Landslide monitoring and early warning method and system based on multi-source data fusion

Abstract

The invention discloses a landslide monitoring and early warning method and system based on multi-source data fusion. The method comprises the steps that position data, displacement data, inclinometry data, water level data, rainfall data and crack data of a target slope body are acquired; dividing the target slope body into a plurality of three-dimensional grids; determining the attribute of each three-dimensional grid according to the position data; the zero value is interpolated into the three-dimensional grid with the attribute of no slope body, the displacement data, the rainfall data and the crack data are interpolated into the three-dimensional grid with the attribute of the slope body surface, and the inclinometry data and the water level data are interpolated into the three-dimensional grid with the attribute of the interior of the slope body to form a target tensor; and inputting the target tensor into the 3D-CNN + LSTM model, and outputting to obtain a landslide early warning result. According to the method, the slope body is subjected to three-dimensional gridding, monitoring point data is expanded from points to surfaces, the method can be conveniently applied to a 3D-CNN + LSTM model, the data size can be reduced, calculation can be simplified, cost can be reduced, and generalization of the model is facilitated.

Description

Technical Field

[0001] The present invention relates to the technical field of geological disaster monitoring, and more particularly to a landslide monitoring and early warning method and system based on multi-source data fusion. Background Art

[0002] Traditional highway slope monitoring makes independent judgments and comprehensive decisions based on multi-source sensor data such as Global Navigation Satellite System (GNSS) receivers, inclinometers, and rain gauges. That is, each type of sensor data has its own early warning method, and finally, manual comprehensive judgment is carried out. Due to the point-like distribution of monitoring sensors, traditional monitoring methods have problems such as limited monitoring range, slow information update, and insufficient adaptability to complex terrain environments.

[0003] To solve the above problems, a method of using a 3D-Convolutional Neural Networks (3D-CNN) + Long Short-Term Memory (LSTM) model for landslide prediction has been proposed in the prior art. However, due to the small number of displacement monitoring sensors, the direct application to the model has an unsatisfactory effect. Therefore, in existing cases, multi-source environmental data (topography, geology, meteorology), remote sensing images, etc. are usually combined, resulting in a large amount of data, complex calculations, and high costs. Summary of the Invention

[0004] The present invention aims to at least solve one of the technical problems existing in the prior art. For this purpose, the present invention proposes a landslide monitoring and early warning method based on multi-source data fusion. By three-dimensionally meshing the slope body, the monitoring point data is expanded from points to surfaces, which is convenient for application to the 3D-CNN + LSTM model, can reduce the amount of data, simplify calculations, reduce costs, and is beneficial to the generalization of the model.

[0005] The present invention also provides a landslide monitoring and early warning system based on multi-source data fusion, a control device for executing the above-mentioned landslide monitoring and early warning method based on multi-source data fusion, and a computer-readable storage medium.

[0006] According to the landslide monitoring and early warning method based on multi-source data fusion according to the first aspect embodiment of the present invention, the method includes: Obtain the position data, displacement data, inclinometer data, water level data, rainfall data, and crack data of the target slope body; Divide the target slope body into W×H×D three-dimensional grids, where W represents the width of the target slope body, represents the height of the target slope body, and D represents the depth of the target slope body; Determine the attributes of each of the three-dimensional grids according to the position data, where the attributes include non-slope body, slope surface, and slope interior; Interpolate zero values into each three-dimensional grid with the attribute of non-slope body, interpolate the displacement data, the rainfall data, and the crack data into each three-dimensional grid with the attribute of slope surface, and interpolate the inclinometer data and the water level data into each three-dimensional grid with the attribute of slope interior to form a target tensor; Input the target tensor into a 3D-CNN+LSTM model to output a landslide warning result.

[0007] The landslide monitoring and warning method based on multi-source data fusion according to the embodiments of the present invention has at least the following beneficial effects: By dividing the target slope into W×H×D three-dimensional grids, determining the attributes of each three-dimensional grid according to the position data, interpolating zero values into each three-dimensional grid with the attribute of non-slope body, interpolating the displacement data, the rainfall data, and the crack data into each three-dimensional grid with the attribute of slope surface, and interpolating the inclinometer data and the water level data into each three-dimensional grid with the attribute of slope interior to form a target tensor as the input of the 3D-CNN+LSTM model, interpolation processing is performed on the attributes of each three-dimensional grid and the corresponding effective sensor monitoring data, expanding the monitoring point data from points to surfaces, facilitating application to the 3D-CNN+LSTM model, capable of reducing the data volume, simplifying the calculation, reducing the cost, and being beneficial to the generalization of the model.

