Intelligent sensing system and method for monitoring landslide disaster risk in real time

Abstract

The invention relates to the technical field of landslide disaster risk assessment and early warning, in particular to an intelligent sensing system and method for landslide disaster risk real-time monitoring. The method comprises the steps of collecting and preliminarily processing multi-dimensional data, further performing standardization processing, and performing fusion processing on the standardized data to obtain fusion features; outputting a predicted landslide risk value based on the fusion features; calculating a prediction error, adjusting the space-time coupling factor according to the prediction error, and optimizing the predicted landslide risk value to obtain a new predicted landslide risk value; based on the new predicted landslide risk value, obtaining a landslide risk value after time-space co-evolution; and further judging whether to trigger an early warning mechanism or not according to a preset threshold value. The problems that the complex nonlinear relationship between the landslide risk and various environmental factors is neglected, the adaptive modeling capability of the deep learning technology on nonlinear features is lacked, and the weights of different data cannot be effectively adjusted according to real-time data are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of landslide disaster risk assessment and early warning, and in particular to an intelligent perception system and method for real-time monitoring of landslide disaster risk. Background Art

[0002] Landslides are a natural disaster that is sudden, destructive, widespread and unpredictable, posing a serious threat to human life and property safety and social and economic development. Especially in mountainous areas, hilly areas and urban-rural fringe areas where landslides are prone to occur, landslide disasters often have a serious impact on people's production and life. Therefore, how to accurately assess landslide risks and provide disaster warnings in a timely manner has become a key issue in reducing losses and improving emergency response efficiency.

[0003] Most of the existing landslide risk assessment methods are based on geological exploration data, historical disaster records and expert experience, but these traditional methods have problems such as low accuracy, poor timeliness and inability to adjust dynamically. With climate change and the intensification of human activities, the frequency and intensity of landslide disasters have increased year by year, making it increasingly difficult for traditional assessment methods to cope with the increasingly complex landslide risk assessment needs. Therefore, there is an urgent need to develop a landslide risk assessment system that can combine multi-source sensor data, intelligent processing technology and prediction algorithms. Summary of the invention

[0004] The present invention provides an intelligent perception system and method for real-time monitoring of landslide disaster risks, so as to solve the problems that the prior art ignores the complex nonlinear relationship between landslide risk and various environmental factors, resulting in the accuracy and practicality of the prediction results failing to meet the needs of real-time monitoring and early warning; lacks the adaptive modeling capability of deep learning technology for nonlinear characteristics, and cannot achieve accurate risk early warning; and cannot effectively adjust the weights of different sensor data according to real-time data, resulting in lag and error accumulation in the assessment of landslide risk.

[0005] The present invention provides an intelligent sensing system and method for real-time monitoring of landslide disaster risks, which specifically includes the following technical solutions:

[0006] An intelligent perception method for real-time monitoring of landslide disaster risk includes the following steps:

[0007] S1. Divide the monitoring area into two or more data blocks, and collect multi-dimensional data related to the landslide in each data block through each sensor; perform denoising, filtering and standardization on the multi-dimensional data collected by each sensor, and fuse the standardized data of each sensor through the spatiotemporal coupling factor to obtain fusion features;

[0008] S2. Perform nonlinear mapping on the fusion features based on the deep learning model and output the predicted landslide risk value; calculate the prediction error based on the predicted landslide risk value, adjust the spatiotemporal coupling factor according to the prediction error, and optimize the predicted landslide risk value to obtain a new predicted landslide risk value; based on the new predicted landslide risk value, combine the spatiotemporal co-evolution model to evaluate the landslide risk and obtain the landslide risk value after spatiotemporal co-evolution; combine the landslide risk value after spatiotemporal co-evolution with the predetermined threshold to determine whether to trigger the early warning mechanism.

