Landslide resistivity prediction imaging method, device, equipment, medium and program product
By integrating data from multiple resistivity acquisition devices and dynamically allocating weights based on geological conditions, the problem of data distortion in traditional landslide detection has been solved, enabling high-precision resistivity prediction and status monitoring of landslides in accumulated layers.
Patent Information
- Application Number
- CN202611165797.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-03
- Publication Date
- 2026-11-06
Smart Images
Figure CN123052102A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of data processing technology, and in particular to a method, apparatus, equipment, medium, and program product for predicting the resistivity of landslide bodies. Background Technology
[0002] Accumulated landslides are one of the most common types of geological hazards in mountainous areas. Their sliding surfaces are mostly distributed within 30m of the surface at the interface between the overburden and bedrock, characterized by strong concealment, rapid evolution, and high destructiveness. Resistivity imaging technology, with its advantages of non-destructive, continuous, and visual detection, has become a core technical means for detecting the internal structure and monitoring deformation of landslides.
[0003] Traditional techniques employ three types of resistivity acquisition devices for landslide detection: the Winner device, the dipole-dipole device, and the Winner-Schlumberger device. These devices exhibit significant differences in detection sensitivity: the Winner device is suitable for detecting horizontally distributed geological structures, the dipole-dipole device for vertically distributed geological structures, and the Winner-Schlumberger device can detect structures directly beneath the geological mass. Traditional monitoring schemes typically use a single acquisition device to collect resistivity data, followed by conventional filtering algorithms for data processing and imaging, achieving static detection of the electrical structure of the landslide body.
[0004] However, traditional monitoring schemes use only a single resistivity acquisition device, which cannot match the special geological structure characteristics of landslides extending downwards in the accumulation layer. The sensitivity of a single device is limited, and the acquired resistivity data has a low degree of matching with the actual electrical structure of the landslide, which easily leads to data distortion and inaccurate structural characterization. Summary of the Invention
[0005] Therefore, it is necessary to provide a landslide resistivity prediction imaging method, device, equipment, medium, and program product that can improve the accuracy of the above-mentioned technical problems.
[0006] In a first aspect, this application provides a landslide resistivity prediction imaging method, the method comprising:
[0007] The raw resistivity data collected by each device and the geological parameters of the landslide body are obtained, and the raw resistivity data collected by each device are fused based on the geological parameters of the landslide body to obtain multi-source resistivity data.
[0008] Based on the multi-source resistivity data, predicted resistivity data is obtained, and based on the multi-source resistivity data and the predicted resistivity data, the landslide state of the landslide body is determined;
[0009] Based on the landslide state, a corresponding target observation model is determined, and the multi-source resistivity data is processed using the determined target observation model to obtain target resistivity data;
[0010] A predicted image is generated based on the target resistivity data.
[0011] In some optional embodiments, the raw resistivity data collected by each device is fused based on the geological parameters of the landslide body to obtain multi-source resistivity data, including:
[0012] Based on the sensitivity distribution characteristics of each device, a mapping relationship between device sensitivity and spatial location is established;
[0013] The geological parameters of the landslide and the mapping relationship are input into a deep neural network, and combined with nonlinear regression analysis, the fusion weights of each device under the geological parameters are obtained.
[0014] The raw resistivity data collected by each device are weighted based on the fusion weight to obtain multi-source resistivity data.
[0015] In some optional embodiments, obtaining predicted resistivity data based on the multi-source resistivity data, and determining the landslide state of the landslide body based on the multi-source resistivity data and the predicted resistivity data, includes:
[0016] The initial transition probabilities between the observation models corresponding to each device are obtained. The initial transition probabilities are input into the Markov transition matrix to perform observation model interaction, thereby obtaining the prediction probability of each observation model. The prediction probability of each observation model and the state prediction value of each observation model are weighted and fused to obtain the predicted resistivity data.
[0017] An information vector is obtained based on the multi-source resistivity data and the predicted resistivity data;
[0018] A discriminant function is constructed based on the information vector and the covariance matrix corresponding to the information vector.
[0019] The detection threshold is corrected based on the geological disturbance parameters of the landslide body;
[0020] The landslide state is determined by comparing the value of the discriminant function with the detection threshold.
[0021] In some optional embodiments, determining the landslide state based on a comparison between the value of the discriminant function and the detection threshold includes:
[0022] If the value of the discrimination function is less than or equal to the detection threshold, the landslide state is determined to be a stable state.
[0023] If the value of the discrimination function is greater than the detection threshold, the landslide state is determined to be an accelerated deformation state on the verge of landslide.
[0024] In some optional embodiments, the step of determining the corresponding target observation model based on the landslide state, and processing the multi-source resistivity data using the determined target observation model to obtain target resistivity data, includes:
[0025] When the landslide is in a stable state, the single-device observation model selected according to the geological conditions is used as the target observation model. The multi-source resistivity data is input into the target observation model for filtering to obtain the target resistivity data.
[0026] When the landslide is in a state of accelerated deformation and imminent landslide, the filtered values and covariance of the observation models corresponding to each device at the previous time step are mixed and interacted with to obtain the interaction input values of each observation model. Each interaction input value is then input into each observation model, and the current observation value is combined with the current observation value to obtain the current filtering result. The probabilities of each observation model are then updated. Based on the probabilities of each observation model, the current filtering results are weighted and fused to obtain the target resistivity data.
