Landslide deformation early warning system and method

Abstract

The present invention discloses a landslide deformation early warning system and method, which belongs to the technical field of landslide deformation early warning. The present invention forms a three-dimensional monitoring network by arranging Nx inclinometers horizontally, Ny inclinometers vertically, and Nz inclinometers vertically in the area to be detected. The present invention extracts the spatial angle value of each inclinometer, constructs a three-dimensional spatial angle value matrix, and screens out abnormal areas according to these angle values, and calculates abnormal coefficients. Then, the three-dimensional spatial angle value matrix at each moment is separated into horizontal-vertical, vertical-vertical and horizontal-vertical angle matrices, and the corresponding gradient matrix is ​​constructed based on the angle matrices at adjacent moments. Finally, the landslide deformation early warning subsystem processes these matrices, adds abnormal coefficients for reinforcement, and obtains the landslide deformation degree value at each moment. The present invention significantly improves the identification and early warning capabilities of landslide risks through multi-dimensional monitoring and precise evaluation.

Description

Technical Field

[0001] The invention relates to the technical field of landslide deformation early warning, and in particular to a landslide deformation early warning system and method. Background Art

[0002] At present, landslide monitoring technology mainly relies on traditional geological exploration and manual observation. These methods are not only time-consuming and labor-intensive, but also difficult to achieve real-time monitoring of the dynamic changes of landslides.

[0003] In the prior art, inclinometers are widely used as an important monitoring tool for landslide monitoring. The prior art usually uses a single inclinometer to detect the landslide situation at a point. This monitoring method has significant limitations. Since landslide is a complex geological dynamic process, single-point monitoring is difficult to fully reflect the deformation characteristics and potential risks of the entire landslide area. Traditional monitoring methods can often only obtain local and discrete angle data, and cannot accurately evaluate the overall landslide deformation situation in a landslide area. Summary of the invention

[0004] In view of the above-mentioned deficiencies in the prior art, a landslide deformation early warning system and method provided by the present invention solves the problem of low accuracy in assessing the overall landslide deformation situation in a landslide area in the prior art.

[0005] In order to achieve the above-mentioned invention object, the technical solution adopted by the present invention is: a landslide deformation early warning system, including: a layout subsystem, a three-dimensional matrix construction subsystem, an abnormal movement extraction subsystem, a three-dimensional matrix separation subsystem, a gradient matrix construction subsystem and a landslide deformation early warning subsystem;

[0006] The arrangement subsystem is used to arrange Nx inclinometers horizontally, Ny inclinometers longitudinally, and Nz layers of inclinometers vertically in the area to be detected, forming Three-dimensional monitoring network, where Nx, Ny and Nz are positive integers;

[0007] The three-dimensional matrix construction subsystem is used to extract the spatial angle value of each inclinometer in the three-dimensional monitoring network and construct a three-dimensional spatial angle value matrix;

[0008] The abnormal movement extraction subsystem is used to screen out the abnormal movement area and calculate the abnormal movement coefficient according to the spatial angle value of each inclinometer in the three-dimensional monitoring network;

[0009] The three-dimensional matrix separation subsystem is used to separate the three-dimensional space angle value matrix at each moment into a horizontal-vertical angle matrix, a vertical-vertical angle matrix and a horizontal-vertical angle matrix;

[0010] The gradient matrix construction subsystem is used to construct a horizontal-vertical gradient matrix, a vertical-vertical gradient matrix and a horizontal-vertical gradient matrix according to the horizontal-vertical angle matrix, the vertical-vertical angle matrix and the horizontal-vertical angle matrix at adjacent moments;

[0011] The landslide deformation early warning subsystem is used to process the horizontal-vertical angle matrix, the vertical-vertical angle matrix, the horizontal-vertical angle matrix, the horizontal-vertical gradient matrix, the vertical-vertical gradient matrix and the horizontal-vertical gradient matrix, and obtains the landslide deformation degree value at each moment based on the enhancement of the anomaly coefficient.

[0012] Further, the abnormality extraction subsystem includes: an abnormality inclinometer extraction unit and an abnormality coefficient calculation unit;

[0013] The abnormal inclinometer extraction unit is used to take the inclinometer whose spatial angle value is greater than the spatial angle threshold in the three-dimensional monitoring network as the abnormal inclinometer;

[0014] The fluctuation coefficient calculation unit is used to calculate the fluctuation coefficient according to the number of fluctuation inclinometers at adjacent moments.