[0009] Preferably, the S1 specifically includes:

[0010] The weights are dynamically adjusted according to the spatiotemporal importance of the standardized data of each sensor, and the standardized data of each sensor are weighted based on the weights. The weighted data are aggregated and summed to form fusion features. The mathematical expression of the fusion feature is:

[0011]

[0012] Among them, F(t) is the fusion feature at time t; n is the number of sensors; m is the number of data blocks; is the spatiotemporal coupling factor between the ith sensor and the ith data block; x′ i (t) is the data of the i-th sensor after standardized processing at time t.

[0013] Preferably, the S2 specifically includes:

[0014] In the deep learning model, the fused features are input into each layer of the neural network for processing, and the input is weighted and offset through the inter-layer weights and bias items. Finally, after passing through the nonlinear activation function, the predicted landslide risk value is output.

[0015] Preferably, the S2 specifically includes:

[0016] Based on the deviation between the predicted landslide risk value and the actual landslide risk value, a prediction error is generated; the prediction error is used as a feedback signal to dynamically adjust the spatiotemporal coupling factor.

[0017] Preferably, the S2 specifically includes:

[0018] When the prediction error is greater than the preset threshold, it means that the current risk assessment fails to capture the risk factors of landslide occurrence. The feedback mechanism will adjust the calculation of the spatiotemporal coupling factor. The specific formula is:

[0019]

[0020] in, is the adjusted spatiotemporal coupling factor; is the spatiotemporal coupling factor between the i-th sensor and the j-th data block; β is the adjustment factor, which indicates the sensitivity of the spatiotemporal coupling factor adjustment; E(t) is the prediction error at time t.

[0021] Preferably, the S2 specifically includes:

[0022] The adjusted spatiotemporal coupling factors are used to fuse the comprehensive features to obtain new fused features. The new fused features are used as the input of the deep learning model to obtain a new predicted landslide risk value.

[0023] Preferably, the S2 specifically includes:

[0024] The newly predicted landslide risk value is transferred to the spatiotemporal coevolution model. Through the spatiotemporal coevolution process, the newly predicted landslide risk value is corrected to obtain the landslide risk value after spatiotemporal coevolution. The specific formula is:

[0025]

[0026] Where S(t) is the landslide risk value after spatiotemporal co-evolution; n is the number of sensors; m is the number of data blocks; γi, j (t) represents the space-time synergy factor between the i-th sensor and the j-th data block; R(t) is the predicted landslide risk value at time t; δ i，j is the time variation factor; is the landslide risk change rate.

[0027] An intelligent perception system for real-time monitoring of landslide disaster risks includes the following parts:

[0028] Spatial division and data acquisition module, data preprocessing module, data fusion module, deep learning model module, spatiotemporal coupling factor adjustment module, spatiotemporal co-evolution model module, landslide risk warning module;

[0029] The spatial division and data acquisition module divides the monitoring area into two or more data blocks; each data block collects multi-dimensional data related to the landslide in real time through various sensors, and transmits the collected multi-dimensional data to the data preprocessing module;

[0030] The data preprocessing module performs preliminary processing and standardization on the multi-dimensional data collected by each sensor, and transmits the standardized data of each sensor to the data fusion module;

[0031] The data fusion module fuses the standardized data of each sensor, dynamically adjusts the weight according to the spatiotemporal coupling factor, forms fusion features, and transmits the fusion features to the deep learning model module;

[0032] The deep learning model module performs nonlinear mapping on the fusion features to obtain the predicted landslide risk value, and calculates the prediction error based on the predicted landslide risk value and the actual landslide risk value; the predicted landslide risk value is passed to the spatiotemporal co-evolution model module, and the prediction error is passed to the spatiotemporal coupling factor adjustment module;

[0033] The spatiotemporal coupling factor adjustment module dynamically adjusts the spatiotemporal coupling factor according to the prediction error and transmits the adjusted spatiotemporal coupling factor to the data fusion module;

[0034] The spatiotemporal co-evolution model module, based on the predicted landslide risk value, corrects the predicted landslide risk value through the spatiotemporal co-evolution process, outputs the landslide risk value after the spatiotemporal co-evolution, and transmits the landslide risk value after the spatiotemporal co-evolution to the landslide risk warning module;

[0035] The landslide risk warning module compares the landslide risk value after spatiotemporal co-evolution with the preset threshold to determine whether to trigger the warning mechanism; when the landslide risk value after spatiotemporal co-evolution exceeds the preset threshold, the landslide disaster warning is automatically triggered.