[0008] According to some embodiments of the present invention, the displacement data includes at least one first eigenvalue, the inclinometer data includes at least one second eigenvalue, the water level data includes at least one third eigenvalue, the rainfall data includes at least one fourth eigenvalue, and the crack data includes at least one fifth eigenvalue; The step of interpolating zero values into each three-dimensional grid with the attribute of non-slope body, interpolating the displacement data, the rainfall data, and the crack data into each three-dimensional grid with the attribute of slope surface, and interpolating the inclinometer data and the water level data into each three-dimensional grid with the attribute of slope interior to form a target tensor includes: Interpolate zero values into the three-dimensional grid where each attribute is the slope-free body, interpolate the at least one first eigenvalue, the at least one fourth eigenvalue, and the at least one fifth eigenvalue into the three-dimensional grid where each attribute is the slope surface, and interpolate the at least one second eigenvalue and the at least one third eigenvalue into the three-dimensional grid where each attribute is the interior of the slope body to form the target tensor of T×W×H×D×C, where T represents the time step and C represents the total number of eigenvalues of the first eigenvalue, the second eigenvalue, the third eigenvalue, the fourth eigenvalue, and the fifth eigenvalue.

[0009] According to some embodiments of the present invention, the displacement data includes four of the first eigenvalues, and the four first eigenvalues are respectively: the horizontal displacement in the first slope direction, the horizontal displacement velocity, the elevation displacement, and the elevation displacement velocity; the inclinometer data includes two of the second eigenvalues, and the two second eigenvalues are respectively: the horizontal displacement in the second slope direction and the displacement acceleration; the water level data includes one of the third eigenvalues, and the third eigenvalue is: the water level depth; the rainfall data includes one of the fourth eigenvalues, and the fourth eigenvalue is: the rainfall; the crack data includes two of the fifth eigenvalues, and the two fifth eigenvalues are respectively: the crack width and the crack change velocity.

[0010] According to some embodiments of the present invention, the interpolating zero values into the three-dimensional grid where each attribute is the slope-free body, interpolating the at least one first eigenvalue, the at least one fourth eigenvalue, and the at least one fifth eigenvalue into the three-dimensional grid where each attribute is the slope surface, and interpolating the at least one second eigenvalue and the at least one third eigenvalue into the three-dimensional grid where each attribute is the interior of the slope body to form the target tensor of T×W×H×D×C, where T represents the time step and C represents the total number of eigenvalues of the first eigenvalue, the second eigenvalue, the third eigenvalue, the fourth eigenvalue, and the fifth eigenvalue, includes: Interpolate zero values into the three-dimensional grid where each attribute is the slope-free body, interpolate the horizontal displacement in the first slope direction, the horizontal displacement velocity, the elevation displacement, the elevation displacement velocity, the rainfall, the crack width, and the crack change velocity into the three-dimensional grid where each attribute is the slope surface, and interpolate the horizontal displacement in the second slope direction, the displacement acceleration, and the water level depth into the three-dimensional grid where each attribute is the interior of the slope body to form the target tensor of T×W×H×D×C, where T represents the time step and C is equal to 10.

[0011] According to some embodiments of the present invention, after obtaining the position data, displacement data, inclinometer data, water level data, rainfall data, and crack data of the target slope body, it further includes: Normalize the displacement data, the inclinometer data, the water level data, the rainfall data and the crack data to eliminate the dimensional differences.

[0012] According to some embodiments of the present invention, the target slope is further provided with an intelligent camera, and the landslide warning result is a warning level; The landslide monitoring and warning method based on multi-source data fusion further includes: Dynamically adjust the sensitivity of the intelligent camera based on the warning level, wherein the warning level and the sensitivity are in a positive correlation relationship.

[0013] According to some embodiments of the present invention, the interpolation algorithm of the three-dimensional grid adopts the Kriging interpolation algorithm.

[0014] According to the landslide monitoring and warning system based on multi-source data fusion in the second aspect embodiment of the present invention, the system includes: A data acquisition unit, configured to acquire the position data, displacement data, inclinometer data, water level data, rainfall data and crack data of the target slope; A grid division unit, configured to divide the target slope into W×H×D three-dimensional grids, where W represents the width of the target slope, represents the height of the target slope, and D represents the depth of the target slope; A grid attribute determination unit, configured to determine the attribute of each three-dimensional grid according to the position data, and the attribute includes no slope, slope surface and slope interior; A grid interpolation unit, configured to interpolate zero values into each three-dimensional grid with the attribute of no slope, interpolate the displacement data, the rainfall data and the crack data into each three-dimensional grid with the attribute of slope surface, and interpolate the inclinometer data and the water level data into each three-dimensional grid with the attribute of slope interior to form a target tensor; A landslide warning unit, configured to input the target tensor into a 3D-CNN+LSTM model and output a landslide warning result.

[0015] The landslide monitoring and warning system based on multi-source data fusion according to the embodiments of the present invention has at least the following beneficial effects: By dividing the target slope into W×H×D three-dimensional grids, determining the attributes of each three-dimensional grid according to the position data, interpolating zero values into each three-dimensional grid with no slope body attribute, interpolating displacement data, rainfall data and crack data into each three-dimensional grid with slope surface attribute, and interpolating inclinometer data and water level data into each three-dimensional grid with slope interior attribute, a target tensor is formed as the input of the 3D-CNN+LSTM model. Interpolation processing is performed on the attributes of each three-dimensional grid and the corresponding effective sensor monitoring data, expanding the monitoring point data from points to surfaces, facilitating its application to the 3D-CNN+LSTM model, reducing the data volume, simplifying the calculation, reducing costs, and being beneficial to the generalization of the model.

