Geological disaster monitoring and early warning method and system based on multi-source sensor data fusion

Abstract

The invention provides a geological disaster monitoring and early warning method and system based on multi-source sensor data fusion, and relates to the technical field of intelligent sensing and information processing.The method comprises the steps that at least three geodetic survey control piles are arranged in a landslide mass monitoring area, and a geodetic survey control network with space coordinates is formed; based on space coordinates provided by a geodetic measurement control network, carrying out space calibration and networking deployment on all displacement, environmental parameter and structural stability monitoring sensors so as to construct a sensor network with a unified space reference; and on the basis of the sensor network with the unified space reference, synchronous triggering and data acquisition of each sensor are executed to obtain a multi-source monitoring data sequence. Based on multi-source sensor data fusion, a closed-loop geological disaster monitoring and early warning method from unified reference data acquisition, space-time registration and intelligent fusion and risk assessment to collaborative response is constructed.

Description

Technical Field

[0001] This invention relates to the field of intelligent sensing and information processing technology, and in particular to a geological disaster monitoring and early warning method and system based on multi-source sensor data fusion. Background Technology

[0002] Landslides and other geological disasters seriously threaten the safety of mountainous areas. Although existing monitoring systems have incorporated various types of sensors, there are systemic deficiencies in data processing, which may lead to low accuracy and poor timeliness in early warning systems.

[0003] Taking landslide monitoring along a highway in Southwest China as an example, various sensors such as GNSS, inclinometers, and rain gauges are deployed. In practice, the system sometimes experiences false alarms and missed alarms. False alarms include situations where heavy rainfall triggers an alarm but the slope remains stable, while missed alarms include situations where deep creep does not reach the threshold but a landslide occurs. This exposes a core weakness in data processing: First, the spatiotemporal references are inconsistent, making data fusion difficult. GNSS, inclinometer, and environmental sensor data exist in different coordinate systems and timestamps, lacking a unified spatiotemporal reference. The system cannot accurately perform spatiotemporal correlation and fusion analysis of heavy rainfall at a certain point, the increase in pore water pressure at that point, and the acceleration of deep displacement at the same point, which may lead to data silos. Secondly, the fusion strategy is static and cannot adapt dynamically. Early warning models often use fixed weights to fuse multi-source data. However, the dominant signals are different at different stages of disaster evolution. In the early stage, hydrological parameters are emphasized, while in the imminent landslide stage, displacement is emphasized. Static weights cannot be adjusted adaptively, which may lead to insensitivity to weak early signals (missed reports) and overreaction to non-disaster interference (false reports). Thirdly, the processing chain is lengthy, and the early warning delay is significant. Data needs to go through multiple links such as collection, remote transmission, and processing by the central server, and the cumulative delay often reaches several minutes. For landslides where the deformation acceleration to instability may only take tens of minutes, this delay is enough to miss the golden window for emergency response, resulting in insufficient real-time performance. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a geological disaster monitoring and early warning method and system based on multi-source sensor data fusion. Based on multi-source sensor data fusion, a closed-loop geological disaster monitoring and early warning method is constructed, from unified benchmark data acquisition, spatiotemporal registration and intelligent fusion, risk assessment to collaborative response.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows: Firstly, a geological disaster monitoring and early warning method based on multi-source sensor data fusion, the method comprising: No fewer than three geodetic control stakes are set up within the landslide monitoring area to form a geodetic control network with spatial coordinates. Based on the spatial coordinates provided by the geodetic control network, all displacement, environmental parameter, and structural stability monitoring sensors are spatially calibrated and networked to construct a sensor network with a unified spatial reference. On the basis of the sensor network with a unified spatial reference, synchronous triggering and data acquisition of each sensor are performed to obtain multi-source monitoring data sequences. Using the spatial coordinate constraints provided by the geodetic control network, spatiotemporal registration is performed on multi-source monitoring data sequences to generate heterogeneous monitoring datasets with unified spatiotemporal references; The heterogeneous monitoring dataset with unified spatiotemporal reference is input into the preset confidence synthesis and fuzzy inference fusion model, and the fusion calculation is performed according to the preset fusion weight strategy corresponding to different disaster evolution stages to obtain the comprehensive evaluation index of landslide stability. The comprehensive stability assessment index of the landslide body is input into a logistic regression probability model trained on historical disasters for calculation to obtain the real-time occurrence probability value; the real-time occurrence probability value is compared with the preset graded dynamic probability threshold to make a decision and obtain an early warning instruction. Based on the early warning command, the system synchronously drives the designated remote and local terminals through a preset hybrid communication channel, and links with relevant emergency systems to execute a closed-loop early warning response.

[0006] Secondly, a geological disaster monitoring and early warning system based on multi-source sensor data fusion includes: The data acquisition module is used to deploy no fewer than three geodetic control stakes within the landslide monitoring area to form a geodetic control network with spatial coordinates. Based on the spatial coordinates provided by the geodetic control network, all displacement, environmental parameter, and structural stability monitoring sensors are spatially calibrated and networked to construct a sensor network with a unified spatial reference. On the basis of the sensor network with a unified spatial reference, the synchronous triggering and data acquisition of each sensor are performed to obtain a multi-source monitoring data sequence. The spatiotemporal registration module is used to perform spatiotemporal registration operations on multi-source monitoring data sequences using the spatial coordinate constraints provided by the geodetic control network, and generate heterogeneous monitoring datasets with unified spatiotemporal references. The fusion module is used to input heterogeneous monitoring datasets with unified spatiotemporal references into a preset confidence synthesis and fuzzy inference fusion model, and perform fusion calculations according to preset fusion weight strategies corresponding to different disaster evolution stages to obtain a comprehensive evaluation index of landslide stability. The decision-making module is used to input the comprehensive stability assessment index of the landslide body into a logistic regression probability model trained on historical disasters for calculation to obtain the real-time occurrence probability value; the real-time occurrence probability value is compared with the preset graded dynamic probability threshold to make a decision and obtain an early warning instruction; The execution module is used to synchronously drive designated remote and local terminals through a preset hybrid communication channel based on the early warning command, and to link with relevant emergency systems to execute a closed-loop early warning response.

[0007] Thirdly, a computing device includes: One or more processors; A storage device for storing one or more programs that, when executed by one or more processors, cause the one or more processors to implement the method.

[0008] Fourthly, a computer-readable storage medium storing a program that, when executed by a processor, implements the method.

[0009] The above-described solution of the present invention has at least the following beneficial effects: Construct a geodetic control network and complete sensor spatial calibration and networking to establish a unified spatial benchmark; synchronously trigger the acquisition of multi-source monitoring data sequences to ensure the spatiotemporal consistency and homogeneity of the initial acquisition of multi-source data; improve the standardization and compatibility of sensor network data to provide a regular basic data support for subsequent data processing; perform spatiotemporal registration based on the spatial coordinate constraints of the geodetic control network to eliminate spatiotemporal deviations of multi-source monitoring data; integrate and form a heterogeneous monitoring dataset with a unified spatiotemporal benchmark to strengthen the correlation and fusion of multi-source data, laying a precise data foundation for subsequent fusion calculations; adopt a confidence synthesis and fuzzy inference fusion model, combined with a dynamic weighting strategy for disaster stages, to carry out fusion calculations; achieve the adaptation of fusion weights to disaster evolution laws, integrate multi-dimensional heterogeneous data information; and output landslide stability data. A qualitative comprehensive assessment index simplifies the information processing dimensions for subsequent decision-making and enhances the systematicness and adaptability of multi-source data fusion; a stability index is calculated using a logistic regression probability model trained on historical disaster data, achieving standardized quantitative calculation of the probability of disaster occurrence; risk comparison and decision-making are completed by combining hierarchical dynamic probability thresholds, achieving refined risk classification; early warning instructions are generated, completing the transformation of data processing results into practical decision-making information and improving the standardization of risk assessment; early warning instructions are synchronously distributed using a pre-set hybrid communication channel, ensuring the stability and full coverage of information transmission; closed-loop response is executed in conjunction with multiple emergency systems, realizing the transformation of early warning information into multi-system collaborative handling actions; a complete closed loop from data processing to emergency execution is formed, improving the practical empowerment value of data processing results and the collaborative efficiency of emergency response. Attached Figure Description

[0010] Figure 1 This is a flowchart illustrating a geological disaster monitoring and early warning method based on multi-source sensor data fusion, provided by an embodiment of the present invention.

[0011] Figure 2 This is a schematic diagram of a geological disaster monitoring and early warning system based on multi-source sensor data fusion, provided by an embodiment of the present invention. Detailed Implementation

[0012] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0013] like Figure 1 As shown in the figure, an embodiment of the present invention proposes a geological disaster monitoring and early warning method based on multi-source sensor data fusion, the method comprising the following steps: Step 100: Deploy no fewer than three geodetic control stakes within the landslide monitoring area to form a geodetic control network with spatial coordinates; Based on the spatial coordinates provided by the geodetic control network, spatially calibrate and network all displacement, environmental parameter, and structural stability monitoring sensors to construct a sensor network with a unified spatial reference; On the basis of the sensor network with a unified spatial reference, perform synchronous triggering and data acquisition of each sensor to obtain a multi-source monitoring data sequence; Step 200: Using the spatial coordinate constraints provided by the geodetic control network, perform spatiotemporal registration on the multi-source monitoring data sequence to generate a heterogeneous monitoring dataset with unified spatiotemporal reference. Step 300: Input the heterogeneous monitoring dataset with unified spatiotemporal reference into the preset confidence synthesis and fuzzy inference fusion model, and perform fusion calculation according to the preset fusion weight strategy corresponding to different disaster evolution stages to obtain the comprehensive evaluation index of landslide stability. Step 400: Input the comprehensive evaluation index of landslide stability into the logistic regression probability model trained on historical disasters for calculation to obtain the real-time occurrence probability value; compare the real-time occurrence probability value with the preset graded dynamic probability threshold to make a decision and obtain an early warning instruction. Step 500: Based on the early warning command, the designated remote and local terminals are synchronously driven through a preset hybrid communication channel, and relevant emergency systems are linked to execute a closed-loop early warning response.

[0014] In this embodiment of the invention, a control network is formed by deploying geodetic control stakes to provide a unified spatial reference for the sensors. Spatial calibration and network deployment ensure the consistency of spatial coordinates of the sensor network, and synchronous triggering of data acquisition ensures the temporal synchronization of multi-source monitoring data sequences, providing a standardized and collaborative data source for subsequent data processing. Spatiotemporal registration is performed using the spatial coordinate constraints of the geodetic control network to achieve a unified spatiotemporal reference for multi-source monitoring data sequences, eliminating spatiotemporal deviations of heterogeneous data, enabling the fusion of heterogeneous monitoring data, and improving the effectiveness of subsequent data processing. A confidence synthesis and fuzzy inference fusion model is adopted, combined with the corresponding stages of different disaster evolution. The system employs a weighted fusion strategy to perform fusion calculations, adapting to the dynamic characteristics of disaster evolution. This makes the fusion process of heterogeneous data more aligned with actual monitoring needs, enhancing the ability of the fusion results to characterize the stability state of landslides. Real-time occurrence probability values ​​are calculated using a logistic regression probability model trained on historical disaster data. This, combined with hierarchical dynamic probability thresholds, enables the generation of early warning instructions based on quantitative probability analysis, improving the targeting and rationality of early warning decisions. Furthermore, the system synchronously drives remote and local terminals via a hybrid communication channel, ensuring the timeliness and comprehensiveness of early warning instruction transmission. This links the emergency response system to execute a closed-loop early warning response, achieving efficient integration of early warning information transmission and emergency response.