[0027] In some alternative embodiments, the method further includes:
[0028] Obtain independent true data through borehole sampling and stratified soil moisture monitoring;
[0029] Based on the predicted image and the independent ground truth data, the fusion weights of the raw resistivity data collected by each device, the state equation parameters of each observation model, and the noise covariance parameters in the filtering process are iteratively adjusted.
[0030] Secondly, this application also provides a landslide resistivity prediction imaging device, the device comprising:
[0031] The data acquisition module is used to acquire the raw resistivity data collected by each device and the geological parameters of the landslide body, and to fuse the raw resistivity data collected by each device based on the geological parameters of the landslide body to obtain multi-source resistivity data.
[0032] The state determination module is used to obtain predicted resistivity data based on the multi-source resistivity data, and to determine the landslide state of the landslide body based on the multi-source resistivity data and the predicted resistivity data.
[0033] The observation module is used to determine the corresponding target observation model based on the landslide state, and to process the multi-source resistivity data through the determined target observation model to obtain the target resistivity data;
[0034] A generation module is used to generate a predicted image based on the target resistivity data.
[0035] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described in any of the above embodiments.
[0036] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the method described in any of the above embodiments.
[0037] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the method described in any of the above embodiments.
[0038] The aforementioned landslide resistivity prediction imaging method, device, equipment, medium, and program products acquire raw resistivity data collected by each device and geological parameters of the landslide. Based on the geological parameters of the landslide, the raw resistivity data collected by each device is fused to obtain multi-source resistivity data. Based on the multi-source resistivity data, predicted resistivity data is obtained, and the landslide state of the landslide is determined based on the multi-source resistivity data and the predicted resistivity data. Based on the landslide state, a corresponding target observation model is determined, and the multi-source resistivity data is processed through the determined target observation model to obtain target resistivity data. A predicted image is generated based on the target resistivity data. This method integrates data from three types of devices: Wenner-Schlumberger, Wenner, and dipole-dipole, and dynamically allocates weights based on on-site geological conditions. This solves the problems of insufficient sensitivity and data distortion of single devices, adapts to complex landslide geological scenarios, and improves accuracy. Attached Figure Description
[0039] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.
[0040] Figure 1 This is a diagram illustrating the application environment of a landslide resistivity prediction imaging method in one embodiment.
[0041] Figure 2 This is a flowchart illustrating a landslide resistivity prediction imaging method in one embodiment;
[0042] Figure 3 Here is a flowchart of step S202 in one embodiment;
[0043] Figure 4This is a diagram showing the three device polarity running modes and the distribution of measurement data points in one embodiment;
[0044] Figure 5 This is a sensitivity distribution diagram of three devices in one embodiment;
[0045] Figure 6 Here is a flowchart of step S206 in one embodiment;
[0046] Figure 7 This is a structural block diagram of a landslide resistivity prediction imaging device in one embodiment;
[0047] Figure 8 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation
[0048] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0049] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.
[0050] The landslide resistivity prediction imaging method provided in this application can be applied to, for example... Figure 1 In the application environment shown, each device 102 communicates with the server 104 via a network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated onto the server 104, or it can be located in the cloud or on another network server.
[0051] Each device can collect raw resistivity data. The server 104 acquires the raw resistivity data collected by each device and the geological parameters of the landslide body. Based on the geological parameters of the landslide body, the server merges the raw resistivity data collected by each device to obtain multi-source resistivity data. Based on the multi-source resistivity data, predicted resistivity data is obtained. Based on the multi-source resistivity data and the predicted resistivity data, the landslide state of the landslide body is determined. Based on the landslide state, the corresponding target observation model is determined, and the multi-source resistivity data is processed through the determined target observation model to obtain target resistivity data. Based on the target resistivity data, a predicted image is generated. This integrates data from three types of devices: Wenner-Schlumberger, Wenner, and dipole-dipole. The weights are dynamically allocated according to the on-site geological conditions to solve the problems of insufficient sensitivity and data distortion of a single device. This adapts to complex landslide geological scenarios and improves accuracy.
[0052] Among them, each device 102 refers to a different acquisition device used for resistivity detection, including at least a Wenner device, a dipole-dipole device, and a Wenner-Schlumberger device. The server 104 can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.
[0053] In one exemplary embodiment, such as Figure 2 As shown, a landslide resistivity prediction imaging method is provided, which can be applied to... Figure 1 Taking the server in the example, the explanation includes the following steps S202 to S208. Wherein:
[0054] S202: Acquire the raw resistivity data collected by each device and the geological parameters of the landslide, and fuse the raw resistivity data collected by each device based on the geological parameters of the landslide to obtain multi-source resistivity data.
[0055] The raw resistivity data consists of resistivity values measured by various devices at the landslide site using different electrode running methods. Geological parameters describe the physical properties of the landslide soil and rock mass, including but not limited to porosity, fracture density, and stratigraphic distribution information. Multi-source resistivity data is obtained by fusing raw resistivity data from different devices at the data level, assigning different weights based on their reliability and adaptability.