[0016] The control device according to the third aspect embodiment of the present invention includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, it implements the landslide monitoring and early warning method based on multi-source data fusion as described in the first aspect embodiment above. Since the control device adopts all the technical solutions of the landslide monitoring and early warning method based on multi-source data fusion in the above embodiment, it at least has all the beneficial effects brought by the technical solutions of the above embodiment.

[0017] The computer-readable storage medium according to the fourth aspect embodiment of the present invention stores computer-executable instructions for executing the landslide monitoring and early warning method based on multi-source data fusion as described in the first aspect embodiment above. Since the computer-readable storage medium adopts all the technical solutions of the landslide monitoring and early warning method based on multi-source data fusion in the above embodiment, it at least has all the beneficial effects brought by the technical solutions of the above embodiment.

[0018] Other features and advantages of the present invention will be described in the subsequent description, and some will become obvious from the description or be understood by implementing the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The above and / or additional aspects and advantages of the present invention will become obvious and easy to understand from the description of the embodiments in conjunction with the following drawings, where: Figure 1 is a flowchart of the landslide monitoring and early warning method based on multi-source data fusion according to an embodiment of the present invention; Figure 2 is a schematic diagram of the target slope according to an embodiment of the present invention; Figure 3 is a schematic diagram of the interpolation of the three-dimensional grid according to an embodiment of the present invention; Figure 4 is a schematic diagram of the 3D-CNN+LSTM model according to an embodiment of the present invention; Figure 5 It is a schematic diagram of the system logic architecture of a landslide monitoring and early warning system based on multi-source data fusion according to an embodiment of the present invention. Detailed implementation manners

[0020] The embodiments of the present invention will be described in detail below. Examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention and should not be construed as a limitation of the present invention.

[0021] In the description of the present invention, if the first, second, etc. are described, it is only for the purpose of distinguishing technical features and should not be construed as indicating or implying relative importance or implicitly indicating the quantity of the indicated technical features or implicitly indicating the sequence of the indicated technical features.

[0022] In the description of the present invention, it should be understood that the orientation or positional relationship indicated by terms such as up, down, etc. is based on the orientation or positional relationship shown in the accompanying drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as a limitation of the present invention.

[0023] In the description of the present invention, it should be noted that unless otherwise clearly defined, terms such as setting, installation, connection, etc. should be understood in a broad sense, and those skilled in the art can reasonably determine the specific meanings of the above terms in the present invention in combination with the specific content of the technical solution.

[0024] Next, in combination with Figures 1 to 5 A clear and complete description of the landslide monitoring and early warning method based on multi-source data fusion according to the embodiments of the present invention will be given. Obviously, the following described embodiments are some embodiments of the present invention, not all embodiments.

[0025] Referring to Figures 1 to 5 , Figure 1 It is a flowchart of the landslide monitoring and early warning method based on multi-source data fusion according to an embodiment of the present invention; Figure 2 It is a schematic diagram of the target slope according to an embodiment of the present invention; Figure 3 It is an interpolation schematic diagram of a three-dimensional grid according to an embodiment of the present invention; Figure 4 It is a schematic diagram of a 3D-CNN + LSTM model according to an embodiment of the present invention; Figure 5 It is a schematic diagram of the system logic architecture of a landslide monitoring and early warning system based on multi-source data fusion according to an embodiment of the present invention.

[0026] According to the landslide monitoring and early warning method based on multi-source data fusion according to the first aspect embodiment of the present invention, the method includes: Obtain the position data, displacement data, inclinometer data, water level data, rainfall data, and crack data of the target slope body; Divide the target slope body into W×H×D three-dimensional grids, where W represents the width of the target slope body, H represents the height of the target slope body, and D represents the depth of the target slope body; Determine the attributes of each three-dimensional grid according to the position data, and the attributes include no slope body, slope body surface, and slope body interior; Interpolate zero values into each three-dimensional grid with the attribute of no slope body, interpolate the displacement data, rainfall data, and crack data into each three-dimensional grid with the attribute of slope body surface, and interpolate the inclinometer data and water level data into each three-dimensional grid with the attribute of slope body interior to form a target tensor; Input the target tensor into the 3D-CNN+LSTM model to output the landslide warning result.

[0027] It can be understood that the position data includes longitude, latitude, and altitude, and the attributes of each three-dimensional grid can be determined according to the longitude, latitude, and altitude.

[0028] Reference Figure 2 , the displacement data is detected by a GNSS receiver. The inclinometer data is detected by an inclinometer. The water level data is detected by a water level gauge. The rainfall data is detected by a rain gauge. The crack data is detected by a crack gauge. It should be noted that the installation positions of the GNSS receiver, inclinometer, water level gauge, rain gauge, and crack gauge can be set according to actual needs, and the setting principle of the positions is the prior art known to those skilled in the art, and will not be elaborated here.

[0029] The selection of the grid unit of the three-dimensional grid must comprehensively consider the calculation amount, measurement accuracy, and model generalization. For example, if the distance between the inclinometer probe heads is 2 meters, a target slope body with a unit of 2 meters and 50×40×20 can be constructed.

[0030] Such as Figure 2 shown, the width of the target slope body is the distance from one side of the slope body where the crack gauge is located to the other side of the slope body where the GNSS receiver is located; the height of the target slope body is the vertical height from the slope top to the slope bottom; the depth of the target slope body is the projection distance of the slope surface on the ground.