[0015] In a preferred embodiment of the present invention, in step 100, at least three geodetic control stakes are deployed within the landslide monitoring area to form a geodetic control network with spatial coordinates; based on the spatial coordinates provided by the geodetic control network, all displacement, environmental parameter, and structural stability monitoring sensors are spatially calibrated and networked to construct a sensor network with a unified spatial reference; based on the sensor network with the unified spatial reference, synchronous triggering and data acquisition of each sensor are performed to obtain a multi-source monitoring data sequence, including: Step 101: Within the landslide monitoring area, select no fewer than three geologically stable points and deploy geodetic control stakes at each point. Specifically, this includes: First, conducting geological surveys within the landslide monitoring area. Based on the regional geological structure distribution characteristics, select no fewer than three geologically stable points unaffected by landslide deformation. Specifically, these points must simultaneously meet the following detailed conditions: First, the lithological conditions: drilling and sampling must confirm that the point is located in a slightly weathered or unweathered hard rock layer area, with a rock integrity coefficient of not less than 0.8, and no fault fracture zones, weak interlayers, or other adverse geological structures passing through it, ensuring that the rock mass where the point is located will not deform due to its own geological defects. Second, the spatial distribution conditions: the points must be evenly distributed in the stable outer area of ​​the landslide monitoring area, and the interior angles of the triangular network units formed by three or more points must be between 30 and 150 degrees, ensuring that the coverage area of ​​the formed control network can... The monitoring points must be designed to completely cover the entire landslide monitoring area, ensuring full coverage for subsequent sensor spatial calibration. Thirdly, the surrounding environment must be carefully considered; the monitoring points must be far from the landslide's rear fracturing zone, front shear zone, and water catchment area, with a horizontal distance of at least 50 meters from the potential sliding boundary to avoid the effects of slope deformation, rainwater erosion, and soaking. Fourthly, the foundation bearing capacity must be carefully assessed. On-site lightweight dynamic testing will be used to calculate the foundation bearing capacity, ensuring it is no less than 150 kPa. Simultaneously, borehole sampling will be used to measure the thickness of the loose overburden layer and calculate the control pile installation depth, ensuring the depth penetrates the loose overburden to stable bedrock, with an embedment depth of at least 1.5 meters. Subsequently, geodetic control piles will be installed at each selected location. The long-term stability of the control piles will be ensured by calculating the burial depth and surface fixing method, providing a solid foundation for subsequent coordinate measurement and control network construction.

[0016] Step 102 involves using satellite positioning technology to synchronously measure all established geodetic control stakes and obtain raw observation data. Specifically, after the geodetic control stakes are deployed, satellite positioning technology is used to synchronously measure all established geodetic control stakes to obtain accurate positional correlation data. Specifically, a satellite positioning receiver is installed at the top of each control stake, with a synchronous observation period of no less than 45 minutes and a sampling interval of 15 seconds. This ensures that all receivers start observation within the same time period, and that receiver signals remain stable during the observation process to avoid obstruction. After the observation is completed, preliminary calculations are performed on the collected satellite signal data. The preliminary relative distance between receivers is calculated using pseudorange observations, and abnormal data exceeding three times the mean square error are removed to reduce time errors caused by observations at different times. This yields raw observation data that reflects the relative positional relationship of each control stake, providing basic observation data for the subsequent determination of the absolute coordinates of the control stakes.

[0017] Step 103 involves performing network adjustment calculations on the original observation data to determine the three-dimensional absolute coordinates of each control point in the national coordinate system. Specifically, this includes: to eliminate observation errors and systematic biases in the original observation data and improve the accuracy of the control point coordinates, network adjustment calculations are performed on the original observation data. In practice, the original observation data is first preprocessed, the standard error of each observation is calculated, and outlier data exceeding three times the standard error are removed. Then, based on the layout of the geodetic control network, a three-dimensional unconstrained adjustment model is selected, and an error equation is constructed. The specific content of this error equation is as follows: the three-dimensional coordinate correction of each control point is used as the core unknown parameter, and the original observation value and the approximate coordinates based on the control points are used as the basis for the calculation. The difference between the calculated theoretical observations is used as the observation residual. The equation as a whole characterizes the functional relationship between the observation residual and the coordinate correction, observation random error, and systematic error. It can quantitatively reflect the corrective effect of the coordinate correction on the observation residual. Its specific form is adapted to the spatial geometric characteristics of the three-dimensional unconstrained adjustment model and is constructed according to the type of observation in the control network. For side length observations, the error equation focuses on the spatial distance dimension, reflecting the linear relationship between the side length observation residual and the three-dimensional coordinate correction of the control stakes at both ends. For angle observations, the error equation focuses on the spatial angle dimension, reflecting the linear relationship between the angle observation residual and the three-dimensional coordinate correction of the three control stakes constituting the angle.

[0018] The calculation process of the error equation is as follows: First, based on the topology of the geodetic control network, the number of control points participating in the adjustment calculation and the approximate three-dimensional coordinates of each control point are determined. Based on the approximate coordinates and spatial geometric formulas, the theoretical values ​​of each observation are calculated. Then, the observation residuals are obtained by subtracting the corresponding theoretical values ​​from the original observation values. Subsequently, the partial derivatives of the coordinate corrections with respect to various observation values ​​are derived to form the coefficient matrix of the error equation. The observation residuals, coefficient matrix, and coordinate corrections are substituted into the adapted basic form of the error equation to complete the construction of the error equation for the entire network. Then, the preprocessed original observation data is substituted into the equation for iterative calculation to gradually eliminate the random and systematic errors generated during the observation process. The unit weight error is checked in real time during the calculation process. When the unit weight error is stable within the allowable range of the standard, the iteration stops. Finally, the three-dimensional absolute coordinates of each control point in the national coordinate system are determined, and the accuracy index of each coordinate component is calculated to ensure that the coordinate accuracy meets the requirements for subsequent control network use, providing coordinate data support for the formation of the subsequent control network.

[0019] Step 104: Based on the three-dimensional absolute coordinates of all control stakes, a geodetic control network with unified known spatial coordinates is formed. This specifically includes: establishing a unified coordinate framework based on the three-dimensional absolute coordinates of all control stakes; calculating the spatial distances and angles between control stakes; verifying the geometric closure error of the control network; and judging the accuracy of the control network shape through the closure error calculation. If the closure error exceeds the allowable range, the coordinates of some control stakes are readjusted until the closure error meets the specification requirements, so that each control stake forms an interconnected spatial network structure under this coordinate framework. This control network can provide a unified spatial reference for the subsequent deployment and spatial calibration of sensors.

[0020] Step 105: Based on the coordinate framework of the geodetic control network, displacement, environmental parameter, and structural stability monitoring sensors are deployed at key geological feature points of the landslide body to obtain the deployed sensor locations. Specifically, this includes: to ensure the targetedness and effectiveness of sensor deployment, based on the coordinate framework of the geodetic control network and combined with the geological survey results of the landslide body, the slope, slope height, and dip angle of the potential sliding surface of the landslide body are calculated to estimate the burial depth and spatial coordinates of the sliding surface, and to determine the coordinate range of key geological feature points such as the sliding surface, weak interlayer, rear edge tensile fracture zone, and front edge shear zone of the landslide body; the spacing between each key point is calculated to ensure that the sensor deployment density can cover the core monitoring area; subsequently, displacement monitoring sensors, environmental parameter monitoring sensors, and structural stability monitoring sensors are deployed at these key geological feature points respectively; by deploying sensors at key points, accurate monitoring of the core deformation area and disaster-induced key areas of the landslide body can be achieved. At the same time, with the help of the coordinate framework of the control network, the sensor deployment has spatial reference, laying the foundation for the subsequent spatial calibration of the sensors.

[0021] Step 106: Based on the deployed sensor locations, use a total station to measure the installation point of each sensor and obtain its relative position observation value relative to the nearest geodetic control stake. Specifically, after completing the sensor deployment, to obtain accurate spatial position information for each sensor installation point, use a total station to accurately measure each sensor installation point based on the deployed sensor locations. During the measurement process, use the control stakes in the geodetic control network as the measurement benchmark, select the geodetic control stake closest to each sensor installation point as the measurement reference point, set up and orient the total station, and observe the horizontal angle, vertical angle, and slope distance of the sensor installation point respectively. Each observation value is observed at least 3 times, the average value of each round of observations is calculated, and observation data with a difference exceeding twice the number of rounds are discarded. Then, based on the horizontal angle, vertical angle, and slope distance, calculate the horizontal distance, elevation difference, and relative coordinate increment of the sensor installation point relative to the corresponding reference control stake to obtain the relative position observation value. This relative position observation value can accurately reflect the spatial positional relationship between the sensor installation point and the control stake, providing a direct basis for subsequently determining the absolute spatial coordinates of the sensor.

[0022] Step 107 involves performing spatial reintersection calculations between the relative position observations and the known absolute coordinates of the corresponding control stakes to determine the absolute spatial coordinates of each sensor. Specifically, this includes: to achieve spatial connection between the sensors and the geodetic control network, performing spatial reintersection calculations between the relative position observations and the known three-dimensional absolute coordinates of the corresponding reference control stakes; specifically, using the three-dimensional absolute coordinates of the control stakes as a known reference, and combining the horizontal angle, vertical angle, and slant distance observation data of the sensor installation point relative to the control stake, constructing the basic equations for spatial reintersection; the specific content of these basic equations is as follows: using the three-dimensional absolute coordinates of the reference control stakes as known quantities and the three-dimensional absolute coordinates of the sensor installation points as unknown quantities, establishing spatial geometric function relationships between the observed data such as horizontal angle, vertical angle, and slant distance and the known and unknown quantities. The core is to establish spatial geometric function relationships between the known reference point coordinates and the observed data. By correlating the measured data and inferring the coordinates of the unknown sensor installation points, the equations as a whole quantitatively characterize the spatial positional constraints among the observed data, known coordinates, and unknown coordinates. Their specific forms are adapted to the computational needs of three-dimensional resection. Corresponding equations are constructed according to the data type of the observed data. For slant distance observation data, the equations focus on the distance dimension between two points in space, reflecting the geometric correlation between the slant distance observation value and the three-dimensional coordinate difference between the sensor installation point and the control stake. For horizontal angle observation data, the equations focus on the angle dimension within the horizontal plane, reflecting the angular correlation between the horizontal angle observation value and the sensor installation point and the control stake in the planar coordinate components. For vertical angle observation data, the equations focus on the angle dimension within the vertical plane, reflecting the angular correlation between the vertical angle observation value, the difference in elevation coordinate components between the sensor installation point and the control stake, and the horizontal distance between the sensor installation point and the control stake.

