Slope deformation prediction method and multi-source heterogeneous data acquisition system

Abstract

The invention discloses a slope deformation prediction method and a multi-source heterogeneous data acquisition system. The slope deformation prediction method comprises the following steps: acquiring multi-source heterogeneous data of each point location of a slope, wherein the multi-source heterogeneous data comprises Beidou GNSS data, MEMS data and InSAR data; preprocessing the multi-source heterogeneous data; carrying out clustering processing on the preprocessed multi-source heterogeneous data; and predicting the deformation of each point position of the slope to complete slope deformation prediction. According to the slope deformation prediction method, comprehensive prediction of slope deformation is achieved through multiple monitoring means and multiple prediction models, and the prediction result is high in accuracy, scientific and reliable.

Description

Technical Field

[0001] The invention relates to a method for predicting slope deformation and a multi-source heterogeneous data acquisition system for realizing the method, belonging to the technical field of slope deformation prediction. Background Art

[0002] Slope stability is affected by its complex geological conditions and has always been a focus of geotechnical engineering. With the rapid development of the economy, the gradual expansion of the scale of engineering projects, and the gradual reduction of optional engineering sites, the geological conditions faced by engineering design have become more complex. With the construction of large-scale projects, the construction of high dams and reservoirs has gradually increased, and the application of high and steep slopes has increased. Therefore, when designing slopes, it is necessary to consider more the impact of the geological conditions of the slope on its stability and its changing trend.

[0003] At present, the evaluation of slope stability still adopts the combination of theoretical analysis, expert group experience and knowledge, and monitoring and control systems. Slope monitoring and prediction is an important link. Among them, the monitoring methods of slope deformation are, for example, measuring the surface deformation of the slope through Beidou, total station, three-dimensional laser, etc., measuring the surface deformation of the slope through remote sensing and optical image interpretation, and measuring the deep deformation of the slope through multi-point displacement meter, inclinometer, etc., and the prediction of slope deformation is, for example, using deterministic prediction model, statistical prediction model, nonlinear prediction model or comprehensive prediction model for prediction. From the actual implementation, it can be seen that the monitoring and prediction of slope deformation at this stage are based on a single monitoring method and a single prediction model, respectively. However, for the complexity of slope geological conditions, the above approach obviously can no longer meet the monitoring and prediction requirements, and it is necessary to develop in the direction of comprehensive prediction of multiple monitoring methods and multiple prediction models. Summary of the invention

[0004] The purpose of the present invention is to provide a slope deformation prediction method and a multi-source heterogeneous data acquisition system. This method realizes comprehensive prediction of slope deformation through multiple monitoring means and multiple prediction models. The prediction result is highly accurate, scientific and reliable.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] A slope deformation prediction method comprises the following steps: 1) acquiring multi-source heterogeneous data of each point on the slope, including Beidou GNSS data, MEMS data and InSAR data; 2) preprocessing the multi-source heterogeneous data; 3) clustering the preprocessed multi-source heterogeneous data; 4) predicting the deformation of each point on the slope to complete the slope deformation prediction.

[0007] A multi-source heterogeneous data acquisition system is used to acquire multi-source heterogeneous data in the slope deformation prediction method, including a base arranged on the slope, a column is provided on the base, a positioning receiver is installed on the column, a reflecting pyramid is installed on the positioning receiver, a flexible displacement meter is provided under the base, the flexible displacement meter extends into the slope to a set depth, the flexible displacement meter and the positioning receiver are electrically connected to a collector arranged on the base, wherein: the positioning receiver is used to obtain the Beidou GNSS data of fixed points on the slope surface; the flexible displacement meter is used to obtain MEMS data of fixed points on the slope surface, and the MEMS data is the internal horizontal displacement data of the slope at different depths in the same vertical direction; the reflecting pyramid is used to assist satellite remote sensing InSAR in obtaining InSAR data of permanent scattering points on the slope surface.