[0031] In some embodiments of the present invention, after obtaining the position data, displacement data, inclinometer data, water level data, rainfall data, and crack data of the target slope body, it further includes: Perform outlier rejection processing on the position data, displacement data, inclinometer data, water level data, rainfall data, and crack data.

[0032] An outlier refers to an extreme value that significantly deviates from most of the observed values in a dataset, which may be caused by measurement errors or real but rare events. Outliers can distort statistical measures (such as the mean and standard deviation), resulting in a distorted distribution of the standardized data, affecting the effect of model training, and reducing model performance. Methods for removing outliers include the quartile method, the isolation forest method, etc. The specific principles and processes are prior arts known to those skilled in the art and will not be elaborated herein.

[0033] In some embodiments of the present invention, referring to Figure 4 , the 3D-CNN+LSTM model includes a 3D-CNN network and an LSTM network; inputting the target tensor into the 3D-CNN+LSTM model, the landslide warning result is output, including: Inputting the target tensor into the 3D-CNN network, a feature matrix is output; Inputting the feature matrix into the LSTM network, the landslide warning result is output.

[0034] The 3D-CNN network has excellent spatial feature learning ability and can be used to capture the spatial correlation relationship between sensor data at different positions in slope monitoring. CNN has been widely used in image recognition. CNN consists of several convolutional layers and pooling layers. The role of convolution is to extract local features of the image. By sliding a window (convolution kernel) to match each pixel on the input feature map, it automatically learns the local features of the image (such as edges, corners, textures, etc.); the main role of pooling is to reduce the dimension and compress the feature map. By taking the maximum or average value within the region, it reduces the spatial dimension, reduces the amount of calculation, and at the same time makes the model robust to position changes. A color image is actually a 2D 3-channel (RGB color) pixel matrix, while the monitoring grid composed of slope monitoring sensors is a 3D N-channel matrix. Referring to Figure 3 , this matrix implies the sliding characteristics of the slope body, so CNN can be used to learn such characteristics. Taking a matrix of H×W×C as an input example, the convolutional layer uses a convolutional kernel of K×K, the stride is S, and the padding is P. The output of the convolutional layer is a matrix of (H−K+2P)×(W−K+2P)×C. This matrix will be compressed by the pooling layer. If the input is H×W×C, the pooling kernel is N×N, and the stride S=N, the output matrix is: (H−N+1)×(W−N+1)×C.

[0035] The LSTM network is a variant of the recurrent neural network (RNN), which is particularly suitable for dealing with long-term dependence problems in time series data and has excellent time feature learning ability. It is usually used for predicting changes in time series data. The input of the LSTM network is a matrix of T×M, where T is the number of time steps and M is the feature dimension of each time step.

[0036] The complete structure of the 3D-CNN+LSTM model is as follows: Input layer: T×W×H×D×C, where T is the number of time steps, W represents the width of the target slope, H represents the height of the target slope, and D represents the depth of the target slope; Convolutional layer: Convolution kernel K×K; Pooling layer: Pooling kernel N×N; Flattening: After pooling, the output matrix is T×W’×H’×D’×C, which becomes a one-dimensional T×M after flattening, where M = W’×H’×D’×C; LSTM layer: Input T×M, output 1×Y, where Y is the warning level, that is, the number of risk levels. For example, if 4 risk levels are designed, the output is designed as a one-dimensional matrix of 1×4.

[0037] Combining the 3D-CNN network and the LSTM network can capture the spatial and temporal features of the three-dimensional sliding of the slope, improving the accuracy of landslide risk warning.

[0038] It should be noted that the focus of the improvement of the present invention lies in the processing of the input data of the 3D-CNN+LSTM model. The specific principle of the 3D-CNN+LSTM model is the prior art known to those skilled in the art and will not be elaborated here.

[0039] The landslide warning result is the warning level. The higher the warning level, the higher the landslide risk.

[0040] As Figure 2 shown, the number of monitoring sensors (GNSS receivers, inclinometers, water level gauges, rain gauges, and crack gauges) of the target slope is small and it is single-point detection. However, the monitoring grid composed of slope monitoring sensors in the 3D-CNN+LSTM model is a 3D N-channel matrix. Directly applying it to the 3D-CNN+LSTM model will result in the inability to fill data for some monitoring grids, and the model effect is not ideal. The present invention divides the target slope into W×H×D three-dimensional grids, determines the attributes of each three-dimensional grid according to the position data, interpolates zero values into each three-dimensional grid with the attribute of no slope body, interpolates displacement data, rain data, and crack data into each three-dimensional grid with the attribute of the slope surface, and interpolates inclinometer data and water level data into each three-dimensional grid with the attribute of the slope interior to form a target tensor as the input of the 3D-CNN+LSTM model. Interpolation processing is carried out for the attributes of each three-dimensional grid and the corresponding effective sensor monitoring data, expanding the monitoring point data from points to surfaces, facilitating application to the 3D-CNN+LSTM model, being able to reduce the data volume, simplify the calculation, reduce the cost, and be beneficial to the generalization of the model.

