of emergency response. This hierarchical warning mechanism ensures the comprehensive identification and effective response of landslide risks, minimizing the losses caused by potential disasters. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 The figure is a flow chart of the overall method of the present invention. DETAILED DESCRIPTION

[0018] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] Example 1: Please refer to Figure 1 As shown, this embodiment provides a wireless Starlink real-time warning data processing method for landslides, including the following steps: S1. Collect data affecting the occurrence of landslides through sensors deployed on the mountain, and transmit the collected data affecting the occurrence of landslides to a ground base station via radio waves; In S1, the data that affect the occurrence of landslides include geological data, meteorological data, seismic data and topographic data; geological data at least include soil moisture and soil pressure; meteorological data at least include precipitation and temperature; seismic data at least include earthquake magnitude; topographic data at least include slope angle and slope direction; S2. The ground base station performs real-time preprocessing on the received data that may affect the occurrence of landslides. During the transmission process, noise interference or data missing may occur, and data cleaning is required. Use interpolation algorithms to fill in missing values, and apply smoothing methods to reduce the impact of noise; in order to avoid the impact between different data scales, normalization or standardization processing methods can be used. Common methods include maximum value normalization or Z-score standardization; the time when different sensors collect data may not be completely consistent, and all data need to be synchronized according to a unified timestamp through a time series alignment algorithm. Preliminary judgment is made on whether each data that affects the occurrence of landslides exceeds its own safety threshold. If one data exceeds its own safety threshold, a first warning signal is generated and transmitted to the warning system through the wireless starlink, otherwise a landslide prediction is made; S3. Use the historical data on landslide occurrence stored in the ground base station to learn the time series pattern, select key time series features from different time windows, and introduce the mutual influence of soil moisture and precipitation and the mutual influence of earthquake magnitude and soil pressure in the process of selecting key time series features for optimization, and finally assign weights to each key time series feature; In S3, the historical data affecting the occurrence of landslides stored in the ground base station are used to learn the time series patterns and select key time series features from different time windows, as follows: Set time window and landslide occurrence tags ,in, Indicates that a landslide has occurred. Indicates that the landslide has not occurred. The data in S2 that affect the occurrence of landslides after processing constitutes a feature collection ; Feature Collection Includes soil moisture, soil pressure, precipitation, temperature, earthquake magnitude, slope angle and aspect; Calculate the tags for landslide occurrence With Features The mutual information between: ; in, is mutual information; is the time step Labels where landslides occurred; is the time step Previous Features It is a dynamic weighting factor, reflecting the relative importance of different time step characteristics to landslide prediction. It can be a function that decays over time, such as exponential decay , or weights that are adaptively adjusted according to specific scenarios; is the time step Labels on landslide occurrence Information entropy of is the time step Previous Features Information entropy of is the time step The label of the upper landslide occurred with the The joint information entropy of the features; ; is the length of the time window; The calculated mutual information values ​​are sorted, several features with higher mutual information values ​​are selected as key time series features, and the key time series features are used as inputs of the prediction model.

[0020] Traditional mutual information calculation usually assumes that the relationship between each time series feature and the target label (land slide occurrence) is static, that is, the time series dependency between features is ignored. In order to consider the dynamic changes of time series features and their mutual influence, we introduce a dynamic mutual information adjustment mechanism based on time series. By introducing a time weighting factor, the influence of features in different time steps on the target event is dynamically adjusted to enhance the accuracy of time series feature selection.

[0021] By introducing dynamic weighting factors, the formula can capture the dynamic relationship between features and landslide occurrence at different time steps. For example, when a landslide is about to occur, recent meteorological, earthquake and other features may have a more important impact on the landslide, while long-term data may contribute less. Therefore, by adjusting the weighting factor , we can make the model pay more attention to those features that are closer in time to the landslide, thereby improving the accuracy of the prediction.