[0023] The calculation process of the equation is as follows: First, the known three-dimensional absolute coordinates of the reference control stake and the relative position observation data of the sensor installation point are determined, and the approximate three-dimensional coordinates of the sensor installation point are initially set. Based on spatial geometric relationships, the calculation formulas for the theoretical observation values ​​corresponding to the slope distance, horizontal angle, and vertical angle are derived respectively. The approximate coordinates and the known control stake coordinates are substituted into the calculation formulas to obtain the theoretical values ​​of various observation data. Then, the difference between the observation value and the theoretical value is used as the residual to construct an error equation containing the correction of the unknown coordinates. The error equation is solved by the least squares method to obtain the correction of the approximate coordinates of the sensor installation point. The correction is superimposed on the approximate coordinates to obtain the preliminary absolute spatial coordinates of the sensor installation point. Then, the above calculation process is repeated, and the preliminary coordinates are iteratively corrected by calculating the residuals of the observation values ​​until the residuals meet the accuracy requirements. Finally, the absolute spatial coordinates of each sensor are calculated, and the coordinate accuracy index is calculated to ensure that the coordinate accuracy meets the monitoring requirements. This ensures that each sensor has a precise spatial identification under a unified coordinate system, providing data support for the subsequent unification of the spatial reference of the sensor network.

[0024] Step 108: Based on the absolute spatial coordinates of all sensors, the spatial reference of the sensor network is unified, resulting in a sensor network with a unified spatial reference. This specifically includes: first, calculating the deviation between the absolute spatial coordinates of each sensor and the coordinate frame of the geodetic control network, including the deviations of the X, Y, and Z coordinate components; then, calculating correction coefficients based on the deviation values ​​to correct the absolute spatial coordinates of each sensor; after correction, calculating the deviation values ​​again to verify the correction effect, until the deviations of all sensor coordinates from the control network coordinate frame are less than the allowable value, ensuring that all sensors in the network follow the same spatial reference; this eliminates spatial coordinate deviations caused by differences in the deployment locations of different sensors, achieving a unified spatial reference for the sensor network. This unified sensor network ensures that data collected by different types of sensors are spatially comparable.

[0025] Step 109: Based on the sensor network with a unified spatial reference, send a synchronous acquisition command to all sensors in the network. Specifically, this includes: based on the sensor network with a unified spatial reference, the network control center sends a synchronous acquisition command to all sensors in the network; first, calculate the signal transmission distance between the network control center and each sensor, calculate the command transmission delay time according to the electromagnetic wave propagation speed, and then compensate for the transmission time of the synchronous trigger signal based on the delay time; so that the unified time trigger signal carried by the synchronous acquisition command can reach each sensor simultaneously, ensuring that all types of sensors in the network can receive a consistent acquisition start command, providing command support for subsequent time-synchronous acquisition of multi-source data.

[0026] Step 110: In response to the synchronous acquisition command, each sensor is triggered to perform synchronous data acquisition operations, and the raw electrical and digital signals generated by each sensor after performing the synchronous data acquisition operations are received. Specifically, this includes: upon receiving the synchronous acquisition command, each sensor immediately responds to the command and performs synchronous data acquisition operations; different types of sensors, based on their own acquisition characteristics, simultaneously start observing the monitored object and generate corresponding raw electrical or digital signals; during the acquisition process, the sampling time interval of the sensors is calculated to ensure that the sampling frequency meets the preset requirements, and then each sensor transmits the acquired raw electrical and digital signals to the data processing terminal in real time; during the transmission process, the data transmission check value is calculated to verify the data integrity; through synchronous acquisition operations, raw data of different monitoring parameters at the same time dimension can be obtained, avoiding data time misalignment caused by differences in acquisition time between different sensors, and at the same time, complete reception of raw signals can provide a comprehensive data source for subsequent data processing.

[0027] Step 111 involves amplifying, filtering, and performing analog-to-digital conversion on the original electrical signal and digital signal to obtain an initial digital sampling sequence. Specifically, this includes: first, distinguishing the types of the original electrical signal and digital signal, and performing appropriate preprocessing operations accordingly; for the original electrical signal, analyzing its amplitude range, calculating the required amplification factor, and amplifying the weak original electrical signal to enhance the effective signal strength; then, determining the noise frequency range in the original signal through spectrum analysis, calculating the appropriate filter cutoff frequency, and using low-pass or band-pass filtering to remove environmental noise from the signal, such as electromagnetic interference and vibration interference; finally, calculating the appropriate sampling frequency to ensure that the sampling frequency is not lower than [the specified value]. The highest frequency of the signal is twice that of the Nyquist sampling theorem. An analog-to-digital conversion (ADC) is then performed to convert it into a standard digital signal. For the original digital signal, its data format and sampling frequency are first analyzed and uniformly converted to a preset standard data format. Then, a digital filtering algorithm is used to remove high-frequency noise and abnormal fluctuations, completing the noise reduction and normalization of the digital signal. Finally, the processed analog-to-digital signal is integrated with the normalized original digital signal to form an initial digital sampling sequence. Simultaneously, the signal-to-noise ratio (SNR) of the initial digital sampling sequence is calculated to verify the preprocessing effect. This preprocessing process effectively improves data reliability and lays the foundation for subsequent data standardization conversion.

[0028] Step 112 involves calling the calibration coefficients of each sensor to convert the initial digital sampling sequence into physical quantity values ​​with standard units. Specifically, this includes: to unify the dimensions of output data from different types of sensors and ensure the data has a unified physical meaning, calling the calibration coefficients set at the factory for each sensor. These calibration coefficients match the sensor's measurement principle and output characteristics. First, the sensor's calibration certificate is consulted to obtain the corresponding calibration coefficients and zero-point offset values. Then, the initial digital sampling sequence is substituted into a preset conversion formula for correlation calculation. That is, the initial digital sampling value is multiplied by the calibration coefficient, and then the zero-point offset value is added to obtain the physical quantity value with standard physical units. For example, the digital sampling value of a displacement sensor is multiplied by a millimeter-level calibration coefficient and added to the zero-point offset value to convert it into millimeter-level displacement; the digital sampling value of a rain gauge sensor is multiplied by a millimeter-level calibration coefficient and added to the zero-point offset value to convert it into millimeter-level rainfall. After the calculation is completed, the error range of the converted physical quantity value is calculated to verify the conversion accuracy. This achieves dimension unification of data from different sensors, facilitating subsequent cross-sensor data comparison and fusion.

[0029] Step 113 involves binding the physical quantity value corresponding to each sensor to its absolute spatial coordinates to obtain physical quantity data with three-dimensional spatial coordinates. Specifically, this includes: first, establishing a data index table, calculating the association identifier between the acquisition timestamp of each physical quantity data and the corresponding sensor absolute coordinates, and storing the physical quantity values ​​and absolute coordinates of the same sensor at the same acquisition time in a one-to-one correspondence; simultaneously calculating the uniqueness check value of the association identifier to avoid association errors; ensuring that each set of monitoring data carries the corresponding three-dimensional spatial location information. Through this binding operation, the source location of each monitoring data can be located, providing support for subsequent correlation analysis of different monitoring parameters at the same location.

[0030] Step 114: Based on the physical quantity data with three-dimensional spatial coordinates, perform timestamp synchronization and alignment operations on it using a unified time base, and use interpolation methods to fill data gaps caused by different sensor sampling rates to obtain data that has completed time synchronization and gap filling. Specifically, this includes: performing timestamp synchronization and alignment operations on all data based on the physical quantity data with three-dimensional spatial coordinates using a unified time base; specifically, setting Coordinated Universal Time (UTC) as the unified time standard, calculating the difference between the original timestamp of each sensor data and UTC, and converting different timestamps into time values ​​under the unified time base; for data gaps caused by differences in the sampling rates of different sensors, first locate the spatial position of the sensor corresponding to the gap data, then search for other valid sensor nodes within a preset range of that position, calculate the spatial straight-line distance between the gap position and each valid sensor node, and assign corresponding weight coefficients according to the spatial distance, with the weight coefficient being larger for closer distances. The greater the distance, the smaller the weighting coefficient. Simultaneously, the changing trends of adjacent valid data within the time period of the missing data are analyzed, and the rate of change of adjacent data is calculated. Combining the spatial distance weighting coefficient and the rate of change over time, an appropriate linear or nonlinear interpolation method is selected to weight the physical quantity values ​​at the missing time. During the calculation process, the temporal change rates of each valid sensor node are fused according to their corresponding weighting coefficients to generate interpolation results that balance spatial correlation and temporal continuity, filling the data gaps. After filling, the deviation between the filled data and adjacent valid data is calculated, and the spatial consistency between the filled data and the data from surrounding sensors at the same time is verified to ensure that the filled data conforms to the spatiotemporal changing trends of the monitoring parameters. Finally, data that has completed time synchronization and gap filling is obtained. This data possesses a unified time benchmark and a continuous time series, while also incorporating spatial correlation information, providing a more accurate temporal guarantee for subsequent multi-source data integration and fusion analysis.

[0031] Step 115: Based on the data that has been synchronized and filled in time, the physical quantity data of displacement, environment, stress, and tilt angle categories are integrated in chronological order into landslide surface and deep displacement sequences, rainfall and soil water content sequences, and soil internal stress and slope tilt angle change sequences, which together constitute a multi-source monitoring data sequence. Specifically, based on the data that has been synchronized and filled in time, classification labels for displacement, environmental parameters, and structural stability data are first defined. The parameter characteristic values ​​of each data set are calculated, and the data category is determined based on the characteristic values. Then, the data of each category are sorted in chronological order, and the time series length of each category is calculated. Consistent sampling intervals were used to regularize the time series. Displacement-related data were integrated chronologically into landslide surface and deep displacement sequences; environmental parameter-related data were integrated chronologically into rainfall and soil moisture state sequences; and structural stability-related data were integrated chronologically into internal stress and slope inclination change sequences. After integration, data integrity indices for each type of sequence were calculated to verify the sequence's effectiveness. These three types of sequences together constitute a multi-source monitoring data sequence. This structured multi-source monitoring data sequence can clearly present the temporal evolution patterns of different types of monitoring parameters, providing regular input data for subsequent accurate multi-source data fusion analysis.

[0032] In this embodiment of the invention, control stakes are deployed at geologically stable locations to ensure their stability and provide reliable foundational support for the subsequent construction of the control network. Satellite positioning technology is used to synchronously measure the control stakes, reducing time-varying errors during observation and ensuring the synchronization and consistency of the original observation data. Network adjustment calculations are performed on the original observation data to eliminate observation errors and system biases, improving the accuracy and reliability of the three-dimensional absolute coordinates of the control stakes. A unified spatial coordinate control network is formed based on the absolute coordinates, establishing a unified coordinate framework to provide a benchmark for subsequent sensor spatial calibration. Sensors are deployed at key geological feature points according to the control network coordinate framework, ensuring sensor coverage of the core monitoring area of ​​the landslide and improving the relevance and effectiveness of the monitoring data. A total station is used to measure the sensor observations relative to the control stakes, obtaining accurate relative position data through high-precision measurement methods, providing reliable input for sensor absolute coordinate calculation. Spatial resection is used to calculate the sensor absolute spatial coordinates, efficiently linking the relative observations with the known coordinates of the control stakes, achieving precise spatial connection between the sensors and the control network. The unified spatial benchmark of the sensor network is completed, eliminating spatial coordinate deviations between different sensors and enabling spatial comparability and fusion of multi-source sensor data.