[0056] In this embodiment, the raw resistivity data collected by each device can be fused based on the geological parameters of the landslide to obtain multi-source resistivity data. For example, the Wenner-Schlumberger device is preferentially used as the core acquisition device, and the data collected by the Wenner device and the dipole-dipole device are integrated simultaneously to construct the IMM dataset (multi-source resistivity data). Combined with the geological conditions of the soil and rock mass, such as porosity, fracture density, and stratigraphic distribution, the weights of various device datasets are dynamically allocated to achieve multi-source resistivity data fusion optimization.
[0057] S204: Obtain predicted resistivity data based on multi-source resistivity data, and determine the landslide state of the landslide body based on the multi-source resistivity data and the predicted resistivity data.
[0058] The landslide state is used to characterize whether the landslide body is currently in a stable stage or in an accelerated deformation stage before instability. Specifically, a simplified interactive multi-model algorithm is used, employing Markov transition matrices to interact with multiple models to obtain predicted resistivity data. A discriminant function is then constructed, and its value is compared with an adaptive detection threshold to determine the landslide state. The detection threshold is adaptively adjusted based on geological disturbance parameters (such as porosity and fracture density) at the landslide site to avoid false alarms or missed alarms due to geological noise.
[0059] Based on the principle of Simplified Interactive Multi-Model (SIMM) algorithm, a Markov transition matrix is constructed to complete the interaction of multiple model modes. The information content of the rate of change of monitoring data and the discrimination function are defined. Combined with geological interference parameters such as porosity and fissure of soil and rock, the false alarm probability and the missed detection probability are corrected to complete the adaptive calibration of the landslide instability detection threshold and realize the landslide state discrimination.
[0060] S206: Determine the corresponding target observation model based on the landslide state, and process the multi-source resistivity data through the determined target observation model to obtain the target resistivity data.
[0061] Specifically, when the landslide is determined to be in a stable state, a single-device observation model is used to directly perform unscented Kalman filtering to improve the real-time performance of data processing. When the landslide is determined to be in a state of accelerated deformation before sliding, an interactive multi-model mechanism is activated. The weights of each model are updated through Markov transition probabilities, and the observation data from multiple devices are fused to perform interactive filtering, outputting high-precision time-series resistivity data. For example, using unscented Kalman filtering (UKF) as the basic filtering unit, a fusion prediction model is built. When the landslide is in a stable state, a single model is used to directly perform UKF filtering to improve real-time performance. During the accelerated deformation stage before sliding, the SIMM multi-model interactive mechanism is activated. The weights of each model are updated through Markov transition probabilities, and the observation data from multiple devices are fused to perform interactive filtering, outputting high-precision time-series resistivity data.
[0062] S208: Generate a predicted image based on the target resistivity data.
[0063] Specifically, based on continuous temporal filtered data, a predictive imaging map of the spatial distribution of resistivity within the landslide body at future times is generated through inversion. This accurately depicts the evolutionary characteristics of low-resistivity anomaly zones and slip surfaces within the landslide. For example, in continuous time series, the target resistivity data obtained at each time step is mapped onto a two-dimensional or three-dimensional spatial grid of the landslide body using an inversion algorithm, forming a dynamic resistivity distribution field. Then, through visualization techniques such as interpolation and coloring, the temporal resistivity distribution data is used to generate a predictive imaging map of the spatial distribution of resistivity within the landslide body. This image clearly demonstrates the evolutionary trends of low-resistivity anomaly zones and potential slip surfaces within the landslide body over time.
[0064] The aforementioned landslide resistivity prediction imaging method acquires raw resistivity data collected by various devices and geological parameters of the landslide. Based on the geological parameters of the landslide, it fuses the raw resistivity data collected by each device to obtain multi-source resistivity data. Based on the multi-source resistivity data, it obtains predicted resistivity data and determines the landslide state based on the multi-source resistivity data and predicted resistivity data. Based on the landslide state, it determines the corresponding target observation model and processes the multi-source resistivity data through the determined target observation model to obtain target resistivity data. Based on the target resistivity data, it generates a predicted image. This method integrates data from three types of devices: Wenner-Schlumberger, Wenner, and dipole-dipole, and dynamically allocates weights according to the on-site geological conditions. This solves the problems of insufficient sensitivity and data distortion of single devices, adapts to complex landslide geological scenarios, and improves accuracy.
[0065] In some alternative embodiments, combined with Figure 3 As shown, Figure 3 This is a flowchart of step S202 in one embodiment. Step S202 involves fusing the raw resistivity data collected by each device based on the geological parameters of the landslide to obtain multi-source resistivity data, including:
[0066] S302: Based on the sensitivity distribution characteristics of each device, establish the mapping relationship between device sensitivity and spatial location.