[0022] ; ; ; in, for The marginal distribution of for The marginal distribution of For X and The joint distribution of In S3, the mutual influence of soil moisture and precipitation as well as the mutual influence of earthquake magnitude and soil pressure are introduced in the process of selecting key time series features for optimization. The specific results after optimization are as follows: Joint information entropy of soil moisture and precipitation: ; in, Soil moisture and precipitation at time The joint information entropy at , reflects the joint uncertainty of these two features; For in time The soil moisture characteristic value on ; For in time Characteristic value of precipitation on ; For in time The joint probability distribution of soil moisture and precipitation; This formula represents the mutual information between soil moisture and precipitation, reflecting the mutual dependence between the two features. The larger the mutual information, the stronger the relationship between the two features, and the smaller the mutual information, the weaker the relationship between them.

[0023] Mutual information between soil moisture and precipitation: ; in, is the mutual information between soil moisture and precipitation, reflecting the mutual dependence between soil moisture and precipitation. The larger the value, the stronger the relationship between them, and the smaller the value, the weaker the relationship between them. is the individual information entropy of soil moisture, describing the randomness or uncertainty of the soil moisture characteristics themselves; is the individual information entropy of precipitation, describing the randomness or uncertainty of the precipitation characteristics themselves; is the joint information entropy of soil moisture and precipitation, describing the randomness or uncertainty of the combination of these two features; Joint information entropy of earthquake vibration and soil pressure: ; in, is the earthquake magnitude and soil pressure at time The joint information entropy of For in time The characteristic value of earthquake magnitude on ; For in time Characteristic value of soil pressure on ; For in time The joint probability distribution of earthquake magnitude and soil pressure; This formula represents the mutual information between earthquake magnitude and soil pressure, reflecting the mutual dependence between these two features. The larger the value of mutual information, the stronger the relationship between earthquake magnitude and soil pressure, and the higher the predictive value.

[0024] Mutual information of earthquake vibrations and soil pressure: ; in, is the mutual information between earthquake magnitude and soil pressure; is the individual information entropy of the earthquake magnitude, describing the randomness or uncertainty of the earthquake magnitude characteristics themselves; is the individual information entropy of soil pressure, describing the randomness or uncertainty of the soil pressure characteristics themselves; is the joint information entropy of earthquake magnitude and soil pressure, describing the randomness or uncertainty of the combination of these two features; Key timing feature selection: ; in, is the optimized mutual information; is the characteristic value of soil moisture; is the characteristic value of precipitation; is the characteristic value of earthquake magnitude; is the characteristic value of soil pressure; is the mutual information between the label of landslide occurrence and soil moisture and precipitation; is the mutual information between the label of landslide occurrence and earthquake magnitude and soil pressure; This formula represents the optimized mutual information, which is used to select key time series features. The optimized mutual information takes into account the mutual information relationship between landslide occurrence and each key feature, and combines the mutual information between soil moisture and precipitation, and earthquake magnitude and soil pressure. Through the comprehensive evaluation of the relationship between multiple features, the most valuable features for landslide prediction can be selected more effectively, thereby improving the accuracy of the prediction model. Finally, the optimized mutual information value will be used for feature sorting, and the most important features will be selected as the input of the prediction model to improve the prediction performance of the model.

[0025] The optimized mutual information value is introduced into the sorting list, and several features ranked high in the optimized sorting list are selected as key time series features, and these key time series features are used as the input of the prediction model. The role of these formulas is to evaluate the importance of each feature and its combination in landslide prediction by calculating the joint information entropy and mutual information between different features. Through this optimization process, the features that have the greatest impact on the prediction results can be selected, and the accuracy of landslide probability and time window prediction can be further improved.

[0026] In S3, weights are assigned to each key timing feature, specifically: ; in, For the The weight of the key timing features; Features and Features The interaction between features is quantified by calculating the mutual information between features, which reflects the interaction between different features in time. For example, the interaction between soil moisture and precipitation, or the interaction between earthquake shaking and soil pressure. Features with strong interactions may be more important for predicting landslide occurrence; This formula enables us to calculate the weight of each feature based on its independent relationship with landslide occurrence and the interaction between features. This ensures that in the subsequent prediction model, each key time series feature is weighted according to its actual impact on landslide prediction. Specifically, features with greater mutual information and stronger interactions will be given higher weights, ensuring that these features occupy a more important position in the prediction process, thereby improving the accuracy and robustness of the prediction model.