[0015] Furthermore, the expression of the abnormal coefficient calculation unit is:

[0016] , where ε t is the coefficient of change at time t, Q t is the number of inclinometers at time t, Q t-1 is the number of the abnormal inclinometers at the t-1th moment, and t is the number of the moment.

[0017] Furthermore, the expression of the elements in the horizontal-vertical angle matrix is:

[0018] , where x i,j is the element in the i-th row and j-th column of the horizontal-vertical angle matrix, x i,j,n is the element of the i-th horizontal position, j-th vertical position and n-th vertical depth in the three-dimensional space angle value matrix, where i, j and n are positive integers;

[0019] The expression of the elements in the vertical-vertical angle matrix is:

[0020] , where x j,n is the element in the jth row and nth column of the vertical-vertical angle matrix;

[0021] The expression of the elements in the horizontal-vertical angle matrix is:

[0022] , where x i,n It is the element in the i-th row and n-th column of the horizontal-vertical angle matrix.

[0023] Further, the gradient matrix construction subsystem includes: a matrix subtraction unit, a first gradient matrix construction unit, a second gradient matrix construction unit and a third gradient matrix construction unit;

[0024] The matrix subtraction unit is used to subtract the horizontal-vertical angle matrix at time t from the horizontal-vertical angle matrix at time t-1 to obtain a horizontal-vertical angle difference matrix, subtract the vertical-vertical angle matrix at time t from the vertical-vertical angle matrix at time t-1 to obtain a vertical-vertical angle difference matrix, and subtract the horizontal-vertical angle matrix at time t from the horizontal-vertical angle matrix at time t-1 to obtain a horizontal-vertical angle difference matrix;

[0025] The first gradient matrix construction unit is used to calculate the element gradient values ​​according to the horizontal-vertical angle difference matrix and construct the horizontal-vertical gradient matrix; the second gradient matrix construction unit is used to calculate the element gradient values ​​according to the vertical-vertical angle difference matrix and construct the vertical-vertical gradient matrix; the third gradient matrix construction unit is used to calculate the element gradient values ​​according to the horizontal-vertical angle difference matrix and construct the horizontal-vertical gradient matrix.

[0026] Furthermore, the formula for calculating the element gradient value is:

[0027] , where α grad,i,j is the gradient value of the element in the i-th row and j-th column, α i,j is the element in the i-th row and j-th column of the horizontal-vertical angle difference matrix, the vertical-vertical angle difference matrix, or the horizontal-vertical angle difference matrix, φ max is the maximum spatial angle value.

[0028] Furthermore, the landslide deformation early warning subsystem includes: 3 feature extraction units, a mean feature fusion unit, a maximum feature fusion unit, a first feature enhancement unit, a second feature enhancement unit and a fully connected layer;

[0029] The first input end of the first feature extraction unit is used to input the horizontal-vertical angle matrix, and the second input end thereof is used to input the horizontal-vertical gradient matrix; the first input end of the second feature extraction unit is used to input the vertical-vertical angle matrix, and the second input end thereof is used to input the vertical-vertical gradient matrix; the first input end of the third feature extraction unit is used to input the horizontal-vertical angle matrix, and the second input end thereof is used to input the horizontal-vertical gradient matrix;

[0030] The input end of the mean feature fusion unit is connected to the first output ends of the three feature extraction units respectively; the input end of the maximum feature fusion unit is connected to the second output ends of the three feature extraction units respectively;

[0031] The first input end of the first feature enhancement unit is connected to the output end of the mean feature fusion unit, and the second input end thereof is used to input the variance coefficient;

[0032] The first input end of the second feature enhancement unit is connected to the output end of the maximum feature fusion unit, and the second input end thereof is used to input the variance coefficient;

[0033] The input end of the fully connected layer is connected to the output end of the first feature enhancement unit and the output end of the second feature enhancement unit respectively, and the output end thereof serves as the output end of the landslide deformation early warning subsystem.

[0034] Furthermore, the expressions of the feature extraction units are:

[0035] , , where Y1 is the output of the first output terminal of the feature extraction unit, Y2 is the output of the second output terminal of the feature extraction unit, X1 is the input of the first input terminal of the feature extraction unit, and X2 is the input of the second input terminal of the feature extraction unit. for The convolutional layer of size, AvgPool is the average pooling layer, and MaxPool is the maximum pooling layer.