[0033] Sending synchronous acquisition commands ensures that all sensors within the network start acquiring data synchronously, guaranteeing the temporal coordination of multi-source data from the source; triggering synchronous acquisition and receiving raw signals to obtain raw data from multiple sensors in parallel acquisition, providing a complete raw data source for subsequent data standardization processing; amplifying, filtering, and performing analog-to-digital conversion on the raw signals to remove signal noise interference, enhance effective signals, and convert analog signals into processable digital signals, improving data quality; calling calibration coefficients to convert to standard unit physical quantity values, unifying the dimensions of output data from different sensors, giving the data a unified physical meaning, and facilitating cross-sensor comparison; binding physical quantity values ​​to absolute spatial coordinates, ensuring that each set of monitoring data carries precise spatial information, determining the monitoring location corresponding to the data, and providing support for spatial dimension data analysis; performing timestamp synchronization alignment and interpolation to fill gaps, eliminating time deviations caused by sampling rate differences, filling data gaps, and ensuring the temporal continuity and uniformity of multi-source data; integrating into multi-source monitoring data sequences by category, making the scattered multi-source data structured and organized, clearly distinguishing different types of monitoring data, and providing a regular data input for subsequent targeted fusion calculations.

[0034] In a preferred embodiment of the present invention, step 200 above, utilizing the spatial coordinate constraints provided by the geodetic control network, performs a spatiotemporal registration operation on the multi-source monitoring data sequence to generate a heterogeneous monitoring dataset with unified spatiotemporal reference, including: Step 201 involves extracting the landslide surface and deep displacement sequences, rainfall and soil moisture state sequences, and soil internal stress and slope inclination change sequences from the multi-source monitoring data sequences. Specifically, this includes: firstly, extracting the landslide surface and deep displacement sequences, rainfall and soil moisture state sequences, and soil internal stress and slope inclination change sequences from the multi-source monitoring data sequences; more specifically, based on the monitoring parameter category identifiers corresponding to each data point, filtering out all data entries under each category by traversing the multi-source monitoring data sequences, and simultaneously performing integrity checks on each extracted sequence to remove invalid data entries, ensuring that each sequence can fully reflect the change process of the corresponding monitoring parameters, thus laying the data foundation for subsequent registration processing in the time and spatial dimensions.

[0035] Step 202 involves assigning a timestamp to each data point in the landslide surface and deep displacement sequence, the rainfall and soil moisture state sequence, and the soil internal stress and slope inclination change sequence, based on the recorded time information and a unified clock reference. Specifically, to eliminate the problem of inconsistent time representation standards between different monitoring sequences, and to connect the landslide surface and deep displacement sequence, the rainfall and soil moisture state sequence, and the soil internal stress and slope inclination change sequence, for each data point in each type of sequence, first, the original recorded time information is retrieved. Then, using the unified clock reference (such as Coordinated Universal Time) set in step 109 as a reference, the time difference between the original time information and the unified clock reference is calculated. The original time information is calibrated based on this difference, and then a unique and standardized timestamp is assigned to each data point. During this process, it is necessary to ensure that the timestamp of the same data point is consistent with the time reference synchronized in step 114, so that all data points in the three types of sequences have a unified time representation standard, providing a unified time reference for subsequent time series mapping and alignment.

[0036] Step 203 involves mapping all data points to a unified time axis based on the timestamps, identifying data points with inconsistent sampling times. Specifically, this includes: first, establishing a continuous unified time axis based on a unified clock reference, the time accuracy of which must match the highest sampling frequency of each sensor; then, mapping all timestamped data points from the three sequences to their corresponding positions on the unified time axis according to their timestamp values; and finally, calculating the time interval between adjacent data points on the time axis to set a time consistency judgment threshold. When the time interval between a data point and its adjacent data points exceeds this threshold, it is determined to be a data point with inconsistent sampling times. This process accurately identifies time-misaligned data points, determining the core objects for subsequent targeted interpolation processing.

[0037] Step 204: For data points with inconsistent sampling times, interpolation calculations are performed within the interval formed by adjacent data points to generate interpolated data at the synchronized time. Specifically, this includes: First, based on the data points with inconsistent sampling times, the validity of the adjacent valid data points before and after the data point is verified to ensure that there are no abnormal deviations between adjacent data points; then, according to the changing trend of adjacent valid data points, an appropriate interpolation method is selected. If the data shows a linear changing trend, linear interpolation is used; if it shows a non-linear changing trend, non-linear interpolation is used; based on the selected interpolation method, interpolation calculations are performed within the time interval formed by adjacent valid data points to generate interpolated data at the synchronized time with a unified time axis. At the same time, the deviation between the interpolated data and adjacent valid data points is calculated to ensure that the deviation is within the allowable range, so that the interpolated data can conform to the changing pattern of the actual monitoring parameters.

[0038] Step 205 involves merging all mapped data points with the interpolated data at the synchronization time to obtain a multi-source data sequence with complete temporal alignment. Specifically, this includes: to achieve temporal unification of the multi-source data, connecting the mapped valid data points with the interpolated data, and merging and sorting the two types of data in ascending order of timestamps; during the merging process, data points with duplicate timestamps need to be removed, and data with higher precision needs to be retained; after merging, the temporal continuity of the obtained multi-source data sequence is checked, and the time interval between adjacent data points is calculated to ensure that all data points are evenly distributed on a unified time axis without any time gaps, ultimately obtaining a multi-source data sequence with complete temporal alignment, effectively eliminating the temporal misalignment problem of multi-source data, and creating a temporal coordination foundation for subsequent spatial registration processing.

[0039] Step 206: Using the spatial coordinate constraints of the geodetic control network, constrained adjustment and optimization calculations are performed on the sensor spatial coordinates associated with the time-aligned multi-source data sequences to obtain the adjusted sensor spatial coordinates. Specifically, this includes: First, extracting the original spatial coordinates of each sensor associated with the time-aligned multi-source data sequences; then, using the spatial coordinates provided by the geodetic control network formed in step 104 as the constraint benchmark, calculating the deviation values ​​between the original spatial coordinates of each sensor and the control network coordinate frame, including the deviations of the X, Y, and Z coordinate components; constructing a constrained adjustment model based on these deviation values, specifically using a least-squares spatial adjustment geometric optimization algorithm for calculation; first, determining the weight matrix of the observation values ​​based on the measurement accuracy index of each sensor, with sensors having higher measurement accuracy corresponding to larger weights, thereby quantifying the different sensor coordinates. The reliability of the observed values ​​is assessed. Then, the error relationship between the original spatial coordinates of the sensors and the control network coordinate framework is constructed, incorporating the coordinate deviation values ​​into this relationship as the observation residuals. Based on this, the least squares principle is used to solve for the correction amount of each sensor's spatial coordinates. This correction amount is then superimposed on the original spatial coordinates to obtain the preliminary adjusted sensor spatial coordinates. Next, the sum of squared residuals between the preliminary adjusted coordinates and the control network baseline is calculated, and it is verified whether this sum of squared residuals is less than a preset accuracy threshold. If it is not met, the weight matrix parameters and adjustment model constraints are adjusted, and iterative calculations are performed again. After each iteration, the sum of squared residuals and the deviations of each coordinate component are calculated synchronously until both the sum of squared residuals and the deviations of each coordinate component meet the preset accuracy requirements. Finally, the adjusted sensor spatial coordinates are obtained, improving the accuracy and consistency of the sensor spatial coordinates.

[0040] Step 207 involves re-associating the adjusted sensor spatial coordinates with the multi-source data sequence to generate a spatiotemporal registration fusion dataset. Specifically, this includes: establishing a one-to-one correspondence between the sensor spatial coordinates and the data sequence based on the adjusted sensor spatial coordinates and the time-aligned multi-source data sequence; specifically, using the sensor's unique identifier as a link, binding the adjusted spatial coordinates with all monitoring data points collected by the corresponding sensor, and establishing a coordinate-data association index table; after the association is completed, verifying the association to ensure that each data point corresponds to a unique sensor spatial coordinate, with no association errors or omissions, thereby generating a spatiotemporal registration fusion dataset and achieving preliminary coordination between the time and spatial dimensions of the monitoring data.

[0041] Step 208 involves constructing a regular three-dimensional spatial grid covering the monitoring area based on the spatiotemporal registration and fusion dataset. Specifically, this includes: First, based on the spatiotemporal registration and fusion dataset and the coordinate framework of the geodetic control network, determining the spatial extent of the monitoring area and calculating the maximum and minimum values ​​of the monitoring area in the X, Y, and Z coordinate directions; then, according to the monitoring accuracy requirements, calculating the node spacing of the regular three-dimensional spatial grid to ensure that the grid node density reflects the spatial variation characteristics of the monitoring parameters; and finally, based on the determined spatial extent and node spacing, constructing a regular three-dimensional spatial grid covering the entire monitoring area, ensuring that the grid's coordinate system is consistent with the geodetic control network, and providing a standardized spatial carrier for subsequent spatial resampling of multi-source data.

[0042] Step 209 involves resampling the physical quantity data in the spatiotemporal registration fusion dataset to each node of the three-dimensional spatial grid to obtain gridded multi-physical quantity field data. Specifically, this includes: analyzing the spatial distribution characteristics of each physical quantity data based on the regular three-dimensional spatial grid and the spatiotemporal registration fusion dataset, and selecting an appropriate spatial resampling method; for data with relatively uniform spatial distribution, using nearest neighbor interpolation, and for data with relatively gentle spatial changes, using linear interpolation; based on the selected method, resampling each physical quantity data in the spatiotemporal registration fusion dataset to each node of the three-dimensional spatial grid; after resampling, calculating the deviation between the data at each node and the original monitoring data, verifying the rationality of the resampling results, and finally obtaining gridded multi-physical quantity field data, thus achieving the integration of multi-source data in the spatial dimension.

[0043] Step 210: Based on the gridded multi-physical field data, organize and aggregate it according to time and space dimensions into a sequence of multi-physical spatial field snapshots arranged in time order, and finally encapsulate it into a heterogeneous monitoring dataset with a unified spatiotemporal reference. Specifically, this includes: First, based on the gridded multi-physical field data, setting a time aggregation interval, which is determined according to the rate of change of monitoring parameters and the timeliness requirements of early warning; then, dividing the gridded data according to the time dimension, and aggregating the physical quantity data of all grid nodes within the same time interval into a multi-physical spatial field snapshot; sorting all snapshots in time order to form a multi-physical spatial field snapshot sequence; finally, standardizing and encapsulating the sequence, determining key information such as data format, coordinate system, and time reference, and finally generating a heterogeneous monitoring dataset with a unified spatiotemporal reference. This dataset can clearly present the distribution and change patterns of each monitoring parameter at different times and spatial locations, providing regular and unified input data for subsequent confidence synthesis and fuzzy inference fusion calculation.