[0027] S4. Generate a prediction model based on the real-time data that affects the occurrence of landslides, combined with the selected key time series features and the weights of the key time series features. The prediction model can predict the time window and probability of landslide occurrence, and use fuzzy reasoning rules to combine the probability with the key time series features to evaluate the risk level in real time; In S4, a prediction model is generated based on the real-time data that affects the occurrence of landslides, combined with the selected key time series features and the weights of the key time series features. The prediction model can predict the time window and probability of landslide occurrence. The specific steps are as follows: Prediction of landslide occurrence time window: ; in, is the predicted time window for landslide occurrence; is the number of key timing features; ; Label for landslide occurrence With Features In time The mutual information on the landslide also includes the interaction between soil moisture and precipitation and the interaction between seismic vibration and soil pressure on the occurrence of landslides. By weighting the mutual information of these features, the time point that best reflects the occurrence of landslides can be selected. , that is, selecting the predicted landslide occurrence time window. This helps to accurately locate the possible time of landslide occurrence under dynamic environmental conditions.

[0028] Select by maximizing weighted mutual information , i.e. select a time point , making the feature the strongest predictor of landslide occurrence.

[0029] ; ; in, It is the weighted sum of all characteristics, indicating the risk of landslide occurrence; is a nonlinear transformation function (such as the output of a deep neural network) used to capture complex relationships between features; The predicted landslide occurrence time window Probability of occurrence; , is the total number of key time series features in the predicted landslide occurrence time window A collection of; Through this three-step formula, the model can effectively predict the time window for landslides and calculate the probability of landslides. The first step is to select the time point that can best predict the occurrence of landslides through weighted mutual information, the second step is to calculate the landslide risk through nonlinear transformation of weighted features, and the last step is to calculate the probability of landslides through logistic regression. These steps form a complete prediction process for real-time assessment of landslide risks and providing accurate warning information.

[0030] In S4, fuzzy inference rules are used to combine probability with key time series features to assess risk levels in real time, as follows: Since landslides are usually affected by multiple uncertain factors, fuzzy logic models can handle these uncertainties. Fuzzy logic models fuzzify different risk factors (such as rainfall, soil moisture, etc.) and convert them into the "probability" value of landslides. Fuzzy set theory is used to process input data (such as precipitation, soil moisture, etc.) and fuzzy reasoning is used to obtain the risk level of landslides.

[0031] S41. Standardize the key features and predicted probabilities that affect landslides so that different types of data can be scaled to facilitate the subsequent fuzzy reasoning process; S42, select key features related to landslide occurrence (such as soil moisture, precipitation, etc.) and predicted probabilities as input variables for fuzzy reasoning and convert them into fuzzy sets, such as "low", "medium", and "high"; S43. Based on the experience of domain experts or data analysis, a fuzzy rule base is developed to describe the fuzzy relationship between input variables, such as the impact of soil moisture and precipitation on the probability of landslide occurrence; common fuzzy rules are as follows: Rule 1: If soil moisture is high and precipitation is high, the probability of landslides is high; Rule 2: If the predicted probability is medium and the earthquake vibration is strong, the probability of landslide occurrence is high; Rule 3: If the predicted probability is low, and both soil moisture and precipitation are low, the probability of landslide occurrence is low; Rule 4: If all characteristics are low, the probability of landslide occurrence is low.

[0032] S44, by fuzzifying the input data and applying fuzzy rules to perform reasoning, an intermediate result is obtained, and then the result is converted into an output value of landslide risk by defuzzification; S45. Evaluate the risk level of landslide based on the output of fuzzy reasoning; S5. Once the probability of a landslide occurring within the predicted landslide occurrence time window exceeds the safety probability value or the risk level exceeds the safety risk level, a second warning signal is generated, and the second warning signal is transmitted to the warning system through multiple channels using wireless Starlink. The warning system issues warnings through multiple channels for the received first warning signal and second warning signal; the multiple channels include mobile phone APP, text messages, emails and social platforms.