[0067] Specifically, in combination Figure 4 As shown, Figure 4This is a diagram showing the polarization methods and measurement data point distribution of three devices in one embodiment. Due to differences in power supply and polarization methods, the sensitivity of various resistivity acquisition devices to soil and rock masses with different distribution directions varies significantly. Taking the commonly used Wenner device, dipole-dipole device, and Wenner-Schlumberger device as examples, the sensitivity ranges of the three devices are: horizontal distribution, vertical distribution, and directly below distribution geological structures, respectively. Applicable target shapes correspond to: horizontal targets, targets directly below the center, and targets extending downwards to a certain extent. The slip surface of accretionary landslides is generally located between the overburden and bedrock, within 30m below the surface. The Wenner-Schlumberger device with a small polarization is most effective for detecting accretionary landslides. The IMM dataset uses resistivity data and other monitoring technology data collected by various devices, deep neural network and nonlinear regression analysis results as input models, and assigns weights to each dataset according to different geological conditions on site. The polarization methods and measurement point distribution of the three devices are shown in the diagram. Figure 4 In this step, the sensitivity values of each device at different electrode distances and spatial locations are pre-determined through numerical simulation or field calibration experiments, thereby establishing a mapping table or mathematical function describing the relationship between "device type-spatial location-sensitivity". This mapping relationship forms the basis for subsequent weight allocation.
[0068] Taking an electrode count of 80 as an example (a=1m), the Wenner device has AM=MN=NB, the dipole-dipole device has AB=BM=MN, and the Wenner-Schlumberger device is between the Wenner and Schlumberger devices, with a spacing of 3a layers. The Schlumberger method is used for electrode running in layers 1-3, and the MN spacing becomes 3a in layers 2-6, and so on, resulting in an inverted trapezoidal cross-sectional diagram. The sensitivity distribution diagrams for the Wenner, dipole-dipole, and Wenner-Schlumberger devices are shown below. Figure 5 .
[0069] S304: Input the geological parameters and mapping relationships of the landslide body into a deep neural network, and combine it with nonlinear regression analysis to obtain the fusion weights of each device under the geological parameters.
[0070] Specifically, the mapping relationship and geological parameters of the landslide body (such as porosity and fracture density) obtained on-site are used as the input feature vectors of the deep neural network. This deep neural network is pre-trained using multi-device detection data under a large amount of known geological conditions, enabling it to learn the complex nonlinear relationship between geological parameters, device sensitivity, and optimal fusion weights. After forward propagation and nonlinear regression calculations, the output layer of the neural network outputs a fusion weight for each of the Wenner, dipole-dipole, and Wenner-Schlumberger devices. This fusion weight represents the reliability and importance of the data from each device under the current geological conditions. For example, for landslides with non-uniform porosity in the deposited layer, the neural network might assign a higher fusion weight to the Wenner-Schlumberger device, whose sensitivity is concentrated directly below.
[0071] S306: The raw resistivity data collected by each device is weighted based on the fusion weight to obtain multi-source resistivity data.
[0072] Specifically, the raw resistivity data collected by the Wenner device is multiplied by the fusion weight assigned to it, the raw data of the dipole-dipole device is multiplied by the corresponding fusion weight, and the raw data of the Wenner-Schlumberger device is multiplied by the corresponding fusion weight. Then, these three weighted data are superimposed point by point or fused at the data level according to the measurement grid to finally generate a set of multi-source resistivity data with high signal-to-noise ratio that can comprehensively reflect the electrical structure of the landslide.
[0073] In this embodiment, by dynamically allocating weights through a deep neural network, the data source can be adaptively optimized according to different on-site geological conditions, fundamentally solving the problem of detection distortion caused by the limited sensitivity of a single acquisition device, and improving the accuracy and adaptability of data acquisition.
[0074] In some optional embodiments, the process of obtaining predicted resistivity data based on multi-source resistivity data and determining the landslide state based on the multi-source resistivity data and the predicted resistivity data includes: obtaining the initial transition probabilities between the observation models corresponding to each device; inputting the initial transition probabilities into the Markov transition matrix to perform observation model interaction to obtain the predicted probabilities of each observation model; and performing weighted fusion based on the predicted probabilities of each observation model and the state prediction values of each observation model to obtain the predicted resistivity data; obtaining an information vector based on the multi-source resistivity data and the predicted resistivity data; constructing a discriminant function based on the information vector and the covariance matrix corresponding to the information vector; correcting the detection threshold based on the geological interference parameters of the landslide body; and determining the landslide state based on the comparison result between the value of the discriminant function and the detection threshold.
[0075] The MM algorithm (Interactive Multiple Model Algorithm) determines the Markov transition matrix between models, uses the filtered values of each observation model as the input values for the next interaction, and obtains a weighted interaction output as the final result. The Interactive Multiple Model Algorithm uses threshold detection to determine the rate of change (information vector) of the multi-model fusion monitoring data. The relationship between the size of the landslide and the threshold determines whether multi-model interaction is needed. When the landslide is in a stable state, the rate of change will not exceed the threshold value. To increase the real-time performance of the system monitoring, a single model (the set data acquisition device model) can be used for filtering output. When the landslide is in the imminent or accelerated deformation stage, the rate of change will exceed the threshold value. At this time, the IMM algorithm is used for interactive output. This improves the real-time performance of the system while ensuring the effectiveness of monitoring and early warning data, meeting the needs of on-site geological disaster monitoring. Information Vector Defined as:
[0076] (3-1)
[0077] in, This is the multi-source resistivity data at the current moment. This is the predicted resistivity data based on the prediction from the previous time step.