[0044] In this embodiment of the invention, three types of data sequences are extracted according to the monitoring parameter categories to achieve structured decomposition of multi-source monitoring data. This enables different types of monitoring data to form independent and well-organized processing units, laying the foundation for subsequent targeted operations such as time alignment and spatial registration, and improving the systematicity and efficiency of data processing. Timestamps are assigned to each data point based on a unified clock reference, unifying the time representation standard of different sequence data. This eliminates the differences in the original time records of different sensors, making various types of data comparable in the time dimension, and providing a unified time reference for subsequent time alignment. Data points are mapped to a unified time axis, and data points with inconsistent sampling times are identified to locate multi-source data. Addressing the issue of time-series misalignment; identifying the core processing objects for time-series alignment to avoid invalid data processing operations and improve the targeting and accuracy of time-series synchronization; generating synchronized time data by interpolation within adjacent data point intervals to fill time gaps caused by inconsistent sampling times; ensuring the continuity and integrity of time series so that different types of data can form effective correspondences at the same time node, creating conditions for the time-series fusion of multi-source data; merging mapped data and interpolated data to obtain a time-series aligned sequence, completing the unification of multi-source data in the time dimension; eliminating the interference of time misalignment on data fusion, enabling various monitoring data to be collaboratively correlated along the time dimension, and improving the time coordination of multi-source data.

[0045] This method utilizes the coordinate constraints of the geodetic control network to adjust and optimize the spatial coordinates of sensors, correcting sensor coordinate deviations with the high-precision spatial benchmark of the control network. This improves the accuracy and consistency of sensor spatial coordinates and strengthens the spatial benchmark uniformity of multi-source data. The adjusted coordinates are re-associated with multi-source data sequences to ensure the matching of monitoring data with precise spatial locations. Spatial correlation errors caused by coordinate deviations are eliminated, providing reliable spatial location support for subsequent spatial data analysis and fusion. A regular three-dimensional spatial grid covering the monitoring area is constructed, establishing a unified spatial data carrying framework. This allows dispersed monitoring data to be incorporated into a consistent spatial reference system, providing a standardized spatial carrier for the spatial integration and global analysis of multi-source data. Physical quantity data is resampled to three-dimensional grid nodes, realizing the transformation of discrete monitoring data into continuous spatial data. The spatial distribution characteristics of physical quantities within the monitoring area are clearly presented, facilitating the analysis of the spatial correlation of various physical quantities from a global perspective. Multi-physical quantity spatial field snapshot sequences are organized and aggregated and packaged into heterogeneous monitoring datasets, completing the structured integration of multi-source data in the spatiotemporal dimensions. This ensures that the data simultaneously possesses a unified time and spatial benchmark, forming standardized input data that can be directly used for subsequent fusion calculations, improving data usability and the efficiency of subsequent processing.

[0046] In a preferred embodiment of the present invention, step 300 involves inputting a heterogeneous monitoring dataset with unified spatiotemporal references into a preset confidence synthesis and fuzzy inference fusion model, and performing fusion calculations according to a preset fusion weight strategy corresponding to different disaster evolution stages to obtain a comprehensive landslide stability assessment index, including: Step 301: Extract displacement, environmental, stress, and tilt data of spatial grid nodes at each time point from the heterogeneous monitoring dataset with unified spatiotemporal reference as raw features. Specifically, this includes: First, establishing feature filtering rules based on data category identifiers. By traversing the heterogeneous monitoring dataset row by row, comparing the category identifier of each data point with the preset feature category list, filtering out all surface and deep deformation data related to displacement, rainfall and soil moisture state data related to the environment, and internal stress and slope tilt data related to structural stability. Then, classify and organize the data according to the time dimension and spatial grid nodes, first extracting each data point... Based on the timestamp and grid node coordinate index, the time interval corresponding to the timestamp is calculated. Time segments are divided according to equal time intervals, and data corresponding to the same grid node coordinate index within the same time segment are grouped together to ensure that each group of data corresponds to a clear time and spatial location. At the same time, the validity of the extracted data is verified by calculating the mean and standard deviation of each group of data and setting a data fluctuation threshold. When the deviation of a data point from the mean of its group exceeds 3 times the standard deviation, it is judged as abnormal data and removed. Finally, a well-organized multi-dimensional original feature set is formed, providing accurate input data for subsequent uncertainty quantification and fusion calculation.

[0047] Step 302: Based on the original features, calculate the support degree of each feature for each disaster evolution stage according to the preset fuzzy membership function, and convert it into a quantitative measure of uncertainty for each disaster evolution stage. Specifically, this includes: to effectively handle the fuzziness and uncertainty commonly found in geological monitoring data, connecting the original features, calculating the support degree of each feature for each disaster evolution stage based on the preset fuzzy membership function, and converting it into a quantitative measure of uncertainty for each disaster evolution stage; the core function of this fuzzy membership function is to quantify and map the fuzzy relationship between the original features and each disaster evolution stage, which is difficult to define, and can effectively characterize the randomness and fuzziness of the monitoring data, providing a unified quantitative evaluation standard for subsequent fusion calculations; the generated value is the support degree, which is used to quantitatively characterize the degree of fit and support strength of a single original feature for a specific disaster evolution stage. The higher the value, the more the feature tends to be the state feature of the corresponding disaster stage, providing a quantitative basis for subsequent dynamic weighting and confidence synthesis.

[0048] Specifically, the different stages of disaster evolution are first identified, including the stable stage, initial deformation stage, accelerated deformation stage, and near-slip stage. For each stage, a corresponding fuzzy membership function is pre-set. The pre-setting process combines historical disaster case data, geological survey reports, and industry expert experience in the monitoring area. Based on the physical meaning and variation patterns of different types of original features, different types of fuzzy membership functions are adapted. For displacement features, which exhibit a staged linear growth trend, a triangular membership function is adapted, using the critical values ​​of the features at each disaster stage as the function's vertex parameters. For environmental features, which have influence threshold ranges, a trapezoidal membership function is adapted, using the upper and lower limits of the effective influence of the features as the function's interval parameters. For stress and tilt angle features, which exhibit a normal distribution, a Gaussian membership function is adapted, using the mean and fluctuation range of the statistical features in historical disasters as the function's distribution parameters.

[0049] Subsequently, the historical variation range of each original feature and the threshold interval of the disaster stage are analyzed. The degree of overlap between the specific value of each feature and the corresponding threshold interval of the stage is calculated. This degree of overlap is substituted into the fuzzy membership function of the corresponding stage to obtain the support value of the feature for that stage. Furthermore, the support values ​​of all features are standardized, and the proportion of each support value to the sum of the support values ​​of the feature across all stages is calculated. This ensures that the standardized values ​​are within a unified value range of 0 to 1, transforming them into an uncertainty metric that can quantify the degree of correlation between the feature and the disaster stage. Through this process, the originally difficult-to-define fuzzy correlation is transformed into a calculable quantitative indicator, providing a unified quantitative basis for subsequent fusion calculations.

[0050] Step 303: Based on the preset fusion weight strategy corresponding to different disaster evolution stages in the fusion model, dynamically weight the uncertainty quantification of each feature to construct a weighted evidence vector; iteratively synthesize the weighted evidence vector based on the confidence synthesis rule to obtain the comprehensive confidence level of each disaster evolution stage. Specifically, this includes: First, retrieving the preset dynamic weight strategy. The preset process of this strategy is determined by combining historical disaster case data of the monitoring area, geological disaster evolution mechanism, and industry expert experience. Specifically, it involves: firstly, extensively collecting complete monitoring data of past landslide disasters, disaster occurrence process records, and emergency response archives within the monitoring area, and classifying and sorting them according to different stages of disaster evolution; for each stage, statistically analyzing the change range, occurrence sequence, and indicative role of each type of feature in the disaster development process, and calculating the correlation between different features and disaster occurrence. The correlation calculation is based on the time difference between feature change and disaster occurrence, and the correlation between feature change range and disaster scale. On this basis, combined with the judgment of industry experts on the dominant influencing factors of each stage of disaster evolution, the statistically obtained correlation is verified and adjusted, and finally, the weight allocation rule for each stage is determined.

[0051] This strategy determines the contribution ratio of different features at each evolution stage through the aforementioned pre-set process. For example, in the initial deformation stage, the weighting coefficient of environmental features is determined to be 0.6 by calculating the correlation between historical data and disaster occurrence, while the combined weighting coefficients of displacement and stress features are 0.4. In the accelerated deformation and imminent landslide stages, the combined weighting coefficients of displacement and stress features are determined to be 0.7 by calculating the correlation between the rate of change of displacement and stress features and disaster occurrence, while the weighting coefficient of environmental features is 0.3. Subsequently, based on the current disaster evolution stage of the landslide, the corresponding weighting rules are matched, and the uncertainty quantification of each feature is multiplied by the corresponding weighting coefficient to obtain the weighted evidence components. Furthermore, all weighted evidence components are sorted and integrated according to the disaster evolution stage. A weighted evidence vector with consistent dimensions and feature count is constructed. Based on the confidence synthesis rule, the weighted evidence vector is iteratively synthesized. Specifically, the first two evidence vectors are selected, and their intersection and union confidence scores are calculated to obtain a preliminary synthesis result. This preliminary synthesis result is then synthesized with the third evidence vector using the same method, and so on, until all evidence vectors are synthesized. After each round of synthesis, the deviation between the synthesized result and each original evidence vector is calculated, and the deviation is checked to ensure that it is less than a preset consistency threshold, thus ensuring the rationality of the synthesis result. Finally, the comprehensive confidence score for each disaster evolution stage is obtained. This confidence score can comprehensively reflect the degree of comprehensive support of multi-source features for each disaster stage, improving the ability of the fusion result to represent the actual state of the landslide.

[0052] Step 304: Based on the comprehensive confidence level of each disaster evolution stage, defuzzification calculation is performed using preset fuzzy inference rules to integrate and map the comprehensive confidence level of each stage into a single comprehensive landslide stability assessment index. Specifically, this includes: First, setting up fuzzy inference rules. The setting process of these rules is specifically combined with historical assessment data of the monitoring area, landslide disaster evolution mechanisms, and industry expert experience. Specifically, this involves: First, extensively collecting comprehensive confidence level data for each disaster evolution stage of past landslide events within the monitoring area, corresponding records of the actual stability status of the landslide body, and historical assessment results, classifying and sorting them according to stability level; then analyzing the comprehensive confidence level of different disaster evolution stages. The confidence level, and its correlation with the actual stability state of the landslide, is determined based on the priority of the contribution of confidence level at each stage to the stability assessment results. The influence weight of confidence level at different stages in the assessment is calculated. Among them, the pre-landslide stage is the most critical because it is directly related to the urgency of the disaster and has the most significant indication of the stability assessment, so its contribution priority is the highest. The accelerated deformation stage is the second highest, followed by the initial deformation stage, and the stable stage has the lowest contribution priority. On this basis, the calculated weights are verified and adjusted in conjunction with the opinions of industry experts on the influence weight of confidence level at each stage, and finally the weight ratio of comprehensive confidence level at each stage of disaster evolution and its mapping relationship with the stability assessment index are determined.