[0033] In S5, the early warning system issues an immediate early warning for the first early warning signal, issues early warnings of varying degrees according to the risk level of the second early warning signal, and generates an emergency strategy.

[0034] The setting of step S2 effectively improves the response capability of the early warning system through a two-level early warning mechanism. When the first early warning signal exceeds the safety threshold, an alarm can be quickly issued to ensure timely handling of potential landslide risks. The generation of the second early warning signal is based on the prediction of the probability and risk level of landslides, providing a hierarchical and dynamically adjusted early warning mechanism that can take corresponding emergency response measures at different risk levels, improving the accuracy of the system and the efficiency of emergency response. This hierarchical early warning mechanism ensures the comprehensive identification and effective response of landslide risks, minimizing the losses caused by potential disasters.

[0035] Embodiment 2: This embodiment provides a wireless Starlink real-time warning data processing system for landslides, including a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the computer program to implement any one of the above-mentioned wireless Starlink real-time warning data processing methods for landslides.

[0036] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments, and the above embodiments and descriptions are only preferred examples of the present invention, and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, and these changes and improvements all fall within the scope of the present invention to be protected.

Claims

1. A wireless Starlink real-time warning data processing method for landslides, characterized in that: The following steps are involved: S1. Collect data affecting the occurrence of landslides through sensors deployed on the mountain, and transmit the collected data affecting the occurrence of landslides to a ground base station via radio waves; S2. The ground base station performs real-time preprocessing on the received data affecting the occurrence of landslides, and preliminarily determines whether each data affecting the occurrence of landslides exceeds its own safety threshold. If one data exceeds its own safety threshold, a first warning signal is generated and transmitted to the warning system via wireless starlink. Otherwise, a landslide occurrence prediction is performed. S3. Use the historical data on landslide occurrence stored in the ground base station to learn the time series pattern, select key time series features from different time windows, and introduce the mutual influence of soil moisture and precipitation and the mutual influence of earthquake magnitude and soil pressure in the process of selecting key time series features for optimization, and finally assign weights to each key time series feature; S4. Generate a prediction model based on the real-time data that affects the occurrence of landslides, combined with the selected key time series features and the weights of the key time series features. The prediction model can predict the time window and probability of landslide occurrence, and use fuzzy reasoning rules to combine the probability with the key time series features to evaluate the risk level in real time; S5. Once the probability of a landslide occurring within the predicted landslide occurrence time window exceeds the safety probability value or the risk level exceeds the safety risk level, a second warning signal is generated, and the second warning signal is transmitted to the warning system through multiple channels using wireless Starlink. The warning system issues warnings through multiple channels for the received first warning signal and second warning signal.

2. The wireless Starlink real-time warning data processing method for landslides according to claim 1 is characterized in that: In S1, the data that affect the occurrence of landslides include geological data, meteorological data, seismic data and topographic data; the geological data at least include soil moisture and soil pressure; the meteorological data at least include precipitation and temperature; the seismic data at least include earthquake magnitude; and the topographic data at least include slope angle and slope direction.

3. The wireless Starlink real-time warning data processing method for landslides according to claim 2 is characterized in that: In S3, the historical data affecting the occurrence of landslides stored in the ground base station are used to learn the time series pattern and select key time series features from different time windows, as follows: Set time window and landslide occurrence tags ,in, Indicates that a landslide has occurred. Indicates that the landslide has not occurred. The data in S2 that affect the occurrence of landslides after processing constitutes a feature collection ; Feature Collection Includes soil moisture, soil pressure, precipitation, temperature, earthquake magnitude, slope angle and aspect; Calculate the tags for landslide occurrence With Features The mutual information between: ； in, is mutual information; is the time step Labels where landslides occurred; is the time step Previous Features is the dynamic weighting factor; is the time step Labels on landslide occurrence Information entropy of is the time step Previous Features Information entropy of is the time step The label of the upper landslide occurred with the The joint information entropy of the features; ; is the length of the time window; The calculated mutual information values ​​are sorted, several features with higher mutual information values ​​are selected as key time series features, and the key time series features are used as inputs of the prediction model.