[0008] The advantages of the present invention are:

[0009] The present invention constructs a three-dimensional monitoring system based on multiple monitoring methods such as BDS, flexible displacement meter, and satellite remote sensing InSAR, and performs big data preprocessing (cleaning), clustering processing, etc. on InSAR data, Beidou GNSS data, and MEMS data, as well as recursive prediction of multiple curves such as cumulative curves, so that the predicted slope deformation is more scientific and reliable, and the development trend of slope deformation can be truly predicted, providing a scientific basis for slope safety status evaluation. The present invention is suitable for various engineering slopes such as water conservancy project slopes. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 It is an implementation flow chart of the slope deformation prediction method of the present invention.

[0011] Figure 2A It is a scatter plot of the two-dimensional data of deformation amount and deformation rate at each point in 2018.

[0012] Figure 2B Yes Figure 2A The scatter plot shown is the result after clustering processing.

[0013] Figure 3A It is a scatter plot of the two-dimensional data of deformation amount and deformation rate at each point in 2019.

[0014] Figure 3B Yes Figure 3A The scatter plot shown is the result after clustering processing.

[0015] Figure 4A It is a scatter plot of the two-dimensional data of deformation amount and deformation rate at each point in 2020.

[0016] Figure 4B Yes Figure 4AThe scatter plot shown is the result after clustering processing.

[0017] Figure 5A It is a scatter plot of the two-dimensional data of deformation amount and deformation rate at each point in 2021.

[0018] Figure 5B Yes Figure 5A The scatter plot shown is the result after clustering processing.

[0019] Figure 6 is the deformation curve of the first period.

[0020] Figure 7 is the cumulative curve of the first period.

[0021] Figure 8 It is the deformation curve of the second period.

[0022] Fig. 9 is the cumulative curve of the second period.

[0023] Fig.10 It is a cumulative growth curve obtained based on the growth curves of the first period and the second period, as well as two fitted polynomial curves.

[0024] Fig.11 It is the cumulative prediction curve obtained after predicting the times 11 and 12.

[0025] Fig.12 It is a schematic diagram of the multi-source heterogeneous data acquisition system of the present invention. DETAILED DESCRIPTION

[0026] like Figure 1 The present invention proposes a slope deformation prediction method, comprising the steps of:

[0027] 1) Obtain multi-source heterogeneous data at each point on the slope, where the multi-source heterogeneous data includes Beidou GNSS data, MEMS data, and InSAR data;

[0028] 2) Preprocessing multi-source heterogeneous data;

[0029] 3) Clustering the preprocessed slope multi-source heterogeneous data;

[0030] 4) Predict the deformation of each point on the slope and complete the slope deformation prediction.

[0031] In step 1):

[0032] Beidou GNSS data is deformation data obtained by positioning and differential analysis of each surface fixed point in the specified area of ​​the slope through BDS (China's Beidou Satellite Navigation System). Each surface fixed point is regarded as a point. The collection period of Beidou GNSS data is 1 second to 3600 seconds, depending on actual requirements.

[0033] The MEMS data is the internal horizontal displacement data of the slope obtained by the flexible displacement meter 6 at different depths in the same vertical direction for each surface fixed point in the specified area of ​​the slope. Each depth position of each surface fixed point in the vertical direction is regarded as a point position. The acquisition period of MEMS data is 1 second to 3600 seconds, depending on the actual requirements.

[0034] InSAR data is the time-course deformation data obtained by interpreting satellite remote sensing InSAR images for each permanent scattering point on the surface within a specified area of ​​the slope. Each permanent scattering point on the surface is regarded as a point, and the collection period of InSAR data is 12 days.

[0035] The above-mentioned InSAR data, Beidou GNSS data and MEMS data are three types of data obtained by the present invention, which constitute multi-source heterogeneous data. Of course, other types of data can also be obtained according to needs during actual implementation.