[0078] In the case of a single model, the covariance matrix of the information vector is:
[0079] (3-2)
[0080] In the formula, This is the covariance matrix of the state prediction from the previous step. For the observation matrix, To observe the noise covariance matrix, the discriminant function for determining whether deformation or slippage instability occurs within the deposited layer landslide is defined as follows:
[0081] (3-3)
[0082] The squared Mahalanobis distance of the information vector. It follows the error vector m Distribution. During monitoring, it is necessary to fully consider the possibility that the porosity and fissure of the soil and rock at the potential hazard point may cause abnormal resistivity data, and to adjust parameters such as false alarm probability and missed detection probability to determine the detection threshold.
[0083] The system pre-sets two or more observation models (e.g., Model 1 adapts to the electrical variation under steady-state conditions, and Model 2 adapts to the rapid electrical variation under slippery conditions), and assigns an initial Markov transition probability matrix to each model. This matrix describes the probability of transitioning from one model to another at the next time step. At each time step, the system uses the Markov transition matrix to perform a mixed interaction of the states of each model from the previous time step, calculating the predicted probability of each model at the current time step. Then, an unscented Kalman filter is used to perform a one-step state prediction for each model, obtaining the state prediction value for each model. Finally, using the predicted probabilities of each model as weights, the state prediction values of all models are weighted and summed to obtain a fused predicted resistivity data.
[0084] The information vector Defined as the difference between the actual observed value (i.e., multi-source resistivity data) and the predicted resistivity data at the current moment, specifically according to the formula... calculate.
[0085] The detection threshold is a threshold used to determine the state of a landslide. Considering that geological parameters such as porosity and fissures in the soil and rock mass of a landslide naturally cause fluctuations in resistivity data, this step uses these geological interference parameters to correct the pre-set theoretical detection threshold to prevent false alarms or missed detections caused by these natural fluctuations. For example, by adjusting the set values of the false alarm probability and the missed detection probability, the detection threshold can be adapted to the geological noise level of the specific engineering site. After fully considering geological factors such as the porosity and fissures of the soil and rock mass at the potential hazard point that can cause abnormal resistivity data, the settings of parameters such as the false alarm probability and the missed detection probability are corrected to determine the final detection threshold.
[0086] In some optional embodiments, the landslide state is determined based on the comparison result between the value of the discriminant function and the detection threshold, including: if the value of the discriminant function is less than or equal to the detection threshold, the landslide state is determined to be a stable state; if the value of the discriminant function is greater than the detection threshold, the landslide state is determined to be a state of imminent landslide and accelerated deformation (i.e., imminent landslide or accelerated deformation stage).
[0087] When the calculated discriminant function value does not exceed the corrected detection threshold, it indicates that the deviation between the current multi-source resistivity data and the predicted resistivity data based on the stability model is within the allowable range, and the internal electrical structure of the landslide body has not undergone significant abnormal changes. Therefore, the landslide body is currently in a stable state. In this state, there is no immediate risk of instability, and the system can maintain a low-power, high-efficiency monitoring mode. When the calculated discriminant function value exceeds the corrected detection threshold, it indicates a significant deviation between the actual observed value and the predicted value. This deviation exceeds the range that normal geological noise fluctuations can explain, reflecting rapid changes in the internal structure of the landslide body, such as the expansion of weak zones and the connection of slip surfaces. Therefore, the landslide body is currently in a state of accelerated deformation and is on the verge of collapse. The system needs to immediately switch to a high-precision multi-model interactive filtering mode and issue an early warning signal. This binary classification of landslide states is achieved through a clear threshold comparison method. The judgment rules are simple and clear, the computational load is small, and it can meet the real-time requirements of on-site geological disaster monitoring.
[0088] In the above embodiments, the prediction deviation is quantitatively evaluated by information vector and discriminant function, and the detection threshold is adaptively corrected by geological interference parameters. This can accurately distinguish between the stable state and the accelerated deformation state of landslides, avoid misjudgment caused by geological noise, and provide a reliable basis for subsequent sub-state data processing.
[0089] In some alternative embodiments, combined with Figure 6The method shown involves determining the corresponding target observation model based on the landslide state, and processing multi-source resistivity data using the determined target observation model to obtain target resistivity data. This includes: when the landslide is in a stable state, using a single-device observation model selected based on geological conditions as the target observation model, inputting multi-source resistivity data into the target observation model for filtering to obtain target resistivity data; when the landslide is in a state of accelerated deformation before sliding, mixing and interacting the filtered values and covariance of the observation models corresponding to each device at the previous time step with the Markov transition matrix to obtain interactive input values for each observation model, inputting each interactive input value into each observation model respectively, combining it with the current observation value for filtering to obtain the current filtering result, updating the probability of each observation model, and weighting and fusing the current filtering results based on the probability of each observation model to obtain the target resistivity data.
[0090] In this system, when the discriminant function value does not exceed the detection threshold, indicating that the landslide is in a stable state, the system no longer performs complex multi-model interactive calculations to improve real-time monitoring and reduce redundant consumption of computing resources. Instead, it directly selects a single observation model best suited to the current geological conditions as the target observation model. For accretionary landslides, the slip surface is generally located between the overburden and bedrock, within 30m below the surface. Selecting the observation model corresponding to the small-interval Wenner-Schlumberger device as the target observation model is most effective. Then, using multi-source resistivity data as input, the single model is rapidly filtered using the unscented Kalman filter algorithm. This process includes: performing a one-step state prediction based on the state estimate from the previous moment, calculating the covariance matrix of the predicted value, updating the state estimate using the current multi-source resistivity data as the observation value, and outputting smoothed and denoised resistivity data as the target resistivity data for that moment.