4. The wireless Starlink real-time warning data processing method for landslides according to claim 3 is characterized in that: In S3, the mutual influence of soil moisture and precipitation and the mutual influence of earthquake magnitude and soil pressure are introduced in the process of selecting key time series features for optimization, and the optimization is as follows: Joint information entropy of soil moisture and precipitation: ； in, Soil moisture and precipitation at time The joint information entropy of For in time The soil moisture characteristic value on ; For in time Characteristic value of precipitation on ; For in time The joint probability distribution of soil moisture and precipitation; Mutual information between soil moisture and precipitation: ； in, is the mutual information between soil moisture and precipitation; is the individual information entropy of soil moisture; is the individual information entropy of precipitation; is the joint information entropy of soil moisture and precipitation; Joint information entropy of earthquake vibration and soil pressure: ； in, is the earthquake magnitude and soil pressure at time The joint information entropy of For in time The characteristic value of earthquake magnitude on ; For in time Characteristic value of soil pressure on ; For in time The joint probability distribution of earthquake magnitude and soil pressure; Mutual information of earthquake vibrations and soil pressure: ； in, is the mutual information between earthquake magnitude and soil pressure; is the individual information entropy of the earthquake magnitude; is the individual information entropy of soil pressure; is the joint information entropy of earthquake magnitude and soil pressure; Key timing feature selection: ； in, is the optimized mutual information; is the characteristic value of soil moisture; is the characteristic value of precipitation; is the characteristic value of earthquake magnitude; is the characteristic value of soil pressure; is the mutual information between the label of landslide occurrence and soil moisture and precipitation; is the mutual information between the label of landslide occurrence and earthquake magnitude and soil pressure; The optimized mutual information value is introduced into the sorting list, several features with top rankings in the optimized sorting list are selected as key time series features, and the several key time series features are used as inputs of the prediction model.

5. The wireless Starlink real-time warning data processing method for landslides according to claim 4 is characterized in that: In S3, a weight is assigned to each key timing feature, specifically: ； in, For the The weight of the key timing features; Features and Features The interaction between them.

6. The wireless Starlink real-time warning data processing method for landslides according to claim 5 is characterized in that: In S4, a prediction model is generated based on the real-time data that affects the occurrence of landslides, combined with the selected key time series features and the weights of the key time series features. The prediction model can predict the time window and probability of occurrence of landslides. The specific steps are as follows: Prediction of landslide occurrence time window: ； in, is the predicted landslide occurrence time window; is the number of key timing features; ; Label for landslide occurrence With Features In time Mutual information on ； ； in, is the weighted sum of all features; is a nonlinear transformation function; The predicted landslide occurrence time window Probability of occurrence; , is the total number of key time series features in the predicted landslide occurrence time window A collection of .

7. The wireless Starlink real-time warning data processing method for landslides according to claim 6 is characterized by: In S4, fuzzy inference rules are used to combine probability with key time series features to evaluate the risk level in real time, as follows: S41. Standardize the key features and predicted probabilities that affect landslides; S42, select key features and predicted probabilities related to landslide occurrence as input variables of fuzzy reasoning and transform them into fuzzy sets; S43, formulate a fuzzy rule base to describe the fuzzy relationship between input variables; S44, by fuzzifying the input data and applying fuzzy rules to perform reasoning, an intermediate result is obtained, and then the result is converted into an output value of landslide risk by defuzzification; S45. Based on the output of fuzzy reasoning, assess the risk level of landslide occurrence.

8. The wireless Starlink real-time warning data processing method for landslides according to claim 7 is characterized in that: In S5, the early warning system issues an immediate early warning for the first early warning signal, issues early warnings of varying degrees according to the risk level of the second early warning signal, and generates an emergency strategy.

9. The wireless Starlink real-time warning data processing method for landslides according to claim 8 is characterized in that: In the S5, the multiple channels include mobile APP, SMS, email and social platform.

10. A wireless Starlink real-time warning data processing system for landslides, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: The processor executes a computer program to implement the steps of the wireless Starlink real-time warning data processing method for landslides as described in any one of claims 1-9.

Images