[0036] Furthermore, the expressions of the feature enhancement units are:

[0037] , where H is the output of the feature enhancement unit, R is the input of the first input terminal of the feature enhancement unit, Sigmoid is the activation function, ε t is the coefficient of variation at the tth moment, ω εt is t The weight of b εt is t The offset of , t is the number of the moment.

[0038] A landslide deformation early warning method comprises the following steps:

[0039] S1, arrange Nx inclinometers horizontally, Ny inclinometers vertically, and Nz layers of inclinometers vertically in the area to be detected, forming Three-dimensional monitoring network, where Nx, Ny and Nz are positive integers;

[0040] S2, extracting the spatial angle value of each inclinometer in the three-dimensional monitoring network and constructing a three-dimensional spatial angle value matrix;

[0041] S3, according to the spatial angle value of each inclinometer in the three-dimensional monitoring network, screen out the abnormal area and calculate the abnormal coefficient;

[0042] S4, separating the three-dimensional space angle value matrix at each moment into a horizontal-vertical angle matrix, a vertical-vertical angle matrix and a horizontal-vertical angle matrix;

[0043] S5, constructing a horizontal-vertical gradient matrix, a vertical-vertical gradient matrix and a horizontal-vertical gradient matrix according to the horizontal-vertical angle matrix, the vertical-vertical angle matrix and the horizontal-vertical angle matrix at adjacent moments;

[0044] S6. The landslide deformation early warning subsystem is used to process the horizontal-vertical angle matrix, the vertical-vertical angle matrix, the horizontal-vertical angle matrix, the horizontal-vertical gradient matrix, the vertical-vertical gradient matrix and the horizontal-vertical gradient matrix, and based on the enhancement of the anomaly coefficient, the landslide deformation degree value at each moment is obtained.

[0045] The beneficial effects of the present invention are:

[0046] 1. The present invention constructs a three-dimensional monitoring network by arranging inclinometers in multiple dimensions, horizontally, vertically and vertically in the area to be detected. This three-dimensional layout significantly improves the spatial coverage of the landslide area, can capture the subtle deformation characteristics of the landslide area in an all-round and multi-level manner, overcomes the limitations of traditional single-point monitoring, and improves the accuracy of landslide assessment in the landslide area.

[0047] 2. The anomaly extraction subsystem selects key deformation areas and calculates the anomaly coefficient, which greatly improves the accuracy of landslide risk identification.

[0048] 3. The present invention calculates the horizontal-vertical gradient matrix, the vertical-vertical gradient matrix and the horizontal-vertical gradient matrix according to the horizontal-vertical angle matrix, the vertical-vertical angle matrix and the horizontal-vertical angle matrix at adjacent moments, reflecting the landslide deformation speed conditions in the horizontal-vertical, vertical-vertical and horizontal-vertical aspects.

[0049] 4. The landslide deformation early warning subsystem of the present invention can accurately calculate the landslide deformation value at each moment by comprehensively processing multiple angle matrices and gradient matrices and combining them with the abnormal coefficient for strengthening. The present invention significantly improves the accuracy and reliability of landslide deformation assessment through a multi-dimensional and multi-level comprehensive assessment method. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is a system block diagram of a landslide deformation early warning system;

[0051] Figure 2 This is a structural diagram of the landslide deformation early warning subsystem. DETAILED DESCRIPTION

[0052] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0053] Embodiment 1, as Figure 1 As shown, a landslide deformation early warning system includes: a layout subsystem, a three-dimensional matrix construction subsystem, an abnormal movement extraction subsystem, a three-dimensional matrix separation subsystem, a gradient matrix construction subsystem and a landslide deformation early warning subsystem;

[0054] The arrangement subsystem is used to arrange Nx inclinometers horizontally, Ny inclinometers longitudinally, and Nz layers of inclinometers vertically in the area to be detected, forming Three-dimensional monitoring network, where Nx, Ny and Nz are positive integers;

[0055] The three-dimensional matrix construction subsystem is used to extract the spatial angle value of each inclinometer in the three-dimensional monitoring network and construct a three-dimensional spatial angle value matrix;