[0036] Step 2) specifically includes:

[0037] 2-1) Convert all InSAR data, Beidou GNSS data and MEMS data into the same 2000 National Geodetic Coordinate System (China’s latest national geodetic coordinate system) to achieve a unified coordinate system;

[0038] 2-2) Based on the 3S criterion, the InSAR data, Beidou GNSS data and MEMS data are denoised, data points with a probability of less than 3‰ are removed, and other valid and real data points are retained;

[0039] 2-3) Through the Newton polynomial interpolation method, InSAR data, Beidou GNSS data and MEMS data are interpolated to retain the original change trend and fluctuation characteristics of the data;

[0040] 2-4) For the collection period of each type of data, InSAR data, Beidou GNSS data and MEMS data, each type of data obtained with 1 times collection period, 10 times collection period, 100 times collection period and 1000 times collection period is subjected to thinning processing to obtain data sets of different types of data and different periods, where the period refers to 1 times collection period, 10 times collection period, 100 times collection period and 1000 times collection period;

[0041] 2-5) Perform difference calculations (or difference calculations) on the data sets of each period for different types of data, complete the statistics of the combined difference of data sets of different periods, and then select the best combination of data sets with the smallest difference among the data sets of the same period and different types of data;

[0042] 2-6) Based on the selected optimal data set combination, for the same period and different types of data, a basic combination sequence containing InSAR data, Beidou GNSS data and MEMS data, that is, a fusion of multi-source heterogeneous data, is recombined, that is, based on the optimal data set combination, all types of data with the same period are combined to form a basic combination sequence;

[0043] 2-7) Based on the basic combination sequence, the basic combination sequence of each point in the specified area of ​​the slope with different periods is obtained.

[0044] In the invention, the 3S principle in slope engineering usually refers to the integrated application of Geographic Information System (GIS), Remote Sensing System (RS) and Global Positioning System (GPS). The combination of these three provides a new generation of observation means, description language and thinking tools for the prevention and control and prediction of slope engineering. Coordinate system 1, denoising, Newton polynomial interpolation, thinning and difference calculation are well-known technologies in the field and will not be described in detail here.

[0045] Step 3) specifically includes:

[0046] 3-1) Based on the energy function and the least squares method, each data (three types of data) in the basic combination sequence of each point in different periods is converted into deformation amount and deformation rate (how to convert each type of data into deformation amount representation and further into deformation rate representation in this step is a well-known technology in the art and will not be described in detail);

[0047] 3-2) Clustering is performed on the deformation amount, deformation rate and two-dimensional data composed of the deformation amount and deformation rate at different points in the same period (the period here is self-defined, the period has a different meaning from the above-mentioned period, and the period and period can be the same or different), wherein: the deformation amount, deformation rate and two-dimensional data composed of the deformation amount and deformation rate are all defined as objects, and the objects are composed of sample data, that is, the deformation amount, deformation rate and two-dimensional data composed of the deformation amount and deformation rate are three different objects, and each object is composed of sample data, for example, the deformation amount (object) is composed of many deformation amount data (sample data);

[0048] 3-2-1) Randomly assign the sample data of an object at different locations in the same period to form several random clusters;

[0049] 3-2-2) For each point of sample data, calculate the distance between this sample data and the center point of each cluster, and assign this sample data to the cluster with the closest distance to it to generate a new cluster;

[0050] 3-2-3) Calculate the center point of the new cluster by calculating the average value of all sample data in the new cluster;

[0051] 3-2-4) Repeat steps 3-2-2) to 3-2-3) until the center point of the cluster no longer changes significantly or the preset number of iterations is reached;

[0052] 3-2-5) Generate sample clusters for this object at different locations in the same period;

[0053] 3-3) Based on the sample clusters generated for each object, the cluster distribution of the basic combination sequence of each point in the specified area of ​​the slope with different periods is obtained.

[0054] like FIG. 2A to FIG. 5B As shown in the figure, the clustering processing of the two-dimensional data composed of the deformation amount and deformation rate of each point in the specified area of ​​the slope in four different periods from 2018 to 2021 is performed. Figure 2A The figure shows the scatter plot of the two-dimensional data of deformation amount and deformation rate at each point in 2018. Figure 2B It shows that Figure 2A The cluster distribution results obtained after clustering the scatter plot shown in the figure. Similarly, Figure 3A The figure shows the scatter plot of the two-dimensional data of deformation amount and deformation rate at each point in 2019. Figure 3B It shows that Figure 3A The cluster distribution results obtained after clustering processing of the scatter plot shown. Figure 4A The figure shows the scatter plot of the two-dimensional data of deformation amount and deformation rate at each point in 2020. Figure 4B It shows that Figure 4A The cluster distribution results obtained after clustering processing of the scatter plot shown. Figure 5A The figure shows the scatter plot of the two-dimensional data of deformation amount and deformation rate at each point in 2021. Figure 5B It shows that Figure 5A The cluster distribution results obtained after clustering the scatter plot shown in the figure. The above cluster distribution results reflect the similarity between the points in the specified area of ​​the slope, which is helpful for partitioning the points, that is, dividing the points with high similarity into one partition. Figure 2B , Figure 3B , Figure 4Band Figure 5B The asterisks shown in represent the center points of the clusters.