[0036] The beneficial effects of the technical solution of the present invention are:

[0037] 1. By real-time collection, processing and integration of data from multiple sensors, including soil moisture, meteorological data, seismic acceleration, etc., fusion features are generated to comprehensively reflect the environmental status of the landslide area; by combining time-space coupling factors, landslide risks can be assessed more accurately, and the weight of each data can be dynamically adjusted to ensure a more sensitive and accurate response to different environmental factors.

[0038] 2. A deep learning model is used to perform nonlinear mapping on the fusion features, thereby capturing the complex nonlinear relationship between spatiotemporal data. It can not only accurately identify the potential risks of landslides, but also adaptively adjust the weights to improve the accuracy and real-time performance of landslide risk prediction. By dynamically adjusting the spatiotemporal coupling factors, it can flexibly respond to environmental changes. When the prediction error is large, the intelligent perception system will automatically adjust the weight of each sensor data to optimize the prediction of landslide risks. The adjustment of the spatiotemporal coupling factors takes into account the co-evolution effect of space and time, significantly improving the adaptability of the deep learning model to a changing environment.

[0039] 3. Through the spatiotemporal co-evolution process, combined with the space-time synergy factor and the time variation factor, the predicted landslide risk value can be corrected, thereby improving the accuracy and timeliness of the early warning; by setting a threshold, it is determined whether the landslide risk value after the spatiotemporal co-evolution exceeds the threshold. Once the landslide risk value after the spatiotemporal co-evolution exceeds the threshold, the early warning mechanism is immediately triggered to provide accurate early warning information to relevant departments and effectively avoid the losses caused by landslide disasters. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 This is a structural diagram of an intelligent sensing system for real-time monitoring of landslide disaster risks according to the present invention;

[0041] Figure 2 This is a flow chart of an intelligent perception method for real-time monitoring of landslide disaster risks described in the present invention. DETAILED DESCRIPTION

[0042] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the technical scheme in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiment of the present invention, not all 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.

[0043] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0044] The specific scheme of an intelligent sensing system and method for real-time monitoring of landslide disaster risks provided by the present invention is described in detail below with reference to the accompanying drawings.

[0045] Refer to the attached Figure 1 , which shows a structure diagram of an intelligent perception system for real-time monitoring of landslide disaster risks provided by an embodiment of the present invention, the system includes the following parts:

[0046] Spatial division and data acquisition module, data preprocessing module, data fusion module, deep learning model module, spatiotemporal coupling factor adjustment module, spatiotemporal co-evolution model module, landslide risk warning module;

[0047] The spatial division and data acquisition module divides the monitoring area into multiple spatial units, namely data blocks, according to the geographical characteristics of the monitoring area and the distribution of historical landslide events. Each data block collects multi-dimensional data related to the landslide in real time through multiple sensors, such as soil moisture sensors, meteorological sensors, seismic acceleration sensors, etc., and transmits the collected multi-dimensional data to the data preprocessing module.

[0048] The data preprocessing module performs preliminary processing such as denoising and filtering on the multi-dimensional data collected by each sensor and standardization processing to eliminate the dimensional differences between different sensors, and passes the standardized data of each sensor to the data fusion module;

[0049] The data fusion module fuses the standardized data of each sensor, dynamically adjusts the weight according to the spatiotemporal coupling factor, forms fusion features, and transmits the fusion features to the deep learning model module;

[0050] The deep learning model module performs nonlinear mapping on the fusion features and uses a deep learning network, such as a multi-layer neural network, to capture the complex nonlinear relationship between spatiotemporal data, outputs a predicted landslide risk value, and calculates the prediction error based on the predicted landslide risk value and the actual landslide risk value; the predicted landslide risk value is passed to the spatiotemporal co-evolution model module, and the prediction error is passed to the spatiotemporal coupling factor adjustment module;

[0051] The spatiotemporal coupling factor adjustment module dynamically adjusts the spatiotemporal coupling factor according to the prediction error of the deep learning model to ensure that the intelligent perception system can respond to environmental changes and improve prediction accuracy. The adjusted spatiotemporal coupling factor is passed to the data fusion module to optimize the fusion features.