[0056] The abnormal movement extraction subsystem is used to screen out the abnormal movement area and calculate the abnormal movement coefficient according to the spatial angle value of each inclinometer in the three-dimensional monitoring network;

[0057] The three-dimensional matrix separation subsystem is used to separate the three-dimensional space angle value matrix at each moment into a horizontal-vertical angle matrix, a vertical-vertical angle matrix and a horizontal-vertical angle matrix;

[0058] The gradient matrix construction subsystem is used to construct a horizontal-vertical gradient matrix, a vertical-vertical gradient matrix and a horizontal-vertical gradient matrix according to the horizontal-vertical angle matrix, the vertical-vertical angle matrix and the horizontal-vertical angle matrix at adjacent moments;

[0059] The landslide deformation early warning subsystem is used to process the horizontal-vertical angle matrix, the vertical-vertical angle matrix, the horizontal-vertical angle matrix, the horizontal-vertical gradient matrix, the vertical-vertical gradient matrix and the horizontal-vertical gradient matrix, and obtains the landslide deformation degree value at each moment based on the enhancement of the anomaly coefficient.

[0060] For example, five inclinometers are arranged horizontally in a landslide monitoring area, that is, Nx=5, and four inclinometers are arranged longitudinally in the same monitoring area, that is, Ny=4, which means that there are four inclinometers for monitoring in the length direction of the area. Assume that two layers of inclinometers are arranged vertically in the area, that is, Nz=2, which means that there are two layers of inclinometers in the depth direction of the area to monitor deformation at different depths. The first layer of inclinometers, depth: 5 meters, select the inclinometer with a depth of 5 meters to monitor the shallow deformation of the landslide area. This depth is usually the initial sliding area of ​​soil and rock when a landslide occurs, and it can capture potential deformation signals in time. The second layer of inclinometers, depth: 15 meters, select the inclinometer with a depth of 15 meters to monitor the deep deformation of the landslide area. A total of 40 inclinometers are arranged. This depth usually involves deeper soil and rock structures, which can provide important information about landslide stability and potential sliding surfaces.

[0061] In this embodiment, the calculation formula of the spatial angle value is:

[0062] , where θ s is the spatial angle value, θ x is the X-axis inclination angle output by the inclinometer, θ y is the Y-axis inclination angle output by the inclinometer, θ z It is the Z-axis inclination angle output by the inclinometer.

[0063] In this embodiment, the anomaly extraction subsystem includes: an anomaly inclinometer extraction unit and an anomaly coefficient calculation unit;

[0064] The abnormal inclinometer extraction unit is used to take the inclinometer whose spatial angle value is greater than the spatial angle threshold in the three-dimensional monitoring network as the abnormal inclinometer;

[0065] The fluctuation coefficient calculation unit is used to calculate the fluctuation coefficient according to the number of fluctuation inclinometers at adjacent moments.

[0066] In the present invention, the spatial angle threshold is a threshold set for the spatial angle value.

[0067] In this embodiment, the expression of the variance coefficient calculation unit is:

[0068] , where ε t is the coefficient of change at time t, Q t is the number of inclinometers at time t, Q t-1 is the number of the abnormal inclinometers at the t-1th moment, and t is the number of the moment.

[0069] The present invention counts the number of abnormal inclinometers at each moment, calculates an abnormal coefficient according to the difference in the number of abnormal inclinometers at adjacent moments and the number of abnormal inclinometers at the current moment, and reflects the deformation size of the landslide through the abnormal coefficient.

[0070] In this embodiment, the expression of the elements in the horizontal-vertical angle matrix is:

[0071] , where x i,j is the element in the i-th row and j-th column of the horizontal-vertical angle matrix, x i,j,n is the element of the i-th horizontal position, j-th vertical position and n-th vertical depth in the three-dimensional space angle value matrix, where i, j and n are positive integers;

[0072] The expression of the elements in the vertical-vertical angle matrix is:

[0073] , where x j,n is the element in the jth row and nth column of the vertical-vertical angle matrix;

[0074] The expression of the elements in the horizontal-vertical angle matrix is:

[0075] , where x i,n It is the element in the i-th row and n-th column of the horizontal-vertical angle matrix.

[0076] The present invention calculates the average of the vertical, horizontal and longitudinal directions respectively, so that the three-dimensional data is converted into a two-dimensional matrix, and the three-dimensional matrix is ​​embodied by three two-dimensional matrices.