[0091] Furthermore, when the discriminant function exceeds the detection threshold, indicating that the landslide has entered the accelerated deformation stage, the system immediately initiates a complete interactive multi-model algorithm to accurately capture the rapidly changing electrical structure. First, the filtered state estimates and covariance matrices of each model (e.g., the Wenner-Schlumberger observation model, dipole-dipole observation model, and Wenner-Schlumberger observation model) from the previous time step are input into a pre-defined Markov transition matrix for mixing calculations, yielding the mixed initial state value for each model at the current time. Then, the mixed initial state value of each model is input into its corresponding observation model, and combined with the current multi-source resistivity data as observations, each model is independently updated in terms of time and measurement using an unscented Kalman filter algorithm, resulting in the filtering result and corresponding likelihood function for each model at the current time. Next, the weighted probability of each model at the current time is updated based on its likelihood function. Finally, the updated probabilities of each model are used as weighting coefficients to weight and sum the filtered results from each model, and the fused data is used as the final high-precision target resistivity data.
[0092] In practical applications, the first step is to determine the UKF filter input and detect whether the data exceeds the set threshold based on the false alarm probability and the discriminant function.
[0093] Step 2: When the data change rate does not exceed the set threshold, the data from the acquisition device, set according to geological conditions, is used as input, and filtering is performed directly according to the original filtering parameters. When the data change rate exceeds the set threshold, the initial prediction state is used. Initial covariance matrix and initial observations As input, the Markov transition probabilities of each model are set according to the interactive multi-model algorithm, and the filtered value at time k is output interactively. and .
[0094] The optimal application scenario for landslide prediction based on simplified interactive multi-model and unscented Kalman filtering technology is the monitoring and early warning process of landslide geological hazards in deposited layers. When the composition of the upper deposits is complex and the porosity is uneven, the resistivity imaging data obtained by this method can be corrected by combining multiple methods such as layered soil moisture monitoring and borehole sampling.
[0095] In this embodiment, the real-time monitoring is ensured by using single-model filtering in the stable state, and the accuracy of the early warning is ensured by starting multi-model interaction in the near-slip state. The two work together to achieve a balance between real-time monitoring and accurate early warning, avoiding the problems of large redundant computation in the stable stage and weak anti-interference ability in the near-slip stage of traditional single models.
[0096] In some optional embodiments, the method further includes: acquiring independent ground truth data obtained through borehole sampling and stratified soil moisture monitoring; and iteratively adjusting the fusion weights of the raw resistivity data collected by each device, the state equation parameters of each observation model, and the noise covariance parameters in the filtering process based on the predicted image and the independent ground truth data.
[0097] In this embodiment, a three-level closed-loop system of "measured benchmark - algorithm prediction - independent truth value verification" is established. Based on the on-site measured resistivity imaging, independent truth value data is obtained through borehole sampling and layered soil moisture content monitoring. The algorithm model weights, state equation parameters, and noise covariance parameters are quantitatively verified and corrected to optimize the algorithm prediction accuracy and engineering reliability.
[0098] Specifically, boreholes are deployed in the monitoring area, and soil and rock samples are periodically extracted for indoor resistivity testing to obtain the true resistivity values of the soil and rock mass at different depths. Simultaneously, soil moisture sensors are installed at different depths within the boreholes to continuously monitor the volumetric moisture content data of each layer. Based on Archie's formula or a field-calibrated resistivity-moisture content relationship model, the moisture content data is converted into corresponding resistivity reference values. These data, obtained directly or indirectly through physical means, are used as "true values" independent of the inversion imaging results. When the upper sedimentary composition is complex and the porosity is uneven, the effect of combining layered soil moisture content monitoring and borehole sampling to correct the resistivity imaging data obtained by this method is particularly significant.
[0099] The resistivity values at each spatial location in the predicted imaging image generated in step S208 are compared point-by-point with independent ground truth data such as borehole sampling test values and water content converted resistivity values at the same spatiotemporal location. The quantitative error between the two is calculated, such as root mean square error or mean absolute percentage error. Based on the direction and magnitude of this error, a feedback optimization algorithm (such as gradient descent, extended Kalman filter, or particle filter algorithm) is used to adjust the key parameters in the entire processing flow in reverse. These key parameters include, but are not limited to: the multi-source data fusion weights output by the deep neural network in step S304, the coefficient matrix of the state transition equations of each observation model in the unscented Kalman filter, the process noise covariance matrix, and the observation noise covariance matrix. Through this closed-loop iterative process of "measured benchmark - algorithm prediction - independent ground truth verification," model errors and environmental disturbances are continuously corrected until the fitting error between the predicted image and the independent ground truth data converges to a preset engineering allowable range, thereby significantly improving the prediction accuracy and engineering reliability of the algorithm.