[0053] This rule, through the aforementioned pre-defined process, determines the weighting of the overall confidence level for each stage of disaster evolution. Specifically, the weighting for the pre-slip stage is set at 0.4, the accelerated deformation stage at 0.3, the initial deformation stage at 0.2, and the stable stage at 0.1. It also establishes the mapping relationship between the confidence level and the stability assessment index for each stage. Subsequently, the overall confidence level for each stage is multiplied by its corresponding weighting, and the sum of all products is calculated to obtain an initial overall value. Further, this initial overall value is standardized by first calculating the maximum and minimum values ​​of the initial overall value in historical data to determine the numerical variation. The system first defines a range, then compares the current initial comprehensive value with this range, and calculates the corresponding value within the standardized numerical range of 0 to 10 to obtain a comprehensive evaluation index within a specific numerical range. Among them, a value of 0 to 3 indicates good stability of the landslide, 3 to 7 indicates moderate stability, and 7 to 10 indicates poor stability, so that the value can intuitively reflect the stability state of the landslide. The final comprehensive evaluation index of landslide stability simplifies the complex multi-source fusion results into a single quantitative indicator, simplifies the information processing process for subsequent early warning decision-making, and provides a concise and effective core evaluation basis for subsequent probability calculation and early warning command generation.

[0054] In this embodiment of the invention, multiple types of original features of spatial grid nodes at various time points are extracted from a heterogeneous monitoring dataset with a unified spatiotemporal benchmark, achieving structured screening and extraction of multi-dimensional monitoring data; ensuring that subsequent processing focuses only on core parameters directly related to landslide stability, eliminating redundant data, and improving the targeting and efficiency of data processing; calculating the support of features for each disaster evolution stage based on fuzzy membership functions and converting it into a quantitative measure of uncertainty, thus quantitatively representing the fuzziness and uncertainty of monitoring data; effectively addressing the fuzziness and randomness problems commonly found in geological monitoring data, transforming the originally difficult-to-quantify feature correlations into calculable quantitative indicators, providing a unified quantitative basis for subsequent fusion calculations; and based on the fusion weights of different disaster evolution stages... The strategy dynamically weights evidence vectors and iteratively synthesizes comprehensive confidence through confidence synthesis rules. It achieves dynamic adaptation of fusion weights to the disaster evolution stages, ensuring that the contribution of each feature aligns with the disaster evolution patterns at different stages. Simultaneously, it iterative synthesis integrates multi-source evidence information, strengthening the comprehensive confidence's ability to represent the actual state of the landslide and improving the reliability and rationality of the fusion results. Fuzzy inference rules are used to defuzzify and calculate the data, integrating it into a single comprehensive landslide stability assessment index. This transforms the multi-stage, multi-dimensional comprehensive confidence into an intuitive, single assessment indicator, simplifying the information processing flow for subsequent early warning decisions. It also ensures that the complex multi-source fusion results have clear interpretability and operability, providing a concise and effective core assessment basis for subsequent probability calculations and early warning command generation.

[0055] In a preferred embodiment of the present invention, step 400 involves inputting the comprehensive evaluation index of landslide stability into a logistic regression probability model trained on historical disaster data for calculation to obtain a real-time occurrence probability value; comparing the real-time occurrence probability value with a preset graded dynamic probability threshold to obtain an early warning instruction, including: Step 401: Based on the comprehensive landslide stability assessment index and combined with historical landslide state data, construct a time-series feature vector for input to the logistic regression probability model. Specifically, this includes: First, extracting the comprehensive landslide stability assessment index from multiple consecutive time points to form real-time time-series assessment data; then, retrieving historical landslide state data and selecting historical stability assessment index sequences similar to the environmental conditions and initial state of the slope during the current monitoring period, along with corresponding status indicators indicating whether historical disasters occurred; further, setting a fixed time window length, concatenating the real-time time-series assessment data with the selected historical assessment data in chronological order, removing abnormal fluctuations in the data sequence, and calculating the time-series continuity index to ensure the concatenated sequence has no time breaks; based on this, arranging the comprehensive stability assessment indices of each time point in chronological order to construct a time-series feature vector with dimensions consistent with the time window length, while standardizing all data within the vector to ensure the data is within a uniform numerical range, providing regular and effective input features for subsequent model calculations.

[0056] Step 402: Input the time series feature vector into the logistic regression probability model trained on the historical disaster training sample set. Calculate the real-time probability value of geological disaster occurrence using the logistic regression coefficient matrix and sigmoid activation function of the logistic regression probability model. Specifically, this includes: inputting the time series feature vector into the logistic regression probability model trained on the historical disaster training sample set; the construction and training process of this logistic regression probability model is as follows: First, construct a historical disaster training sample set, extensively collecting time-series monitoring data, stability assessment results, and disaster occurrence status records of past landslide disasters within the monitoring area. Organize these into time-series feature samples according to a preset time window length, and label each sample with a corresponding disaster occurrence status label; then, preprocess the sample set to remove abnormal samples. The sample set, including both training and validation subsets, is divided into training and validation subsets according to a preset ratio. Based on this, the basic structure of a logistic regression probability model is constructed. The dimension of the logistic regression coefficient matrix is ​​determined based on the dimension of the time series feature vectors, with each element in the coefficient matrix corresponding to the influence weight of a time-series feature. An iterative learning algorithm is used to train the model, inputting the training subset into the model. By continuously adjusting the element values ​​of the coefficient matrix, the deviation between the model's output prediction results and the sample labels is gradually reduced. After each iteration, the validation subset is used to verify the model's performance, calculating the prediction accuracy and error index. If the preset performance standard is not met, iterative adjustments continue until the model performance meets the requirements. Finally, the logistic regression coefficient matrix is ​​determined, which can quantitatively characterize the degree of influence of each time-series feature on the probability of disaster occurrence.

[0057] Specifically, the time series feature vector and the logistic regression coefficient matrix are linearly combined to obtain a linear combination result. This result is then input into a sigmoid activation function for mapping calculation. The core function of the sigmoid activation function is to perform a non-linear mapping of the linear combination result, converting linear values ​​of any range into values ​​between 0 and 1, effectively realizing the transformation from linear calculation results to probability values, while smoothly distinguishing the boundaries of different disaster probabilities. The generated value is the real-time probability value of geological disasters, used to quantitatively characterize the likelihood of a geological disaster occurring at the current landslide site within a preset time period. The closer the value is to 1, the higher the probability of disaster occurrence; the closer it is to 0, the lower the probability of disaster occurrence, providing accurate quantitative basis for subsequent risk classification. After calculation, the real-time probability value is validated by comparing it with the probability value range of similar historical conditions. If it exceeds a reasonable range, the model parameters are retrieved again for a second calculation until a real-time probability value within a reasonable range is obtained, ensuring the reliability of the calculation results.

[0058] Step 403: Compare the real-time occurrence probability value with a preset hierarchical dynamic probability threshold to determine the membership status of the probability value to each preset threshold interval; determine the risk classification judgment result corresponding to the real-time occurrence probability value based on the membership status, specifically including: comparing the real-time occurrence probability value with a preset hierarchical dynamic probability threshold to determine the membership status of the probability value to each preset threshold interval; specifically, the preset process of the hierarchical dynamic probability threshold is as follows: first, extensively collect basic data of the monitoring area, including the properties of soil and rock mass, slope structure type and potential disaster scale of the landslide geological survey report, the disaster occurrence probability range corresponding to the historical disaster level record, and the response time and rescue resources covered by the regional emergency response capacity assessment data. The system first defines the source configuration and control scope. Based on this, it analyzes the correlation between the historical probability of disaster occurrence and the actual disaster level, and combines the risk tolerance threshold corresponding to emergency response capabilities to initially divide the probability ranges corresponding to low, medium, high, and extremely high risks. Then, based on the assessment opinions of industry experts on the landslide risk characteristics of the monitoring area, the initially divided ranges are verified and fine-tuned to ensure that the boundary values ​​of each range can accurately distinguish the differences between different risk levels, while reserving space for dynamic adjustment. Finally, it determines the triggering conditions and adjustment rules for dynamic adjustment. When seasonal changes cause changes in the water content of the soil and rock, or when extreme environmental conditions such as heavy rainfall or continuous drought occur, the threshold ranges of each range can be adaptively adjusted according to the preset rules to ensure that the thresholds always conform to the actual risk situation.

[0059] The preset dynamic probability thresholds are formed through the above process, including low-risk, medium-risk, high-risk, and extremely high-risk threshold ranges. The threshold range of each range can be dynamically adjusted according to environmental factors such as seasonal changes and rainfall intensity. The real-time probability value is compared with the upper and lower limits of each threshold range in turn. If the probability value falls within a certain threshold range, it is determined that the probability value belongs to that range. The corresponding risk classification result is matched according to the affiliation status, that is, the low-risk threshold range corresponds to the low-risk level, the medium-risk threshold range corresponds to the medium-risk level, the high-risk threshold range corresponds to the high-risk level, and the extremely high-risk threshold range corresponds to the extremely high-risk level. After the determination is completed, the risk classification result and the corresponding probability value range are recorded to provide a basis for the generation of subsequent early warning instructions.

[0060] Step 404: Based on the risk classification determination results and combined with the landslide spatial location information represented by the heterogeneous monitoring dataset, an early warning instruction containing risk level, spatial location, and emergency guidance is automatically generated. Specifically, this includes: First, extracting the spatial grid node coordinates of the landslide corresponding to the risk classification from the heterogeneous monitoring dataset, converting the coordinates into the actual geographical location information of the monitoring area, and determining the specific spatial range of the risk occurrence; then matching the corresponding emergency guidance content according to the risk classification determination results: low risk level corresponds to routine patrol guidance, medium risk level corresponds to intensified monitoring and patrol guidance, high risk level corresponds to personnel evacuation preparation and key control guidance, and extremely high risk level corresponds to immediate evacuation and area lockdown guidance; integrating the risk level, specific spatial location information, and corresponding emergency guidance content, standardizing them according to a preset instruction format, and supplementing basic information such as monitoring time and data source; finally generating a standardized early warning instruction, which can be directly transmitted to the emergency response terminal, providing clear and accurate action guidance for on-site emergency response, and effectively improving the efficiency and pertinence of emergency response.

[0061] In this embodiment of the invention, a time-series feature vector is constructed by integrating the comprehensive landslide stability assessment index with historical state data, achieving effective correlation between real-time assessment data and historical time-series data. The time-series feature vector is then input into a logistic regression probability model trained on historical disaster data for calculation. Quantification is completed using the logistic regression coefficient matrix and sigmoid activation function generated during model training. This achieves a systematic and standardized solution for time-series features, ensuring that the generation of real-time probability values ​​has a unified and reliable processing logic, guaranteeing that the probability values ​​can accurately quantify the likelihood of current landslide disasters, and improving the standardization and consistency of probability calculation results. A hierarchical dynamic... The system compares the state probability threshold with the real-time occurrence probability value to determine the threshold state and classify the risk level. This enables a refined classification of disaster risks, allowing risk assessment to adapt to different levels of disaster likelihood, improving the pertinence and rationality of risk classification, and providing a basis for generating differentiated early warning instructions. It integrates the risk classification results with the spatial location information of the landslide body to generate early warning instructions, consolidating core information such as risk level, spatial location, and emergency guidance. This effectively transforms data processing results into practical early warning information, ensuring that early warning instructions combine risk level identification, precise spatial orientation, and operable emergency guidance, thus enhancing the practicality and operability of early warning information.