[0077] In this embodiment, the gradient matrix construction subsystem includes: a matrix subtraction unit, a first gradient matrix construction unit, a second gradient matrix construction unit and a third gradient matrix construction unit;

[0078] The matrix subtraction unit is used to subtract the horizontal-vertical angle matrix at time t from the horizontal-vertical angle matrix at time t-1 to obtain a horizontal-vertical angle difference matrix, subtract the vertical-vertical angle matrix at time t from the vertical-vertical angle matrix at time t-1 to obtain a vertical-vertical angle difference matrix, and subtract the horizontal-vertical angle matrix at time t from the horizontal-vertical angle matrix at time t-1 to obtain a horizontal-vertical angle difference matrix;

[0079] The first gradient matrix construction unit is used to calculate the element gradient values ​​according to the horizontal-vertical angle difference matrix and construct the horizontal-vertical gradient matrix; the second gradient matrix construction unit is used to calculate the element gradient values ​​according to the vertical-vertical angle difference matrix and construct the vertical-vertical gradient matrix; the third gradient matrix construction unit is used to calculate the element gradient values ​​according to the horizontal-vertical angle difference matrix and construct the horizontal-vertical gradient matrix.

[0080] In this embodiment, the formula for calculating the element gradient value is:

[0081] , where α grad,i,j is the gradient value of the element in the i-th row and j-th column, α i,j is the element in the i-th row and j-th column of the horizontal-vertical angle difference matrix, the vertical-vertical angle difference matrix, or the horizontal-vertical angle difference matrix, φ max is the maximum spatial angle value.

[0082] The maximum value of the X-axis inclination, Y-axis inclination, and Z-axis inclination is 180°, so the maximum spatial angle value is °.

[0083] The present invention subtracts the horizontal-vertical angle matrix, the vertical-vertical angle matrix and the horizontal-vertical angle matrix at adjacent moments, and calculates the element gradient value according to the difference, reflecting the angle change at each position and reflecting the instantaneous change.

[0084] like Figure 2As shown, the landslide deformation early warning subsystem includes: 3 feature extraction units, mean feature fusion unit, maximum feature fusion unit, first feature enhancement unit, second feature enhancement unit and full connection layer;

[0085] The first input end of the first feature extraction unit is used to input the horizontal-vertical angle matrix, and the second input end thereof is used to input the horizontal-vertical gradient matrix; the first input end of the second feature extraction unit is used to input the vertical-vertical angle matrix, and the second input end thereof is used to input the vertical-vertical gradient matrix; the first input end of the third feature extraction unit is used to input the horizontal-vertical angle matrix, and the second input end thereof is used to input the horizontal-vertical gradient matrix;

[0086] The input end of the mean feature fusion unit is connected to the first output ends of the three feature extraction units respectively; the input end of the maximum feature fusion unit is connected to the second output ends of the three feature extraction units respectively;

[0087] The first input end of the first feature enhancement unit is connected to the output end of the mean feature fusion unit, and the second input end thereof is used to input the variance coefficient;

[0088] The first input end of the second feature enhancement unit is connected to the output end of the maximum feature fusion unit, and the second input end thereof is used to input the variance coefficient;

[0089] The input end of the fully connected layer is connected to the output end of the first feature enhancement unit and the output end of the second feature enhancement unit respectively, and the output end thereof serves as the output end of the landslide deformation early warning subsystem.

[0090] The present invention extracts mean features and maximum features through a feature extraction unit, adopts a mean feature fusion unit to fuse the mean features output by three feature extraction units, adopts a maximum feature fusion unit to fuse the maximum features output by three feature extraction units, a first feature enhancement unit uses an anomaly coefficient to enhance the fused mean features, and a second feature enhancement unit uses an anomaly coefficient to enhance the fused maximum features, and then the enhanced features are comprehensively enhanced through a fully connected layer to obtain a landslide deformation degree value.

[0091] In this embodiment, the mean feature fusion unit and the maximum feature fusion unit can be implemented by an adder, by which the three features are added, or by a Concat layer, by which the three features are concatenated.

[0092] In this embodiment, the expressions of the feature extraction units are all:

[0093] , , where Y1 is the output of the first output terminal of the feature extraction unit, Y2 is the output of the second output terminal of the feature extraction unit, X1 is the input of the first input terminal of the feature extraction unit, and X2 is the input of the second input terminal of the feature extraction unit. for The convolutional layer of size, AvgPool is the average pooling layer, and MaxPool is the maximum pooling layer.