[0100] In this embodiment, the algorithm model is closed-loop corrected by introducing independent true data obtained from borehole sampling and layered moisture content monitoring. This enables the fusion weights, state equation parameters, and noise covariance parameters to be dynamically optimized according to actual field conditions. This solves the problems of imperfect verification system and low engineering reliability of traditional algorithms, and ensures that the algorithm can be truly implemented in the monitoring scenario of complex deposited layer landslides.
[0101] In the above embodiments, an instability detection threshold is constructed based on the SIMM algorithm, and the probability of false alarms and missed detections is corrected by combining geological parameters. This fundamentally solves the problem that traditional technologies cannot accurately determine the real-time state of landslides, avoiding the situation where inaccurate state determination leads to deviations in subsequent data processing and insufficient reliability of early warnings, thus providing accurate support for state-specific data processing. A state-specific processing mode is adopted, employing a single-model UKF fast filtering for stable states and a multi-model interactive optimization of the SIMM for landslide-prone states. This mode integrates observation data from multiple devices to output high-precision time-series data, fundamentally solving the problem that traditional single-filter model data processing modes are fixed and cannot balance real-time performance and accuracy. It avoids the problems of large redundant computational loads and poor real-time monitoring in stable stages, and weak anti-interference capabilities and susceptibility to false alarms and missed warnings in landslide-prone stages. Based on continuous time-series filtered data inversion and predictive imaging, this fundamentally solves the problem that traditional technologies only support static imaging and lack forward-looking prediction capabilities. It avoids monitoring lag, the inability to capture the evolution trend of weak zones and slip surfaces within landslides in advance, and the inability to meet the needs of early warning and prediction for landslides, thus providing a sufficient window of opportunity for landslide response. A three-level closed-loop verification system of "measured benchmark - algorithm prediction - independent truth value verification" was established. Independent truth value data were obtained through borehole drilling and moisture content monitoring. This fundamentally solved the problems of imperfection and low engineering reliability of traditional algorithm verification systems, and avoided the situation where algorithm theory is disconnected from engineering practice, the prediction accuracy cannot be quantified, and it is difficult to implement in complex landslide monitoring scenarios.
[0102] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.
[0103] Based on the same inventive concept, this application also provides a landslide resistivity prediction imaging device for implementing the landslide resistivity prediction imaging method described above. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations of one or more landslide resistivity prediction imaging device embodiments provided below can be found in the limitations of the landslide resistivity prediction imaging method described above, and will not be repeated here.
[0104] In one exemplary embodiment, such as Figure 7 As shown, a landslide resistivity prediction imaging device is provided, comprising: a data acquisition module 701, a state determination module 702, an observation module 703, and a generation module 704, wherein:
[0105] The data acquisition module 701 is used to acquire the raw resistivity data collected by each device and the geological parameters of the landslide body, and to fuse the raw resistivity data collected by each device based on the geological parameters of the landslide body to obtain multi-source resistivity data.
[0106] The state determination module 702 is used to obtain predicted resistivity data based on the multi-source resistivity data, and to determine the landslide state of the landslide body based on the multi-source resistivity data and the predicted resistivity data.
[0107] The observation module 703 is used to determine the corresponding target observation model based on the landslide state, and to process the multi-source resistivity data through the determined target observation model to obtain target resistivity data.
[0108] The generation module 704 is used to generate a predicted image based on the target resistivity data.
[0109] In some optional embodiments, the data acquisition module 701 is specifically used to establish a mapping relationship between device sensitivity and spatial location based on the sensitivity distribution characteristics of each device; input the geological parameters of the landslide body and the mapping relationship into a deep neural network, and combine nonlinear regression analysis to obtain the fusion weight of each device under the geological parameters; and weight the raw resistivity data collected by each device based on the fusion weight to obtain multi-source resistivity data.
[0110] In some optional embodiments, the state determination module 702 is specifically used to obtain the initial transition probabilities between the observation models corresponding to each device, input the initial transition probabilities into the Markov transition matrix for observation model interaction, obtain the predicted probabilities of each observation model, and perform weighted fusion based on the predicted probabilities of each observation model and the state prediction values of each observation model to obtain predicted resistivity data; obtain an information vector based on the multi-source resistivity data and the predicted resistivity data; construct a discriminant function based on the information vector and the covariance matrix corresponding to the information vector; correct the detection threshold based on the geological interference parameters of the landslide body; and determine the landslide state based on the comparison result between the value of the discriminant function and the detection threshold.
[0111] In some optional embodiments, the state determination module 702 is specifically used to determine that the landslide state is a stable state when the value of the discrimination function is less than or equal to the detection threshold; and to determine that the landslide state is an accelerated deformation state on the verge of landslide when the value of the discrimination function is greater than the detection threshold.
[0112] In some optional embodiments, the observation module 703 is specifically used to: when the landslide is in a stable state, use a single-device observation model selected according to geological conditions as the target observation model, input the multi-source resistivity data into the target observation model for filtering, and obtain target resistivity data; when the landslide is in a state of accelerated deformation and imminent landslide, mix and interact the filtered values and covariance of the observation models corresponding to each device at the previous moment into the Markov transition matrix to obtain interactive input values for each observation model, input each interactive input value into each observation model respectively, filter it in combination with the observation value at the current moment to obtain the filtering result at the current moment, update the probability of each observation model, and perform weighted fusion of the filtering results at the current moment based on the probability of each observation model to obtain target resistivity data.