[0052] The spatiotemporal co-evolution model module, based on the predicted landslide risk value, corrects the predicted landslide risk value through the spatiotemporal co-evolution process, outputs the landslide risk value after spatiotemporal co-evolution, captures the spatiotemporal dynamic changes of landslide hazards, and transmits the landslide risk value after spatiotemporal co-evolution to the landslide risk warning module;

[0053] The landslide risk warning module compares the landslide risk value after spatiotemporal co-evolution with the preset threshold to determine whether to trigger the warning mechanism; when the landslide risk value after spatiotemporal co-evolution exceeds the preset threshold, the landslide disaster warning is automatically triggered.

[0054] Refer to the attached Figure 2 , which shows a flow chart of an intelligent perception method for real-time monitoring of landslide disaster risks provided by an embodiment of the present invention, the method comprising the following steps:

[0055] S1. Divide the monitoring area into two or more data blocks, and collect multi-dimensional data related to the landslide in each data block through each sensor; denoise, filter and standardize the multi-dimensional data collected by each sensor, and fuse the standardized data of each sensor through the spatiotemporal coupling factor to obtain the fusion feature.

[0056] In the real-time monitoring system of landslide disaster risk, the Geographic Information System (GIS) is an important source of information. In order to achieve efficient risk assessment, the monitoring area needs to be divided into multiple spatial units, namely data blocks, each of which contains characteristic data of a specific geographical location in the current area; the data blocks are defined based on the geographical characteristics of the monitoring area, such as slope, soil moisture, vegetation coverage, rainfall, etc., as well as the distribution of historical landslide events.

[0057] Real-time acquisition of multi-dimensional data related to landslides in each data block, using a variety of sensor networks, covering meteorological data collection, geological data collection, and structural health monitoring. The sensor network includes: soil moisture sensors, deployed in the soil layer of the landslide area, used to monitor the moisture content of the soil in real time. The output of the soil moisture sensor reflects the saturation of the soil. The abnormal increase in soil moisture is often a precursor to landslide disasters; meteorological sensors, including temperature and humidity sensors, air pressure sensors, and precipitation sensors, are used to obtain meteorological data in the landslide area. The output of the precipitation sensor indicates the precipitation at a certain moment, and the output of the temperature and humidity sensor is used to provide temperature and humidity data; seismic acceleration sensors are used to monitor seismic activity in the landslide area. The output of the seismic acceleration sensor is the seismic acceleration value, which represents the impact of seismic activity on the geological structure of the landslide area.

[0058] The multi-dimensional data collected by the sensor is processed by preliminary denoising and filtering to remove environmental noise and interference to ensure the accuracy and reliability of the data. The data collected by each sensor at different times is standardized through the following process:

[0059]

[0060] Among them, x′ i (t) is the data of the ith sensor after standardized processing at time t; x i (t) is the data collected by the i-th sensor at time t; μ i (t) is the mean value of the data collected by the i-th sensor at time t; σ i (t) is the standard deviation of the data collected by the i-th sensor at time t. The purpose of data normalization is to eliminate the dimensional differences between different sensors and ensure that the fusion is not affected by the scale.