[0055] Step 4) specifically includes:

[0056] 4-1) For the same period, all points in the designated area of ​​the slope are divided into multiple partitions according to the cluster distribution of the basic combination sequence of each point;

[0057] 4-2) Based on the deformation curves of each point in the same partition, the cumulative curve of each point is obtained: the deformation of a point in the partition obtained at different times is cumulatively calculated to obtain the cumulative curve;

[0058] 4-3) According to the weights of each point in the same partition at different times (the weights are given according to the contribution correlation rate of the points), the cumulative curve is converted into a growth curve;

[0059] 4-4) Based on the growth curves of each point in different periods, the growth curves of the same point in different periods are connected end to end to form a cumulative growth curve, where the time of each period is continuous (for example, 2018 and 2019, these two periods are continuous in time);

[0060] 4-5) Perform polynomial fitting on the cumulative growth curve of the same point to fit multiple polynomial curves;

[0061] 4-6) Calculate the mean square error of each polynomial curve, and select the polynomial curve with the smallest mean square error from all polynomial curves as the prediction curve. The smaller the mean square error, the greater the significance;

[0062] 4-7) According to the time interval (i.e., the interval between two moments), the prediction curve of each point is predicted for multiple moments backward to obtain a cumulative prediction curve, wherein the moment of data prediction is recorded as the prediction moment;

[0063] 4-8) According to the weight of each point at the prediction time obtained in step 4-3), the cumulative deformation of the cumulative prediction curve of each point at each prediction time is restored, and then the deformation at each prediction time is obtained by a cumulative subtraction calculation;

[0064] 4-9) The deformation of different points in the specified area of ​​the slope at the prediction time is obtained, so as to realize the deformation prediction of each point of the slope based on the cluster distribution and complete the slope deformation prediction.

[0065] Here, since cluster distribution is a kind of big data classification, it can better obtain data rules, facilitate classification and prediction of data, and is more scientific. Therefore, the prediction result obtained by step 4) of the present invention can better reflect the deformation trend of the slope, has the characteristics of zoning, and makes it possible to objectively reflect the deformation of the slope.

[0066] Example:

[0067] For the continuous first period and second period, the deformation of a certain point in the specified area of ​​the slope is predicted, where the first period has 5 moments (moments 1 to 5), the second period has 5 moments (moments 6 to 10), a total of 10 moments, and the deformation obtained at each moment is shown in Table 1 below.

[0068] Table 1

[0069]

[0070] The deformation curve of this point in the first period is shown in Figure 6 Therefore, for the deformation at each moment, through a cumulative calculation (the cumulative deformation at the next moment is the sum of the cumulative deformation at the previous moment and the deformation at the next moment), the cumulative curve of this point in the first period is obtained (see Figure 7 ). Similarly, the deformation curve of this point in the second period can be found in Figure 8 Therefore, for the deformation at each moment, through a cumulative calculation, the cumulative curve of this point in the second period is obtained (see Fig. 9 ).

[0071] Then, according to the weight of this point at different times (given according to the contribution correlation rate of the point), the cumulative curves of the first period and the second period are converted into growth curves (the data of each moment is obtained by multiplying the cumulative deformation at this moment by the weight of this moment), and the two growth curves obtained for the first period and the second period are connected end to end to obtain a cumulative growth curve (see Fig.10 Then a polynomial fit is performed on this cumulative growth curve. For example, Fig.10 , fitting two polynomial curves (see Fig.10 The red dotted line and the green dotted line are shown in the figure). By calculating the mean square error of each polynomial curve, the mean square error of the red dotted line curve is 0.9488, and the mean square error of the green dotted line curve is 0.931. Therefore, the green dotted line curve with the smallest mean square error is selected as the prediction curve (the smaller the mean square error, the greater the significance).