[0113] In some optional embodiments, the above-mentioned apparatus further includes: an adjustment module specifically used to acquire independent ground truth data obtained through borehole sampling and stratified soil moisture content monitoring; and based on the predicted image and the independent ground truth data, iteratively adjusting the fusion weight of the raw resistivity data collected by each device, the state equation parameters of each observation model, and the noise covariance parameters in the filtering process.
[0114] Each module in the aforementioned landslide resistivity prediction imaging device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the computer device's memory as software, so that the processor can call and execute the corresponding operations of each module.
[0115] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 8 As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores the data involved in the aforementioned methods. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements a landslide resistivity prediction imaging method.
[0116] Those skilled in the art will understand that Figure 8 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.
[0117] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.
[0118] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.
[0119] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.
[0120] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.
[0121] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.
[0122] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.
Claims
1. A landslide resistivity prediction imaging method, characterized in that, The method includes: The raw resistivity data collected by each device and the geological parameters of the landslide body are obtained, and the raw resistivity data collected by each device are fused based on the geological parameters of the landslide body to obtain multi-source resistivity data. Based on the multi-source resistivity data, predicted resistivity data is obtained, and based on the multi-source resistivity data and the predicted resistivity data, the landslide state of the landslide body is determined; Based on the landslide state, a corresponding target observation model is determined, and the multi-source resistivity data is processed using the determined target observation model to obtain target resistivity data; A predicted image is generated based on the target resistivity data.
2. The method according to claim 1, characterized in that, The original resistivity data collected by each device is fused based on the geological parameters of the landslide body to obtain multi-source resistivity data, including: Based on the sensitivity distribution characteristics of each device, a mapping relationship between device sensitivity and spatial location is established; The geological parameters of the landslide and the mapping relationship are input into a deep neural network, and combined with nonlinear regression analysis, the fusion weights of each device under the geological parameters are obtained. The raw resistivity data collected by each device are weighted based on the fusion weight to obtain multi-source resistivity data.
3. The method according to claim 1, characterized in that, The step of obtaining predicted resistivity data based on the multi-source resistivity data, and determining the landslide state of the landslide body based on the multi-source resistivity data and the predicted resistivity data, includes: The initial transition probabilities between the observation models corresponding to each device are obtained. The initial transition probabilities are input into the Markov transition matrix to perform observation model interaction, thereby obtaining the prediction probability of each observation model. The prediction probability of each observation model and the state prediction value of each observation model are weighted and fused to obtain the predicted resistivity data. An information vector is obtained based on the multi-source resistivity data and the predicted resistivity data; A discriminant function is constructed based on the information vector and the covariance matrix corresponding to the information vector. The detection threshold is corrected based on the geological disturbance parameters of the landslide body; The landslide state is determined by comparing the value of the discriminant function with the detection threshold.
4. The method according to claim 3, characterized in that, Determining the landslide state based on the comparison result between the value of the discriminant function and the detection threshold includes: If the value of the discrimination function is less than or equal to the detection threshold, the landslide state is determined to be a stable state. If the value of the discrimination function is greater than the detection threshold, the landslide state is determined to be an accelerated deformation state on the verge of landslide.
5. The method according to claim 1, characterized in that, The process of determining the corresponding target observation model based on the landslide state, and processing the multi-source resistivity data using the determined target observation model to obtain target resistivity data, includes: When the landslide is in a stable state, the single-device observation model selected according to the geological conditions is used as the target observation model. The multi-source resistivity data is input into the target observation model for filtering to obtain the target resistivity data. When the landslide is in a state of accelerated deformation and imminent landslide, the filtered values and covariance of the observation models corresponding to each device at the previous time step are mixed and interacted with to obtain the interaction input values of each observation model. Each interaction input value is then input into each observation model, and the current observation value is combined with the current observation value to obtain the current filtering result. The probabilities of each observation model are then updated. Based on the probabilities of each observation model, the current filtering results are weighted and fused to obtain the target resistivity data.
6. The method according to any one of claims 1 to 5, characterized in that, The method further includes: Obtain independent true data through borehole sampling and stratified soil moisture monitoring; Based on the predicted image and the independent ground truth data, the fusion weights of the raw resistivity data collected by each device, the state equation parameters of each observation model, and the noise covariance parameters in the filtering process are iteratively adjusted.
7. A landslide resistivity prediction imaging device, characterized in that, The device includes: The data acquisition module is used to acquire the raw resistivity data collected by each device and the geological parameters of the landslide body, and to fuse the raw resistivity data collected by each device based on the geological parameters of the landslide body to obtain multi-source resistivity data. The state determination module is used to obtain predicted resistivity data based on the multi-source resistivity data, and to determine the landslide state of the landslide body based on the multi-source resistivity data and the predicted resistivity data. The observation module is used to determine the corresponding target observation model based on the landslide state, and to process the multi-source resistivity data through the determined target observation model to obtain the target resistivity data; A generation module is used to generate a predicted image based on the target resistivity data.
8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.
10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.