[0094] The present invention performs the following steps on the input of the first input terminal and the input of the second input terminal of the feature extraction unit: The convolution operation of the size is performed, and then the matrix is ​​added to merge the features corresponding to the angle matrix and the gradient matrix. Then the average pooling layer is used to extract the mean feature, and the maximum pooling layer is used to extract the maximum feature. In this embodiment, the window size of the average pooling layer and the maximum pooling layer is .

[0095] In this embodiment, the expressions of the feature enhancement units are:

[0096] , where H is the output of the feature enhancement unit, R is the input of the first input terminal of the feature enhancement unit, Sigmoid is the activation function, ε t is the coefficient of variation at the tth moment, ω εt is t The weight of b εt is t The offset of , t is the number of the moment.

[0097] The present invention assigns weights and biases to the abnormal coefficients to obtain the strengthening ratio , the mean and maximum characteristics are enhanced according to the enhancement ratio to achieve a more accurate landslide deformation assessment.

[0098] In the present invention, the weights and biases in the landslide deformation early warning subsystem can be obtained by training using the existing gradient descent method.

[0099] In this embodiment, when the landslide deformation degree value exceeds the landslide deformation degree threshold, an early warning may be issued.

[0100] Embodiment 2, a landslide deformation early warning method, comprising the following steps:

[0101] S1, arrange Nx inclinometers horizontally, Ny inclinometers vertically, and Nz layers of inclinometers vertically in the area to be detected, forming Three-dimensional monitoring network, where Nx, Ny and Nz are positive integers;

[0102] S2, extracting the spatial angle value of each inclinometer in the three-dimensional monitoring network and constructing a three-dimensional spatial angle value matrix;

[0103] S3, according to the spatial angle value of each inclinometer in the three-dimensional monitoring network, screen out the abnormal area and calculate the abnormal coefficient;

[0104] S4, separating the three-dimensional space angle value matrix at each moment into a horizontal-vertical angle matrix, a vertical-vertical angle matrix and a horizontal-vertical angle matrix;

[0105] S5, constructing a horizontal-vertical gradient matrix, a vertical-vertical gradient matrix and a horizontal-vertical gradient matrix according to the horizontal-vertical angle matrix, the vertical-vertical angle matrix and the horizontal-vertical angle matrix at adjacent moments;

[0106] S6. The landslide deformation early warning subsystem is used to process the horizontal-vertical angle matrix, the vertical-vertical angle matrix, the horizontal-vertical angle matrix, the horizontal-vertical gradient matrix, the vertical-vertical gradient matrix and the horizontal-vertical gradient matrix, and based on the enhancement of the anomaly coefficient, the landslide deformation degree value at each moment is obtained.

[0107] The specific implementation process of Example 2 is the same as that of Example 1.

[0108] The present invention constructs a three-dimensional monitoring network by arranging inclinometers in multiple dimensions, horizontally, vertically and vertically in the area to be detected. This three-dimensional layout significantly improves the spatial coverage of the landslide area, can capture the subtle deformation characteristics of the landslide area in an all-round and multi-level manner, overcomes the limitations of traditional single-point monitoring, and improves the overall landslide deformation assessment accuracy of the landslide area.

[0109] The anomaly extraction subsystem selects key deformation areas and calculates anomaly coefficients, which greatly improves the accuracy of landslide risk identification.

[0110] The present invention calculates the transverse-vertical gradient matrix, the vertical-vertical gradient matrix and the transverse-vertical gradient matrix according to the transverse-vertical angle matrix, the vertical-vertical angle matrix and the transverse-vertical angle matrix at adjacent moments, reflecting the landslide deformation speed conditions in the transverse-vertical, vertical-vertical and transverse-vertical aspects.

[0111] The landslide deformation early warning subsystem of the present invention can accurately calculate the landslide deformation value at each moment by comprehensively processing multiple angle matrices and gradient matrices and combining them with the anomaly coefficient for reinforcement. The present invention significantly improves the accuracy and reliability of landslide deformation assessment through a multi-dimensional and multi-level comprehensive assessment method.