[0041] The landslide monitoring and early warning method based on multi-source data fusion according to the embodiments of the present invention divides the target slope into W×H×D three-dimensional grids, determines the attributes of each three-dimensional grid according to the position data, interpolates zero values into each three-dimensional grid with the attribute of no slope body, interpolates displacement data, rainfall data and crack data into each three-dimensional grid with the attribute of slope surface, interpolates inclinometer data and water level data into each three-dimensional grid with the attribute of slope interior, forms a target tensor as the input of the 3D-CNN+LSTM model, performs interpolation processing on the attributes of each three-dimensional grid and the corresponding effective sensor monitoring data, expands the monitoring point data from points to surfaces, is convenient for applying to the 3D-CNN+LSTM model, can reduce the data volume, simplify the calculation, reduce the cost, and is beneficial to the generalization of the model.

[0042] In some embodiments of the present invention, the displacement data includes at least one first eigenvalue, the inclinometer data includes at least one second eigenvalue, the water level data includes at least one third eigenvalue, the rainfall data includes at least one fourth eigenvalue, and the crack data includes at least one fifth eigenvalue; Interpolating zero values into each three-dimensional grid with the attribute of no slope body, interpolating displacement data, rainfall data and crack data into each three-dimensional grid with the attribute of slope surface, and interpolating inclinometer data and water level data into each three-dimensional grid with the attribute of slope interior to form a target tensor, including: Interpolating zero values into each three-dimensional grid with the attribute of no slope body, interpolating at least one first eigenvalue, at least one fourth eigenvalue and at least one fifth eigenvalue into each three-dimensional grid with the attribute of slope surface, and interpolating at least one second eigenvalue and at least one third eigenvalue into each three-dimensional grid with the attribute of slope interior to form a target tensor of T×W×H×D×C, where T represents the time step and C represents the total number of eigenvalues of the first eigenvalue, the second eigenvalue, the third eigenvalue, the fourth eigenvalue and the fifth eigenvalue.

[0043] Each three-dimensional grid has C eigenvalues, and C is equal to the sum of the eigenvalues of each type of sensor, that is, the total number of eigenvalues of the first eigenvalue, the second eigenvalue, the third eigenvalue, the fourth eigenvalue and the fifth eigenvalue.

[0044] For a three-dimensional grid without a slope body, the monitoring values of all sensors are invalid data for it, so the eigenvalue is filled with 0; for the three-dimensional grid on the slope surface, the displacement data, rainfall data, and crack data are all reflected on the slope surface, which belong to the effective sensor monitoring data of the three-dimensional grid on the slope surface. Therefore, at least one first eigenvalue, at least one fourth eigenvalue, and at least one fifth eigenvalue are interpolated into each three-dimensional grid with the attribute of the slope surface; for the three-dimensional grid inside the slope body, the inclinometer data and water level data are all reflected inside the slope body, which belong to the effective sensor monitoring data of the three-dimensional grid inside the slope body. Therefore, at least one second eigenvalue and at least one third eigenvalue are interpolated into each three-dimensional grid with the attribute of the slope body interior.

[0045] In some embodiments of the present invention, the displacement data includes four first eigenvalues, which are respectively: the horizontal displacement in the first slope direction, the horizontal displacement speed, the elevation displacement, and the elevation displacement speed; the inclinometer data includes two second eigenvalues, which are respectively: the horizontal displacement in the second slope direction and the displacement acceleration; the water level data includes one third eigenvalue, which is: the water level depth; the rainfall data includes one fourth eigenvalue, which is: the rainfall; the crack data includes two fifth eigenvalues, which are respectively: the crack width and the crack change speed.

[0046] It should be noted that the horizontal displacement in the first slope direction and the horizontal displacement in the second slope direction belong to the same type of data monitored by two different sensors. Through the data processing method of the present invention, their fusion can increase the training data volume of the 3D-CNN+LSTM model and improve the prediction accuracy of the 3D-CNN+LSTM model.

[0047] In some embodiments, the displacement data, inclinometer data, water level data, rainfall data, and crack data may also include more eigenvalues. In addition, other types of sensor data can be added in addition to the displacement data, inclinometer data, water level data, rainfall data, and crack data, which should not be regarded as a limitation of the present invention.

[0048] In some embodiments of the present invention, refer to Figure 3 , zero values are interpolated into each three-dimensional grid with the attribute of no slope body, at least one first eigenvalue, at least one fourth eigenvalue, and at least one fifth eigenvalue are interpolated into each three-dimensional grid with the attribute of the slope surface, and at least one second eigenvalue and at least one third eigenvalue are interpolated into each three-dimensional grid with the attribute of the slope body interior to form a target tensor of T×W×H×D×C, where T represents the time step, and C represents the total number of eigenvalues of the first eigenvalue, second eigenvalue, third eigenvalue, fourth eigenvalue, and fifth eigenvalue, including: Interpolate zero values into a three-dimensional grid where each attribute is a non-slope body, interpolate the horizontal displacement, horizontal displacement velocity, elevation displacement, elevation displacement velocity, rainfall, crack width, and crack change velocity in the first slope direction into a three-dimensional grid where each attribute is the slope surface, and interpolate the horizontal displacement, displacement acceleration, and water level depth in the second slope direction into a three-dimensional grid where each attribute is the interior of the slope body to form a target tensor of T×W×H×D×C, where T represents the time step and C equals 10.