[0072] Then, according to the prediction curve shown by the green dashed line (see Fig.10 ), according to the set time interval, the data of multiple moments are predicted backward to obtain the cumulative prediction curve. Fig.11 The cumulative prediction curve obtained by predicting the time 11 and 12 backward is shown, wherein the time 11 and 12 are the predicted time.

[0073] Then, according to the weights of each prediction moment given by the contribution correlation rate of the point, the accumulated deformation at each prediction moment is restored, and then a cumulative subtraction calculation is performed (the accumulated deformation at the next moment is the accumulated deformation at the next moment × weight minus the accumulated deformation at the previous moment × weight) to obtain the deformation at the prediction moment, thus realizing the deformation prediction of this point. For example, referring to Table 1 above, based on the weights of the given moments 11 and 12, the accumulated deformation at moment 11 is 80, and the accumulated deformation at moment 12 is 102. Then, through a cumulative subtraction calculation, the deformation at moment 11 is restored to 15, and the deformation at moment 12 is 22.

[0074] In practical applications, weights are given according to the contribution correlation rate of the points. This is a well-known technology. Component analysis and principal component analysis methods are usually applied to obtain the contribution correlation rate of points at different times. The contribution correlation rate of each point is then divided by the sum of the contribution correlation rates of all points to normalize the contribution correlation rate of each point, so that the sum of the weights is 1, and the weights of the points at different times are obtained. Of course, the weights can also be reasonably given values ​​according to actual projects without limitation.

[0075] The present invention also proposes a multi-source heterogeneous data acquisition system for acquiring multi-source heterogeneous data in the slope deformation prediction method of the present invention, such as Fig.12 , including a base 5 arranged on the slope, a column 4 is arranged on the base 5, a positioning receiver 2 is installed on the column 4, a reflecting pyramid 1 is installed on the positioning receiver 2, a flexible displacement meter 6 is arranged under the base 5, the flexible displacement meter 6 extends into the set depth of the slope, the flexible displacement meter 6 and the positioning receiver 2 are electrically connected to a collector 3 arranged on the base 5, wherein: the positioning receiver 2 is used to obtain Beidou GNSS data of fixed points on the slope surface; the flexible displacement meter 6 is used to obtain MEMS data of fixed points on the slope surface, and the MEMS data is the internal horizontal displacement data of the slope at different depths in the same vertical direction; the reflecting pyramid 1 is used to assist satellite remote sensing InSAR to obtain InSAR data of permanent scattering points on the slope surface.

[0076] In the present invention, the permanent scattering points on the slope surface include not only the location of the reflection pyramid 1 but also the locations of other objects (such as soil slopes, etc.) located on the slope.

[0077] like Fig.12The base 5 and the column 4 thereon, the positioning receiver 2, the reflection pyramid 1 and the collector 3, and the flexible displacement meter 6 under the base 5 constitute a group of acquisition devices. In actual implementation, multiple groups of acquisition devices are arranged at intervals in the specified area of ​​the slope, wherein: for one group of acquisition devices, the flexible displacement meter 6 and the column 4 are arranged vertically, and the flexible displacement meter 6, the column 4, the positioning receiver 2 and the reflection pyramid 1 are arranged coaxially to ensure that the flexible displacement meter 6, the positioning receiver 2 and the reflection pyramid 1 are in the same position, ensuring that the internal and surface deformations of the slope are consistent.

[0078] In the present invention, the slope refers to the top 9 and the slope surface 8. Specifically, in the art, the slope generally includes the top 9, the slope surface 8 and the bottom 7. The designated area of ​​the slope involved in the present invention refers to a certain area on the top 9 and the slope surface 8, and the bottom 7 may or may not be provided with an acquisition device. In practical applications, each acquisition device should be reasonably arranged on the top 9 and the slope surface 8 according to the shape of the slope, without the need for regular arrangement.

[0079] In the present invention, the flexible displacement meter 6 is a bidirectional flexible displacement meter, which is used to measure the displacement of the slope in the left and right bank directions and the upstream and downstream directions (such as Fig.12 The flexible displacement meter 6 and the positioning receiver 2 are existing devices.