[0062] In a preferred embodiment of the present invention, step 500, based on an early warning command, synchronously drives designated remote and local terminals through a preset hybrid communication channel, and links with relevant emergency systems to execute a closed-loop early warning response, including: Step 501: Parse the warning instruction, extract the risk level, spatial location, and emergency guidance information, and encapsulate the risk level, spatial location, and emergency guidance information into a standard communication message according to a preset communication protocol specification. Specifically, this includes: First, parsing the warning instruction; specifically, according to the preset field structure of the warning instruction, extracting the core information corresponding to the risk level field, spatial location field, and emergency guidance field in sequence, and removing redundant format identification information in the instruction; further, retrieving the preset communication protocol specification. The specific preset process of this specification is as follows: First, comprehensively analyze the transmission characteristics of the satellite communication link and the terrestrial wireless communication network to be used, including the bandwidth limitations, transmission delay, anti-interference capabilities, and data transmission format requirements of the two links; at the same time, analyze the communication between the remote monitoring and warning platform and the on-site emergency terminal. The interface parameters and data parsing capabilities are defined. Based on this, and considering the core information types that the warning instructions need to transmit, the fields that the standard message must include are determined, along with the core functions and data types of each field. The field arrangement order is then established, prioritizing key information fields such as risk level and spatial location to improve transmission and parsing efficiency. Simultaneously, a data encoding format adapted to the transmission characteristics of the two communication links is determined, selecting an encoding method with strong anti-interference capabilities and high encoding efficiency. Furthermore, message integrity verification rules are designed, and an appropriate verification algorithm is selected based on the bit error rate characteristics of the link transmission to ensure that no data loss or errors occur during data transmission. Finally, the adaptability of the specification is verified through multiple sets of link transmission tests. Based on the test results, the field parameters, encoding format, and verification rules are adjusted to ensure that the specification can adapt to the transmission requirements of mixed communication channels, ultimately forming a pre-defined communication protocol specification.

[0063] This specification adapts to the transmission characteristics of subsequent hybrid communication channels, determines the field arrangement order, data encoding format, and verification rules of the message; the extracted risk level, spatial location, and emergency guidance information are filled into the corresponding fields of the standard message according to the above specification, and the integrity check code of the message is calculated and added to the end of the message; after filling is completed, the standard communication message is format verified to check whether the field length and encoding format meet the specification requirements, ensuring that the encapsulated standard communication message can adapt to the transmission requirements of satellite communication links and terrestrial wireless communication networks, laying the foundation for subsequent synchronous distribution and transmission.

[0064] Step 502 involves synchronously distributing the standard communication message to the remote monitoring and early warning platform and the on-site emergency terminal via a hybrid communication channel composed of a satellite communication link and a terrestrial wireless communication network. Specifically, this includes: first, pre-checking the link status of the hybrid communication channel by detecting the signal strength of the satellite communication link and the bandwidth and latency of the terrestrial wireless communication network to ensure both links are available; second, simultaneously loading the standard communication message into the transmission queues of both communication links, setting the transmission priority of both links to equal priority, and initiating a synchronous transmission mechanism; during transmission, monitoring the transmission progress of both links in real time, and immediately triggering a link switching compensation mechanism if one link experiences a transmission interruption, allowing the other link to continue transmitting the remaining data, ensuring that the standard communication message can be synchronously delivered to the remote monitoring and early warning platform and the on-site emergency terminal, avoiding delays or loss of early warning information transmission due to a single channel failure, and guaranteeing the full coverage and reliability of early warning information transmission.

[0065] Step 503: In response to the receipt of the standard communication message, trigger the audible and visual alarm device on site, and drive the on-site emergency terminal to display and broadcast the risk level, location, and guidance information contained in the warning instruction. Specifically, this includes: First, performing an integrity check on the received standard communication message, verifying whether there are any transmission errors by checking the checksum at the end of the message. If an error is found, requesting retransmission from the sender. After successful verification, parsing the message to extract the risk level, spatial location, and emergency guidance information. Based on this, triggering the on-site audible and visual alarm device, setting the corresponding alarm mode according to the risk level: low-risk level corresponds to low-frequency audible and visual prompts, medium-high risk level corresponds to high-frequency audible and visual warnings, and extremely high risk level corresponds to continuous audible and visual alarms with superimposed voice broadcast trigger signals. Simultaneously, driving the on-site emergency terminal to display the parsed risk level, spatial location, and emergency guidance information according to the preset interface layout, and synchronously activating the terminal's voice broadcast function to broadcast the core information in clear and easy-to-understand language logic, ensuring that the warning status can be quickly perceived and accurate emergency action guidance can be obtained.

[0066] Step 504: In response to the receipt of the standard communication message, the remote monitoring and early warning platform updates and records the status of the early warning event, obtaining the updated status of the early warning event. Simultaneously, it pushes alarm information containing all the contents of the early warning instruction to authorized monitoring nodes. Specifically, this includes: to achieve standardized remote control and information synchronization of early warning events, and to connect the remote monitoring and early warning platform's reception of the standard communication message, firstly, parsing the message content and extracting all the core information of the early warning instruction; further, updating and recording the status of the early warning event in the early warning event management module of the remote monitoring and early warning platform, specifically including adding early warning event entries, filling in key information such as the early warning occurrence time, risk level, affected spatial range, and data source, establishing a complete early warning event archive, and achieving traceable management of early warning events; based on this, according to preset authorization rules, selecting authorized monitoring nodes that need to be simultaneously aware of the early warning information, including terminal nodes of monitoring management departments at all levels and emergency command centers; converting the alarm information containing all the contents of the early warning instruction according to the communication protocol adaptation requirements of the authorized nodes, and pushing it to each authorized monitoring node through a dedicated communication link, ensuring that remote control personnel can fully and promptly grasp the details of the early warning event, providing complete information support for remote emergency decision-making.

[0067] Step 505: Based on the updated warning event status and the risk level and spatial location information in the warning instructions, automatically generate linkage control instructions; send the linkage control instructions to the regional traffic signal control system and related emergency department systems; the regional traffic signal control system and related emergency department systems receive and execute the linkage control instructions, triggering preset traffic control and emergency response linkage control logic to complete the closed-loop warning response. Specifically, this includes: based on the updated warning event status and the risk level and spatial location information in the warning instructions, first constructing a linkage control rule mapping table, which determines the linkage control actions corresponding to different risk levels and spatial locations; according to the risk level and spatial location of the warning event, matching the corresponding linkage control logic from the mapping table, automatically generating linkage control instructions, which contain core content such as the control object identifier, control action requirements, and execution time limit; further, encapsulating the linkage control instructions according to the communication protocol of the corresponding system, and sending them to the regional traffic signal control system and related emergency department systems respectively; after receiving the linkage control instructions, each receiving system parses the instruction content and triggers the preset traffic control and emergency response linkage control logic.

[0068] The pre-set process for this traffic control and emergency response linkage control logic is as follows: First, extensively collect traffic control and handling records, emergency response execution procedures, functional parameters and operating specifications of relevant systems for historical landslide disasters within the monitoring area. Simultaneously, analyze the regional road network distribution, key road node locations, emergency rescue force deployment, and emergency resource reserves. Based on this, analyze the core needs of traffic control at each level, considering the disaster impact range and urgency corresponding to different risk levels. Determine the traffic diversion, control, or closure measures required for low, medium, high, and extremely high risk levels, and match the corresponding emergency response levels to determine the necessary emergency departments and specific response actions. Then, combining the technical capabilities and operational permissions of various receiving systems such as the regional traffic signal control system and emergency rescue team dispatch system, translate the above requirements into executable actions for each system. The control logic determines the action sequence and coordination mechanism of each system to avoid conflicts between different systems. Finally, the feasibility and adaptability of the control logic are verified through multi-scenario simulation tests. Logical details are adjusted based on problems discovered during testing, and opinions are solicited from traffic management departments, emergency management departments, and relevant system operation and maintenance units for optimization. This ultimately forms the pre-set traffic control and emergency response linkage control logic. For example, under low-risk levels, traffic flow guidance is initiated; under medium- and high-risk levels, road entrances to landslide-affected areas are closed and vehicles are guided to detour; under extremely high-risk levels, the emergency rescue team dispatch system is linked to initiate rescue preparations. After each system completes its execution, the execution results are fed back to the remote monitoring and early warning platform, which records the execution status, ultimately completing a closed-loop early warning response from early warning issuance to emergency response execution, effectively shortening the emergency response chain and improving the collaborative efficiency of emergency response.

[0069] In this embodiment of the invention, the core information of the warning command is extracted and encapsulated into a standard communication message, realizing the structured sorting and standardized conversion of the warning information; ensuring that the information format conforms to the communication protocol specification, improving the adaptability of the information between different communication links and terminals; synchronously distributing the standard message through a hybrid communication channel composed of satellite and terrestrial wireless, relying on multi-link redundant transmission to improve the stability of information distribution; realizing the synchronous acquisition of information between the remote platform and the field terminal, avoiding information transmission interruption caused by a single channel failure; triggering audible and visual alarms on site and driving the terminal to display and broadcast the warning information, quickly converting the standardized message into a perceptible warning signal and intuitive information on site; realizing the immediate delivery of warning information to on-site personnel, simplifying The on-site personnel information acquisition process ensures that emergency guidelines can be quickly implemented; the remote platform updates and records the warning status and pushes alarm information to authorized nodes, realizing standardized archiving of warning events and multi-dimensional information synchronization; it ensures that remote control personnel have a comprehensive grasp of the warning details, providing complete information support for remote decision-making, while forming a traceable record of warning events, improving the standardization of warning management; it generates linkage control commands and drives the traffic and emergency systems to execute linkage logic, realizing the automatic conversion of warning information into multi-system collaborative handling actions; it integrates regional traffic and emergency resources to form a closed-loop connection from warning issuance to emergency response, improving the coordination and efficiency of emergency response, and enabling data processing results to effectively empower actual emergency management.

[0070] like Figure 2 As shown, embodiments of the present invention also provide a geological disaster monitoring and early warning system based on multi-source sensor data fusion, comprising: The data acquisition module is used to deploy no fewer than three geodetic control stakes within the landslide monitoring area to form a geodetic control network with spatial coordinates. Based on the spatial coordinates provided by the geodetic control network, all displacement, environmental parameter, and structural stability monitoring sensors are spatially calibrated and networked to construct a sensor network with a unified spatial reference. On the basis of the sensor network with a unified spatial reference, the synchronous triggering and data acquisition of each sensor are performed to obtain a multi-source monitoring data sequence. The spatiotemporal registration module is used to perform spatiotemporal registration operations on multi-source monitoring data sequences using the spatial coordinate constraints provided by the geodetic control network, and generate heterogeneous monitoring datasets with unified spatiotemporal references. The fusion module is used to input heterogeneous monitoring datasets with unified spatiotemporal references into a preset confidence synthesis and fuzzy inference fusion model, and perform fusion calculations according to preset fusion weight strategies corresponding to different disaster evolution stages to obtain a comprehensive evaluation index of landslide stability. The decision-making module is used to input the comprehensive stability assessment index of the landslide body into a logistic regression probability model trained on historical disasters for calculation to obtain the real-time occurrence probability value; compare the real-time occurrence probability value with the preset graded dynamic probability threshold to make a decision and obtain an early warning instruction; The execution module is used to synchronously drive designated remote and local terminals through a preset hybrid communication channel based on the early warning command, and to link with relevant emergency systems to execute a closed-loop early warning response.