[0049] Taking a three-dimensional body of 50×40×20 as an example, the number of sensor features is 10, and taking 1 hour as a unit, the target tensor of 24-hour data is 24×50×40×20×10.

[0050] In some embodiments of the present invention, after obtaining the position data, displacement data, inclinometer data, water level data, rainfall data, and crack data of the target slope body, it further includes: Normalize the displacement data, inclinometer data, water level data, rainfall data, and crack data to eliminate the dimension difference.

[0051] The physical quantities measured by different sensors may be different, and the data ranges are also different. For example, the physical quantities of GNSS receivers, inclinometers, and crack gauges are displacements; the physical quantity of a water level gauge is the underground water level height; the physical quantity of a rain gauge is the rainfall. Therefore, normalization processing is required before using it for model training to eliminate the dimension difference. Here, the Z-Score standardization method can be used to make each eigenvalue conform to the standard normal distribution.

[0052] It should be noted that the principle and process of the Z-Score standardization method are prior arts known to those skilled in the art and will not be elaborated here.

[0053] In some embodiments of the present invention, referring to Figure 2 , the target slope body is also provided with an intelligent camera, and the landslide warning result is the warning level; The landslide monitoring and warning method based on multi-source data fusion further includes: Dynamically adjust the sensitivity of the intelligent camera based on the warning level, where the warning level and the sensitivity have a positive correlation.

[0054] The intelligent camera is Figure 2The AI camera in it. The AI camera can identify ongoing rockfalls and landslides and is a supplement to other sensors. However, the AI camera has a certain misjudgment rate for landslides, and the false alarm rate can be adjusted through the sensitivity setting of the AI algorithm. When the sensitivity is high, the missed judgment rate decreases but the misjudgment rate increases; when the sensitivity is set low, the misjudgment rate is low but the missed judgment rate will increase. Use the early warning output of other sensors to dynamically set the sensitivity of the AI camera. When there is no early warning, dynamically lower the sensitivity of the AI camera to reduce misjudgments and false alarm reports; the higher the early warning level, the higher the sensitivity of the AI camera is set to reduce the missed judgment rate.

[0055] In some embodiments of the present invention, the interpolation algorithm of the three-dimensional grid adopts the Kriging interpolation algorithm.

[0056] The Kriging interpolation algorithm originated from geostatistics, which allows us to estimate spatial variables that cannot be directly measured. Compared with traditional interpolation methods, the advantage of Kriging is that it uses a statistical model to predict variable values, and this method is particularly suitable for data with complex spatial correlations.

[0057] The Kriging interpolation algorithm is an optimal unbiased estimation method that not only considers the distance between points but also the spatial distribution pattern. The core idea of Kriging is that the attribute value of any point in a region can be estimated by the attribute values of the known points around that point and the spatial position relationship of these points. Kriging interpolation predicts the value of an unknown point by establishing an optimal prediction model, and its prediction error variance is the smallest.

[0058] Since the Kriging interpolation algorithm not only provides the predicted value of the interpolation point but also provides an assessment of the prediction error, it has great practical value for many application fields that require precise spatial analysis, such as mineral resource assessment, climate science, environmental engineering, and agriculture. The Kriging interpolation algorithm can provide us with an in-depth understanding of the spatial distribution of variables and is an indispensable tool for optimizing decision-making and resource management.

[0059] It should be noted that the principle and process of the Kriging interpolation algorithm are prior arts known to those skilled in the art, and the description of how to apply it in CNN will not be elaborated here.

[0060] According to the landslide monitoring and early warning system based on multi-source data fusion in the second aspect embodiment of the present invention, the system includes a data acquisition unit, a grid division unit, a grid attribute determination unit, a grid interpolation unit, and a landslide early warning unit.

[0061] The data acquisition unit is used to acquire the position data, displacement data, inclinometer data, water level data, rainfall data, and crack data of the target slope body; The mesh division unit is used to divide the target slope into W×H×D three-dimensional meshes, where W represents the width of the target slope, H represents the height of the target slope, and D represents the depth of the target slope; The mesh attribute determination unit is used to determine the attribute of each three-dimensional mesh according to the position data, and the attributes include no slope, slope surface, and slope interior; The mesh interpolation unit is used to interpolate zero values into each three-dimensional mesh with the attribute of no slope, interpolate displacement data, rainfall data, and crack data into each three-dimensional mesh with the attribute of slope surface, and interpolate inclinometer data and water level data into each three-dimensional mesh with the attribute of slope interior to form a target tensor; The landslide warning unit is used to input the target tensor into the 3D-CNN+LSTM model and output the landslide warning result.

[0062] Since the landslide monitoring and warning system based on multi-source data fusion adopts all the technical solutions of the landslide monitoring and warning method based on multi-source data fusion in the above embodiment, it at least has all the beneficial effects brought by the technical solutions of the above embodiment, and will not be elaborated here.