[0112] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A landslide deformation early warning system, characterized in that: include: Layout subsystem, three-dimensional matrix construction subsystem, abnormal movement extraction subsystem, three-dimensional matrix separation subsystem, gradient matrix construction subsystem and landslide deformation early warning subsystem; The arrangement subsystem is used to arrange Nx inclinometers horizontally, Ny inclinometers longitudinally, and Nz layers of inclinometers vertically in the area to be detected, forming Three-dimensional monitoring network, where Nx, Ny and Nz are positive integers; The three-dimensional matrix construction subsystem is used to extract the spatial angle value of each inclinometer in the three-dimensional monitoring network and construct a three-dimensional spatial angle value matrix; The abnormality extraction subsystem is used to screen out abnormality areas and calculate abnormality coefficients based on the spatial angle value of each inclinometer in the three-dimensional monitoring network: , where ε t is the coefficient of change at time t, Q t is the number of inclinometers at time t, Q t-1 is the number of the abnormal inclinometers at the time t-1, and t is the number of the time; The three-dimensional matrix separation subsystem is used to separate the three-dimensional space angle value matrix at each moment into a horizontal-vertical angle matrix, a vertical-vertical angle matrix and a horizontal-vertical angle matrix; The gradient matrix construction subsystem is used to construct a horizontal-vertical gradient matrix, a vertical-vertical gradient matrix and a horizontal-vertical gradient matrix according to the horizontal-vertical angle matrix, the vertical-vertical angle matrix and the horizontal-vertical angle matrix at adjacent moments, wherein the horizontal-vertical angle matrices at adjacent moments are subtracted, and the element gradient values ​​are calculated according to the difference to obtain the horizontal-vertical gradient matrix, the vertical-vertical angle matrices at adjacent moments are subtracted, and the element gradient values ​​are calculated according to the difference to obtain the vertical-vertical gradient matrix, the horizontal-vertical gradient matrices at adjacent moments are subtracted, and the element gradient values ​​are calculated according to the difference to obtain the horizontal-vertical gradient matrix, and the element gradient value is the ratio of the difference to the maximum spatial angle value; The landslide deformation early warning subsystem is used to process the horizontal-vertical angle matrix, the vertical-vertical angle matrix, the horizontal-vertical angle matrix, the horizontal-vertical gradient matrix, the vertical-vertical gradient matrix and the horizontal-vertical gradient matrix, and obtain the landslide deformation degree value at each moment based on the enhancement of the anomaly coefficient; The landslide deformation early warning subsystem includes: 3 feature extraction units, mean feature fusion unit, maximum feature fusion unit, first feature enhancement unit, second feature enhancement unit and full connection layer; The first input end of the first feature extraction unit is used to input the horizontal-vertical angle matrix, and the second input end thereof is used to input the horizontal-vertical gradient matrix; the first input end of the second feature extraction unit is used to input the vertical-vertical angle matrix, and the second input end thereof is used to input the vertical-vertical gradient matrix; the first input end of the third feature extraction unit is used to input the horizontal-vertical angle matrix, and the second input end thereof is used to input the horizontal-vertical gradient matrix; The input end of the mean feature fusion unit is connected to the first output ends of the three feature extraction units respectively; the input end of the maximum feature fusion unit is connected to the second output ends of the three feature extraction units respectively; The first input end of the first feature enhancement unit is connected to the output end of the mean feature fusion unit, and the second input end thereof is used to input the variance coefficient; The first input end of the second feature enhancement unit is connected to the output end of the maximum feature fusion unit, and the second input end thereof is used to input the variance coefficient; The input end of the fully connected layer is connected to the output end of the first feature enhancement unit and the output end of the second feature enhancement unit respectively, and the output end thereof serves as the output end of the landslide deformation early warning subsystem; The expressions of feature enhancement units are: , where H is the output of the feature enhancement unit, R is the input of the first input terminal of the feature enhancement unit, Sigmoid is the activation function, ε t is the coefficient of variation at the tth moment, ω εt is t The weight of b εt is t The offset of , t is the number of the moment.

2. The landslide deformation early warning system according to claim 1, characterized in that: The abnormality extraction subsystem includes: an abnormality inclinometer extraction unit and an abnormality coefficient calculation unit; The abnormal inclinometer extraction unit is used to take the inclinometers with spatial angle values ​​greater than the spatial angle threshold in the three-dimensional monitoring network as abnormal inclinometers; The fluctuation coefficient calculation unit is used to calculate the fluctuation coefficient according to the number of fluctuation inclinometers at adjacent moments.