[0071] It should be noted that this system is a system corresponding to the above method. All implementation methods in the above method embodiments are applicable to this embodiment and can achieve the same technical effect.

[0072] Embodiments of the present invention also provide a computing device, including: a processor and a memory storing a computer program, wherein the computer program, when executed by the processor, performs the method described above. All implementations in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0073] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A geological disaster monitoring and early warning method based on multi-source sensor data fusion, characterized in that, The method includes: Step 100: Deploy no fewer than three geodetic control stakes within the landslide monitoring area to form a geodetic control network with spatial coordinates; Based on the spatial coordinates provided by the geodetic control network, spatially calibrate and network all displacement, environmental parameter, and structural stability monitoring sensors to construct a sensor network with a unified spatial reference; On the basis of the sensor network with a unified spatial reference, perform synchronous triggering and data acquisition of each sensor to obtain a multi-source monitoring data sequence; Step 200: Using the spatial coordinate constraints provided by the geodetic control network, perform spatiotemporal registration on the multi-source monitoring data sequence to generate a heterogeneous monitoring dataset with unified spatiotemporal reference. Step 300: Input the heterogeneous monitoring dataset with unified spatiotemporal reference into the preset confidence synthesis and fuzzy inference fusion model, and perform fusion calculation according to the preset fusion weight strategy corresponding to different disaster evolution stages to obtain the comprehensive evaluation index of landslide stability. Step 400: Input the comprehensive evaluation index of landslide stability into the logistic regression probability model trained on historical disasters for calculation to obtain the real-time occurrence probability value; compare the real-time occurrence probability value with the preset graded dynamic probability threshold to make a decision and obtain an early warning instruction. Step 500: Based on the early warning command, the designated remote and local terminals are synchronously driven through a preset hybrid communication channel, and relevant emergency systems are linked to execute a closed-loop early warning response.

2. The geological disaster monitoring and early warning method based on multi-source sensor data fusion according to claim 1, characterized in that, Step 100 includes: Within the landslide monitoring area, select no fewer than three geologically stable locations and set up geodetic control stakes at each location; Using satellite positioning technology, all established geodetic control stakes were simultaneously measured to obtain raw observation data. The original observation data are subjected to network adjustment calculations to determine the three-dimensional absolute coordinates of each control point in the national coordinate system; Based on the three-dimensional absolute coordinates of all control stakes, a geodetic control network with unified known spatial coordinates is formed; Based on the coordinate framework of the geodetic control network, displacement, environmental parameters and structural stability monitoring sensors are deployed at key geological feature points of the landslide body to obtain the locations of the deployed sensors. Based on the locations of the deployed sensors, a total station is used to measure the installation point of each sensor and obtain its relative position observation value relative to the nearest geodetic control stake. The relative position observations are intersected with the known absolute coordinates of the corresponding control stakes to determine the absolute spatial coordinates of each sensor. Based on the absolute spatial coordinates of all sensors, the spatial reference of the sensor network is unified, resulting in a sensor network with a unified spatial reference.

3. The geological disaster monitoring and early warning method based on multi-source sensor data fusion according to claim 2, characterized in that, Step 100 further includes: Based on the sensor network with a unified spatial reference, a synchronous acquisition command is sent to all sensors within the network. In response to the synchronous acquisition command, each sensor is triggered to perform synchronous data acquisition operation, and the raw electrical signals and digital signals generated by each sensor after performing the synchronous data acquisition operation are received. The original electrical signal and digital signal are amplified, filtered, and converted from analog to digital to obtain an initial digital sampling sequence; The calibration coefficients of each sensor are used to convert the initial digital sampling sequence into physical quantity values ​​with standard units; The physical quantity value corresponding to each sensor is bound to its absolute spatial coordinates to obtain physical quantity data with three-dimensional spatial coordinates; Based on the physical quantity data with three-dimensional spatial coordinates, timestamp synchronization and alignment operations are performed on it using a unified time reference, and interpolation methods are used to fill the data gaps caused by different sensor sampling rates, so as to obtain data that has completed time synchronization and gap filling. Based on the data from the completed time synchronization and gap filling, the physical quantity data of displacement, environment, stress and tilt angle are integrated in chronological order into landslide surface and deep displacement sequences, rainfall and soil water content sequences, and soil internal stress and slope tilt angle change sequences, which together constitute a multi-source monitoring data sequence.

4. The geological disaster monitoring and early warning method based on multi-source sensor data fusion according to claim 3, characterized in that, Step 200 includes: From the multi-source monitoring data sequence, the landslide surface and deep displacement sequence, rainfall and soil moisture state sequence, and soil internal stress and slope dip angle change sequence are extracted respectively. For each data point in the landslide surface and deep displacement sequence, rainfall and soil water content sequence, and soil internal stress and slope dip angle change sequence, a timestamp is assigned based on the recorded time information and a unified clock reference. Based on the timestamp, all data points are mapped to a unified time axis to identify data points with inconsistent sampling times; For data points with inconsistent sampling times, interpolation calculations are performed within the interval formed by adjacent data points to generate interpolated data at the synchronization time. All mapped data points are merged with the interpolated data at the synchronization time to obtain a multi-source data sequence with complete timing alignment.

5. The geological disaster monitoring and early warning method based on multi-source sensor data fusion according to claim 4, characterized in that, Step 200 further includes: Using the spatial coordinate constraints of the geodetic control network, constrained adjustment and optimization calculations are performed on the sensor spatial coordinates associated with the time-aligned multi-source data sequences to obtain the adjusted sensor spatial coordinates. The adjusted sensor spatial coordinates are re-associated with the multi-source data sequence to generate a spatiotemporal registration fusion dataset; Based on the aforementioned spatiotemporal registration and fusion dataset, a regular three-dimensional spatial grid covering the monitoring area is constructed; The physical quantity data in the spatiotemporal registration and fusion dataset are resampled to each node of the three-dimensional spatial grid to obtain gridded multi-physical quantity field data; Based on the gridded multi-physical field data, the data is organized and aggregated according to time and space dimensions into a multi-physical spatial field snapshot sequence arranged in time order, and finally packaged into a heterogeneous monitoring dataset with unified spatiotemporal reference.

6. The geological disaster monitoring and early warning method based on multi-source sensor data fusion according to claim 5, characterized in that, Step 300 includes: Displacement, environment, stress, and tilt angle data of spatial grid nodes at each time point are extracted as raw features from the heterogeneous monitoring dataset with unified spatiotemporal reference. Based on the original features, according to the preset fuzzy membership function, the support degree of each feature for each disaster evolution stage is calculated, and it is transformed into a quantitative measure of the uncertainty of each disaster evolution stage. Based on the fusion weight strategy pre-set in the fusion model corresponding to different disaster evolution stages, the uncertainty quantification of each feature is dynamically weighted to construct a weighted evidence vector; the weighted evidence vector is iteratively synthesized based on the confidence synthesis rule to obtain the comprehensive confidence of each disaster evolution stage; Based on the comprehensive confidence level of each disaster evolution stage, defuzzification calculation is performed through preset fuzzy inference rules, and the comprehensive confidence level of each stage is integrated and mapped into a single comprehensive landslide stability assessment index.

7. The geological disaster monitoring and early warning method based on multi-source sensor data fusion according to claim 6, characterized in that, Step 400 includes: Based on the comprehensive evaluation index of landslide stability, and combined with historical state data of the landslide, a time series feature vector is constructed for input to the logistic regression probability model. The time series feature vector is input into the logistic regression probability model trained on the historical disaster training sample set. The real-time occurrence probability value of geological disaster is calculated by using the logistic regression coefficient matrix and sigmoid activation function of the logistic regression probability model. The real-time occurrence probability value is compared with a preset hierarchical dynamic probability threshold to determine the membership status of the probability value to each preset threshold interval; the risk classification judgment result corresponding to the real-time occurrence probability value is determined based on the membership status. Based on the risk classification results, and combined with the spatial location information of the landslide body represented by the heterogeneous monitoring dataset, an early warning instruction containing risk level, spatial location, and emergency guidance is automatically generated.

8. The geological disaster monitoring and early warning method based on multi-source sensor data fusion according to claim 7, characterized in that, Step 500 includes: The warning instruction is parsed to extract the risk level, spatial location, and emergency guidance information, and then encapsulated into a standard communication message according to a preset communication protocol specification. The standard communication messages are synchronously distributed and transmitted to the remote monitoring and early warning platform and the on-site emergency terminal through a hybrid communication channel consisting of a satellite communication link and a terrestrial wireless communication network. In response to the receipt of the standard communication message, an audible and visual alarm device is triggered on-site, and the on-site emergency terminal is driven to display and broadcast the risk level, location, and guidance information contained in the warning instruction; In response to the receipt of the standard communication message, the remote monitoring and early warning platform updates and records the status of the early warning event, obtains the updated status of the early warning event, and pushes alarm information containing all the contents of the early warning instruction to the authorized monitoring node. Based on the updated warning event status and the risk level and spatial location information in the warning instructions, a linkage control instruction is automatically generated; the linkage control instruction is sent to the regional traffic signal control system and related emergency department systems; the regional traffic signal control system and related emergency department systems receive and execute the linkage control instruction, triggering the preset traffic control and emergency response linkage control logic to complete the closed-loop warning response.

9. A geological disaster monitoring and early warning system based on multi-source sensor data fusion, wherein the system implements the method as described in any one of claims 1 to 8, characterized in that, include: The data acquisition module is used to set up no fewer than three geodetic control stakes within the landslide monitoring area to form a geodetic control network with spatial coordinates. Based on the spatial coordinates provided by the geodetic control network, spatial calibration and network deployment are carried out for all displacement, environmental parameter and structural stability monitoring sensors to build a sensor network with a unified spatial reference. Based on a sensor network with a unified spatial reference, synchronous triggering and data acquisition of each sensor are performed to obtain multi-source monitoring data sequences; The spatiotemporal registration module is used to perform spatiotemporal registration operations on multi-source monitoring data sequences using the spatial coordinate constraints provided by the geodetic control network, and generate heterogeneous monitoring datasets with unified spatiotemporal references. The fusion module is used to input heterogeneous monitoring datasets with unified spatiotemporal references into a preset confidence synthesis and fuzzy inference fusion model, and perform fusion calculations according to preset fusion weight strategies corresponding to different disaster evolution stages to obtain a comprehensive evaluation index of landslide stability. The decision-making module is used to input the comprehensive stability assessment index of the landslide body into the logistic regression probability model trained on historical disasters for calculation, and obtain the real-time probability value of occurrence. The real-time occurrence probability value is compared with the preset hierarchical dynamic probability threshold to make a decision and obtain an early warning instruction; The execution module is used to synchronously drive designated remote and local terminals through a preset hybrid communication channel based on the early warning command, and to link with relevant emergency systems to execute a closed-loop early warning response.

10. A computing device, characterized in that, include: One or more processors; A storage device for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the method as described in any one of claims 1 to 8.

Images