[0063] In some embodiments, referring to Figure 2 and Figure 5 , the landslide monitoring and warning system based on multi-source data fusion of the present invention also introduces edge computing technology, deploys a rockfall landslide recognition model, and the video monitoring is connected to the edge computing box for image recognition in the box. The edge computing box also has the function of controlling the local sound and light alarm system. When a warning is triggered, the sound and light alarm system is turned on, including horn broadcasting, flashing of strobe lights, display of warning information such as "Pay attention to landslide" on the LED sign, and projection of the word "Stop" on the road surface by the ground projection lamp to give a warning to passing vehicles in time. Specifically, when the landslide warning result is of high risk, the warning information is promptly informed to vehicles through the sound and light alarm system and the ground projection lamp to reduce the occurrence of casualties and vehicle accidents.

[0064] The system logic architecture of the landslide monitoring and early warning system based on multi-source data fusion according to the embodiments of the present invention includes a cloud computing layer, an access layer, an edge layer, a sensing layer, and an alarm layer. Among them, the sensing layer includes various types of sensors, including but not limited to AI cameras, GNSS receivers, inclinometers, water level gauges, rain gauges, and crack gauges; the alarm layer includes but not limited to ground projection lights, LED signs, strobe lights, and horn broadcasts; the edge layer includes an ad hoc network and an edge computing gateway; the access layer is 4G access; the cloud computing layer includes but not limited to device management, slope monitoring, AI model training, APIs, databases, video / picture libraries, AI model libraries, and computing power resources. The landslide monitoring and early warning method based on multi-source data fusion of the present invention is implemented through the cloud computing layer.

[0065] According to the landslide monitoring and early warning system based on multi-source data fusion of the embodiments of the present invention, the target slope is divided into W×H×D three-dimensional grids, and the attributes of each three-dimensional grid are determined according to the position data. Zero values are interpolated into each three-dimensional grid with the attribute of no slope, displacement data, rainfall data, and crack data are interpolated into each three-dimensional grid with the attribute of the slope surface, and inclinometer data and water level data are interpolated into each three-dimensional grid with the attribute of the slope interior to form a target tensor as the input of the 3D-CNN+LSTM model. Interpolation processing is performed on the attributes of each three-dimensional grid and the corresponding effective sensor monitoring data, expanding the monitoring point data from points to surfaces, facilitating application to the 3D-CNN+LSTM model, being able to reduce the data volume, simplify the calculation, reduce the cost, and be beneficial to the generalization of the model.

[0066] In addition, an embodiment of the present invention also provides a control device, which includes: a memory, a processor, and a computer program stored on the memory and executable on the processor. The processor and the memory can be connected through a bus or other means.

[0067] The memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. In addition, the memory can include high-speed random access memory, and can also include non-transitory memory, such as at least one disk storage device, a flash memory device, or other non-transitory solid-state storage devices. In some embodiments, the memory can optionally include a memory remotely set relative to the processor, and these remote memories can be connected to the processor through a network. Examples of the above networks include but are not limited to the Internet, enterprise intranets, local area networks, mobile communication networks, and their combinations.

[0068] The non-transitory software programs and instructions required to implement the landslide monitoring and early warning method based on multi-source data fusion of the above embodiments are stored in the memory, and when executed by the processor, they execute the landslide monitoring and early warning method based on multi-source data fusion in the above embodiments.

[0069] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0070] In addition, an embodiment of the present invention further provides a computer-readable storage medium storing computer-executable instructions, which are executed by a processor or a controller, for example, executed by the processor in the above embodiment, enabling the processor to execute the landslide monitoring and early warning method based on multi-source data fusion in the above embodiment.

[0071] Those of ordinary skill in the art can understand that all or some of the steps and systems in the methods disclosed above can be implemented as software, firmware, hardware, and their appropriate combinations. Some physical components or all physical components can be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or can be implemented as hardware, or can be implemented as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, which can include a computer storage medium (or non-transitory medium) and a communication medium (or transitory medium). As is well known to those of ordinary skill in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information, such as computer-readable instructions, data structures, program modules, or other data. Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory, or other memory technologies, CD-ROM, digital versatile disk (DVD), or other optical disk storage, magnetic cassette, tape, magnetic disk storage, or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, as is well known to those of ordinary skill in the art, a communication medium typically includes computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transmission mechanism, and can include any information delivery medium.

[0072] The embodiments of the present invention have been described in detail above in conjunction with the accompanying drawings, but the present invention is not limited to the above embodiments. Various changes can be made without departing from the spirit of the present invention within the scope of knowledge possessed by those of ordinary skill in the art to which the present invention pertains.

Claims

1. A landslide monitoring and early warning method based on multi-source data fusion, characterized in that, The method includes: Obtaining position data, displacement data, inclinometer data, water level data, rainfall data, and crack data of a target slope body; Divide the target slope into W×H×D three-dimensional grids, where W represents the width of the target slope, H represents the height of the target slope, and D represents the depth of the target slope; Determining the attributes of each of the three-dimensional grids according to the position data, where the attributes include no slope body, slope body surface, and slope body interior; Interpolating zero values into each of the three-dimensional grids with the attribute of no slope body, interpolating the displacement data, the rainfall data, and the crack data into each of the three-dimensional grids with the attribute of slope body surface, and interpolating the inclinometer data and the water level data into each of the three-dimensional grids with the attribute of slope body interior to form a target tensor; Inputting the target tensor into a 3D-CNN+LSTM model and outputting a landslide warning result.