3. The landslide deformation early warning system according to claim 1, characterized in that: The expression of the elements in the horizontal-vertical angle matrix is: , where x i,j is the element in the i-th row and j-th column of the horizontal-vertical angle matrix, x i,j,n is the element of the i-th horizontal position, j-th vertical position and n-th vertical depth in the three-dimensional space angle value matrix, where i, j and n are positive integers; The expression of the elements in the vertical-vertical angle matrix is: , where x j,n is the element in the jth row and nth column of the vertical-vertical angle matrix; The expression of the elements in the horizontal-vertical angle matrix is: , where x i,n It is the element in the i-th row and n-th column of the horizontal-vertical angle matrix.

4. The landslide deformation early warning system according to claim 1, characterized in that: The gradient matrix construction subsystem comprises: a matrix subtraction unit, a first gradient matrix construction unit, a second gradient matrix construction unit and a third gradient matrix construction unit; The matrix subtraction unit is used to subtract the horizontal-vertical angle matrix at time t from the horizontal-vertical angle matrix at time t-1 to obtain a horizontal-vertical angle difference matrix, subtract the vertical-vertical angle matrix at time t from the vertical-vertical angle matrix at time t-1 to obtain a vertical-vertical angle difference matrix, and subtract the horizontal-vertical angle matrix at time t from the horizontal-vertical angle matrix at time t-1 to obtain a horizontal-vertical angle difference matrix; The first gradient matrix construction unit is used to calculate the element gradient values ​​according to the horizontal-vertical angle difference matrix and construct the horizontal-vertical gradient matrix; the second gradient matrix construction unit is used to calculate the element gradient values ​​according to the vertical-vertical angle difference matrix and construct the vertical-vertical gradient matrix; the third gradient matrix construction unit is used to calculate the element gradient values ​​according to the horizontal-vertical angle difference matrix and construct the horizontal-vertical gradient matrix.

5. The landslide deformation early warning system according to claim 4, characterized in that: The formula for calculating the element gradient value is: , where α grad,i,j is the gradient value of the element in the i-th row and j-th column, α i,j is the element in the i-th row and j-th column of the horizontal-vertical angle difference matrix, the vertical-vertical angle difference matrix, or the horizontal-vertical angle difference matrix, φ max is the maximum spatial angle value.

6. The landslide deformation early warning system according to claim 1, characterized in that: The expressions of the feature extraction units are all: , , where Y1 is the output of the first output terminal of the feature extraction unit, Y2 is the output of the second output terminal of the feature extraction unit, X1 is the input of the first input terminal of the feature extraction unit, and X2 is the input of the second input terminal of the feature extraction unit. for The convolutional layer of size, AvgPool is the average pooling layer, and MaxPool is the maximum pooling layer.

7. A landslide deformation early warning method, implemented based on the landslide deformation early warning system according to any one of claims 1 to 6, characterized in that: The following steps are involved: S1, arrange Nx inclinometers horizontally, Ny inclinometers vertically, and Nz layers of inclinometers vertically in the area to be detected, forming Three-dimensional monitoring network, where Nx, Ny and Nz are positive integers; S2, extracting the spatial angle value of each inclinometer in the three-dimensional monitoring network and constructing a three-dimensional spatial angle value matrix; S3, according to the spatial angle value of each inclinometer in the three-dimensional monitoring network, screen out the abnormal area and calculate the abnormal coefficient; S4, separating the three-dimensional space angle value matrix at each moment into a horizontal-vertical angle matrix, a vertical-vertical angle matrix and a horizontal-vertical angle matrix; S5, constructing a horizontal-vertical gradient matrix, a vertical-vertical gradient matrix and a horizontal-vertical gradient matrix according to the horizontal-vertical angle matrix, the vertical-vertical angle matrix and the horizontal-vertical angle matrix at adjacent moments; S6. The landslide deformation early warning subsystem is used to process the horizontal-vertical angle matrix, the vertical-vertical angle matrix, the horizontal-vertical angle matrix, the horizontal-vertical gradient matrix, the vertical-vertical gradient matrix and the horizontal-vertical gradient matrix, and based on the enhancement of the anomaly coefficient, the landslide deformation degree value at each moment is obtained.

Images