[0061] The standardized data from different sensors are fused into a comprehensive feature. Since the environmental factors measured by each sensor, such as soil moisture, meteorology, earthquake acceleration, etc., have different effects on landslide risk, the intelligent perception system will dynamically adjust the weight of the standardized data of each sensor according to its spatiotemporal importance, thereby ensuring that under different environmental changes, the intelligent perception system can give priority to more influential data. For example, if the input data of a certain sensor contributes more to the prediction of landslide risk, the intelligent perception system will increase its spatiotemporal coupling factor, thereby increasing the fusion weight of the data. All weighted data will be aggregated at time t and summed to form the final fusion feature to more accurately reflect the current environmental conditions and potential risks, and provide more accurate input for subsequent risk assessment. The mathematical expression of the fusion feature is:

[0062]

[0063] Among them, F(t) is the fusion feature at time t; n is the number of sensors; m is the number of data blocks; is the spatiotemporal coupling factor between the ith sensor and the ith data block, indicating the importance of each data point in the comprehensive feature. Through this method, the weight can be adaptively adjusted according to the real-time data, so as to more accurately integrate the multi-dimensional data and ensure the real-time assessment of landslide hazard risk.

[0064] S2. Perform nonlinear mapping on the fusion features based on the deep learning model and output the predicted landslide risk value; calculate the prediction error based on the predicted landslide risk value, adjust the spatiotemporal coupling factor according to the prediction error, and optimize the predicted landslide risk value to obtain a new predicted landslide risk value; based on the new predicted landslide risk value, combine the spatiotemporal co-evolution model to evaluate the landslide risk and obtain the landslide risk value after spatiotemporal co-evolution; combine the landslide risk value after spatiotemporal co-evolution with the predetermined threshold to determine whether to trigger the early warning mechanism.

[0065] Based on the deep learning model, the fused features are nonlinearly mapped, and the complex nonlinear relationship between spatiotemporal data is captured through the processing of multiple neural network layers; in the deep learning model, the fused features are input into each layer of the neural network for processing, and the input is weighted and offset through the inter-layer weights and bias items, and finally the predicted landslide risk value is output after the nonlinear activation function. The formula is as follows:

[0066]

[0067] Where R(t) is the landslide risk value predicted at time t; f k (F(t)) represents the activation function of the kth layer, which represents the nonlinear mapping of the fusion feature F(t); w k is the weight of the kth layer, indicating the contribution of each layer feature; b k is the bias term of the kth layer, indicating the constant adjustment of this layer; σ is the activation function, usually the Sigmoid function is selected for nonlinear mapping; p is the number of layers of the neural network. The parameter w of the neural network k and b k It is trained by the error back propagation algorithm. The deep learning model can better assess landslide risk by capturing the complex nonlinear relationship of spatiotemporal data.

[0068] The prediction error is calculated to measure the deviation between the predicted landslide risk value and the actual landslide risk value, reflecting the gap between the predicted value and the actual observed value. Therefore, the prediction error is used as a feedback signal to adjust the spatiotemporal coupling factor, which is a key parameter calculated based on different sensor inputs and environmental factors during the monitoring process, reflecting the weight of sensor data in the overall risk assessment. Through the analysis of the prediction error, the spatiotemporal coupling factor can be dynamically adjusted to improve the response speed to environmental changes and the prediction accuracy. Specifically, the calculation formula of the prediction error is:

[0069]

[0070] Where E(t) is the prediction error at time t; is the actual landslide risk value.

[0071] When the prediction error is greater than the threshold preset based on the expert experience method, it means that the current risk assessment may not accurately capture the risk factors of landslides; at this time, the feedback mechanism will adjust the calculation of the spatiotemporal coupling factor and increase the weight of specific sensor data to compensate for the deviation in the prediction. The specific formula is:

[0072]

[0073] in, is the adjusted spatiotemporal coupling factor; β is the adjustment factor, which indicates the sensitivity of the spatiotemporal coupling factor adjustment. The larger the prediction error, the greater the adjustment of the spatiotemporal coupling factor, and the more sensitive the judgment of real-time landslide risk is.