2. The landslide monitoring and early warning method based on multi-source data fusion according to claim 1, characterized in that, The displacement data includes at least one first eigenvalue, the inclinometer data includes at least one second eigenvalue, the water level data includes at least one third eigenvalue, the rainfall data includes at least one fourth eigenvalue, and the crack data includes at least one fifth eigenvalue; The step of interpolating zero values into each of the three-dimensional grids with the attribute of no slope body, interpolating the displacement data, the rainfall data, and the crack data into each of the three-dimensional grids with the attribute of slope body surface, and interpolating the inclinometer data and the water level data into each of the three-dimensional grids with the attribute of slope body interior to form a target tensor includes: Interpolating zero values into each of the three-dimensional grids with the attribute of no slope body, interpolating the at least one first eigenvalue, the at least one fourth eigenvalue, and the at least one fifth eigenvalue into each of the three-dimensional grids with the attribute of slope body surface, and interpolating the at least one second eigenvalue and the at least one third eigenvalue into each of the three-dimensional grids with the attribute of slope body interior to form the target tensor of T×W×H×D×C, where T represents the time step and C represents the total number of eigenvalues of the first eigenvalue, the second eigenvalue, the third eigenvalue, the fourth eigenvalue, and the fifth eigenvalue.

3. The landslide monitoring and early warning method based on multi-source data fusion according to claim 2, characterized in that The displacement data includes four of the first eigenvalues, which are respectively: horizontal displacement in the first slope direction, horizontal displacement speed, elevation displacement, and elevation displacement speed; the inclinometer data includes two of the second eigenvalues, which are respectively: horizontal displacement in the second slope direction and displacement acceleration; the water level data includes one of the third eigenvalues, which is: water level depth; the rainfall data includes one of the fourth eigenvalues, which is: rainfall; the crack data includes two of the fifth eigenvalues, which are respectively: crack width and crack change speed.

4. The landslide monitoring and early warning method based on multi-source data fusion according to claim 3, wherein Interpolating zero values into the three-dimensional grid with each attribute being the slope-free body, interpolating the at least one first eigenvalue, the at least one fourth eigenvalue, and the at least one fifth eigenvalue into the three-dimensional grid with each attribute being the slope surface, and interpolating the at least one second eigenvalue and the at least one third eigenvalue into the three-dimensional grid with each attribute being the slope interior to form the target tensor of T×W×H×D×C, where T represents the time step and C represents the total number of eigenvalues of the first eigenvalue, the second eigenvalue, the third eigenvalue, the fourth eigenvalue, and the fifth eigenvalue, includes: Interpolating zero values into the three-dimensional grid with each attribute being the slope-free body, interpolating the horizontal displacement in the first slope direction, the horizontal displacement velocity, the elevation displacement, the elevation displacement velocity, the rainfall, the crack width, and the crack change velocity into the three-dimensional grid with each attribute being the slope surface, and interpolating the horizontal displacement in the second slope direction, the displacement acceleration, and the water level depth into the three-dimensional grid with each attribute being the slope interior to form the target tensor of T×W×H×D×C, where T represents the time step and C is equal to 10.

5. The landslide monitoring and early warning method based on multi-source data fusion according to claim 1, characterized in that After obtaining the position data, displacement data, inclinometer data, water level data, rainfall data, and crack data of the target slope, it further includes: Normalizing the displacement data, the inclinometer data, the water level data, the rainfall data, and the crack data to eliminate the dimension difference.

6. The landslide monitoring and early warning method based on multi-source data fusion according to claim 1, wherein, The target slope is also provided with an intelligent camera, and the landslide warning result is the warning level; The landslide monitoring and warning method based on multi-source data fusion further includes: Dynamically adjusting the sensitivity of the intelligent camera based on the warning level, where the warning level and the sensitivity have a positive correlation.

7. The landslide monitoring and early warning method based on multi-source data fusion according to claim 1, wherein The interpolation algorithm of the three-dimensional grid adopts the Kriging interpolation algorithm.

8. A landslide monitoring and early warning system based on multi-source data fusion, characterized in that, The system includes: A data acquisition unit for acquiring the position data, displacement data, inclinometer data, water level data, rainfall data, and crack data of the target slope; Mesh division unit, used to divide the target slope into W×H×D three-dimensional meshes, where W represents the width of the target slope, H represents the height of the target slope, and D represents the depth of the target slope; A grid attribute determination unit for determining the attribute of each three-dimensional grid according to the position data, and the attribute includes slope-free body, slope surface, and slope interior; A grid interpolation unit for interpolating zero values into the three-dimensional grid with each attribute being the slope-free body, interpolating the displacement data, the rainfall data, and the crack data into the three-dimensional grid with each attribute being the slope surface, and interpolating the inclinometer data and the water level data into the three-dimensional grid with each attribute being the slope interior to form a target tensor; A landslide warning unit for inputting the target tensor into the 3D-CNN+LSTM model and outputting the landslide warning result.

Images