[0074] The adjusted spatiotemporal coupling factor reflects the mutual influence and co-evolution relationship between different locations and time points in space and time, and is used to capture the coupling effect of geographic space (such as the mutual influence of geological and climatic factors between different regions) and time factors (such as climate change, seasonal changes, etc.). By adjusting the spatiotemporal coupling factor, the intelligent perception system can dynamically adjust the predicted landslide risk value of each location in the spatiotemporal domain. The adjusted spatiotemporal coupling factor is used to fuse the comprehensive features to obtain a new fusion feature, which contains the global risk information after the spatiotemporal coupling relationship is adjusted. The new fusion feature is used as the input of the deep learning model, and the new predicted landslide risk value is obtained through the calculation of the multi-layer neural network, which provides an accurate judgment basis for the actual landslide risk monitoring and early warning.

[0075] The newly predicted landslide risk value is passed to the spatiotemporal co-evolution model. Through the spatiotemporal co-evolution process and combined with the optimization results of the spatiotemporal coupling factors, the intelligent perception system makes weighted corrections to the landslide risk at each spatiotemporal point to reflect the complex interaction between spatial distribution and temporal dynamics, capture the inherent laws of landslide hazards in the spatiotemporal evolution and the dynamic changes of the external environment, and thus obtain a more accurate landslide risk value. The specific formula is:

[0076]

[0077] Where S(t) is the landslide risk value after spatiotemporal co-evolution; γ i，j (t) represents the spatial-temporal synergy factor between the i-th sensor and the j-th data block, which indicates the spatial and temporal dependency between each data point in the comprehensive feature and the landslide risk; δ i，j is the time variation factor, which indicates the changing speed of landslide risk with time evolution; is the landslide risk change rate.

[0078] Combining the landslide risk value after spatiotemporal co-evolution with the predetermined threshold θ, real-time warning of landslide disasters is achieved. Specifically, when S(t) exceeds the threshold θ, the system will automatically trigger the warning mechanism. The warning trigger formula is:

[0079]

[0080] Among them, Alert(t) is the landslide warning status, 1 means warning triggering, 0 means no warning; the predetermined threshold is determined by expert experience or data-driven methods. Warning triggering ensures that when the landslide risk value after spatiotemporal co-evolution exceeds the predetermined threshold, the intelligent perception system can issue a warning in time and provide accurate information to relevant departments.

[0081] In summary, an intelligent perception system and method for real-time monitoring of landslide disaster risks have been completed.

[0082] The order of the embodiments of the invention is for description only and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0083] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

[0084] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. An intelligent perception method for real-time monitoring of landslide disaster risk, characterized in that: The following steps are involved: S1. Divide the monitoring area into two or more data blocks, and collect multi-dimensional data related to the landslide in each data block through each sensor; perform denoising, filtering and standardization on the multi-dimensional data collected by each sensor, and fuse the standardized data of each sensor through the spatiotemporal coupling factor to obtain fusion features; S2, nonlinearly mapping the fused features based on the deep learning model, and outputting the predicted landslide risk value; The prediction error is calculated based on the predicted landslide risk value, the spatiotemporal coupling factor is adjusted according to the prediction error, and the predicted landslide risk value is optimized to obtain a new predicted landslide risk value; based on the new predicted landslide risk value, the landslide risk is evaluated in combination with the spatiotemporal co-evolution model to obtain the landslide risk value after spatiotemporal co-evolution; The landslide risk value after spatiotemporal co-evolution is combined with the predetermined threshold to determine whether the early warning mechanism should be triggered.

2. The intelligent perception method for real-time monitoring of landslide disaster risk according to claim 1 is characterized in that: The S1 specifically includes: The weights are dynamically adjusted according to the spatiotemporal importance of the standardized data of each sensor, and the standardized data of each sensor are weighted based on the weights. The weighted data are aggregated and summed to form fusion features. The mathematical expression of the fusion feature is: Among them, F(t) is the fusion feature at time t; n is the number of sensors; m is the number of data blocks; is the spatiotemporal coupling factor between the ith sensor and the jth data block; x′ i (t) is the data of the i-th sensor after standardized processing at time t.

3. The intelligent perception method for real-time monitoring of landslide disaster risk according to claim 1 is characterized in that: The S2 specifically includes: In the deep learning model, the fused features are input into each layer of the neural network for processing, and the input is weighted and offset through the inter-layer weights and bias items. Finally, after passing through the nonlinear activation function, the predicted landslide risk value is output.

4. The intelligent perception method for real-time monitoring of landslide disaster risk according to claim 3 is characterized in that: The S2 specifically includes: Based on the deviation between the predicted landslide risk value and the actual landslide risk value, a prediction error is generated; the prediction error is used as a feedback signal to dynamically adjust the spatiotemporal coupling factor.

5. The intelligent perception method for real-time monitoring of landslide disaster risk according to claim 4 is characterized in that: The S2 specifically includes: When the prediction error is greater than the preset threshold, it means that the current risk assessment fails to capture the risk factors of landslide occurrence. The feedback mechanism will adjust the calculation of the spatiotemporal coupling factor. The specific formula is: in, is the adjusted spatiotemporal coupling factor; is the spatiotemporal coupling factor between the i-th sensor and the j-th data block; β is the adjustment factor, which indicates the sensitivity of the spatiotemporal coupling factor adjustment; E(t) is the prediction error at time t.

6. The intelligent perception method for real-time monitoring of landslide disaster risk according to claim 5 is characterized in that: The S2 specifically includes: The adjusted spatiotemporal coupling factors are used to fuse the comprehensive features to obtain new fused features. The new fused features are used as the input of the deep learning model to obtain a new predicted landslide risk value.

7. The intelligent perception method for real-time monitoring of landslide disaster risk according to claim 6 is characterized in that: The S2 specifically includes: The newly predicted landslide risk value is transferred to the spatiotemporal coevolution model. Through the spatiotemporal coevolution process, the newly predicted landslide risk value is corrected to obtain the landslide risk value after spatiotemporal coevolution. The specific formula is: Where S(t) is the landslide risk value after spatiotemporal co-evolution; n is the number of sensors; m is the number of data blocks; γ i，j (t) represents the space-time synergy factor between the i-th sensor and the j-th data block; R(t) is the predicted landslide risk value at time t; δ i，j is the time variation factor; is the landslide risk change rate.

8. An intelligent perception system for real-time monitoring of landslide disaster risk, applied to the intelligent perception method for real-time monitoring of landslide disaster risk as claimed in claim 1, characterized in that: Includes the following parts: Spatial division and data acquisition module, data preprocessing module, data fusion module, deep learning model module, spatiotemporal coupling factor adjustment module, spatiotemporal co-evolution model module, landslide risk warning module; The spatial division and data acquisition module divides the monitoring area into two or more data blocks; each data block collects multi-dimensional data related to the landslide in real time through various sensors, and transmits the collected multi-dimensional data to the data preprocessing module; The data preprocessing module performs preliminary processing and standardization on the multi-dimensional data collected by each sensor, and transmits the standardized data of each sensor to the data fusion module; The data fusion module fuses the standardized data of each sensor, dynamically adjusts the weight according to the spatiotemporal coupling factor, forms fusion features, and transmits the fusion features to the deep learning model module; The deep learning model module performs nonlinear mapping on the fusion features to obtain the predicted landslide risk value, and calculates the prediction error based on the predicted landslide risk value and the actual landslide risk value; the predicted landslide risk value is passed to the spatiotemporal co-evolution model module, and the prediction error is passed to the spatiotemporal coupling factor adjustment module; The spatiotemporal coupling factor adjustment module dynamically adjusts the spatiotemporal coupling factor according to the prediction error and transmits the adjusted spatiotemporal coupling factor to the data fusion module; The spatiotemporal co-evolution model module, based on the predicted landslide risk value, corrects the predicted landslide risk value through the spatiotemporal co-evolution process, outputs the landslide risk value after the spatiotemporal co-evolution, and transmits the landslide risk value after the spatiotemporal co-evolution to the landslide risk warning module; The landslide risk warning module compares the landslide risk value after spatiotemporal co-evolution with the preset threshold to determine whether to trigger the warning mechanism; When the landslide risk value after the space-time co-evolution exceeds the preset threshold, the landslide disaster warning is automatically triggered.

Images