[0080] In actual design, column 4 should have a certain height, which is used to layout the surface permanent scattering points formed by the satellite remote sensing InSAR (interferometric synthetic aperture radar) through the reflection pyramid 1 and the surface fixed points formed by the positioning receiver 2, and ensure that these points are not blocked by the surface vegetation of the slope.

[0081] In the present invention, the collector 3 is used to obtain Beidou GNSS data and MEMS data to ensure synchronous collection of the two data. The collector 3 is a well-known electronic device in the art.

[0082] In the present invention, Beidou GNSS (Global Navigation Satellite System) data may also be referred to as BDS data, which is one of the data included in GNSS (Global Navigation Satellite System).

[0083] The slope deformation prediction method of the present invention realizes comprehensive prediction of slope deformation through multiple monitoring means (BDS, flexible displacement meter, satellite remote sensing InSAR) and multiple prediction models (cluster distribution model, various curve models (such as cumulative curve, cumulative growth curve, etc.)), and the prediction result is highly accurate, scientific and reliable.

[0084] The above are preferred embodiments of the present invention and the technical principles used therein. For those skilled in the art, any obvious changes such as equivalent transformations, simple replacements, etc. based on the technical solution of the present invention, without departing from the spirit and scope of the present invention, are within the scope of protection of the present invention.

Claims

1. A slope deformation prediction method, characterized in that: Includes steps: 1) Obtain multi-source heterogeneous data at each point on the slope, including Beidou GNSS data, MEMS data, and InSAR data; 2) preprocessing the multi-source heterogeneous data; 3) performing clustering processing on the preprocessed multi-source heterogeneous data; 4) Predict the deformation of each point on the slope and complete the slope deformation prediction.

2. The slope deformation prediction method according to claim 1, characterized in that: In the step 1): The Beidou GNSS data is deformation data obtained by BDS positioning difference for each surface fixed point in the specified area of ​​the slope. Each surface fixed point is regarded as a point position. The collection period of the Beidou GNSS data is 1 second to 3600 seconds. The MEMS data is the internal horizontal displacement data of the slope obtained by the flexible displacement meter at different depths in the same vertical direction for each surface fixed point in the specified area of ​​the slope. Each depth position of each surface fixed point in the vertical direction is regarded as a point position. The acquisition period of the MEMS data is 1 second to 3600 seconds. The InSAR data are time-course deformation data obtained by interpreting satellite remote sensing InSAR images for each permanent scattering point on the surface within a specified area of ​​the slope. Each permanent scattering point on the surface is regarded as a point. The collection period of the InSAR data is 12 days.

3. The slope deformation prediction method according to claim 2, characterized in that: The step 2) specifically includes: 2-1) converting all of the InSAR data, the Beidou GNSS data and the MEMS data into the same 2000 national geodetic coordinate system to complete a unified coordinate system; 2-2) Based on the 3S criterion, the InSAR data, the Beidou GNSS data and the MEMS data are denoised to remove data points with a probability of less than 3‰; 2-3) performing interpolation processing on the InSAR data, the Beidou GNSS data and the MEMS data by Newton polynomial interpolation method, and retaining the original change trend and fluctuation characteristics of the data; 2-4) for the collection period of each type of data of the InSAR data, the Beidou GNSS data and the MEMS data, each type of data obtained with 1 times the collection period, 10 times the collection period, 100 times the collection period and 1000 times the collection period is subjected to thinning processing to obtain data sets of different types of data and different periods; 2-5) Perform pairwise difference calculations on the data sets of each period between different types of data, complete the statistics of the combined difference variation of the data sets of different periods, and then select the best data set combination with the smallest difference among the data sets of the same period and different types of data; 2-6) Based on the selected optimal data set combination, the basic combination sequence is recombined for the same period and different types of data; 2-7) Based on the basic combination sequence, the basic combination sequence of each point in the specified area of ​​the slope with different periods is obtained.

4. The slope deformation prediction method according to claim 3, characterized in that: The step 3) specifically includes: 3-1) Based on the energy function and the least squares method, the data in the basic combination sequence of each point in different periods are converted into deformation amount and deformation rate; 3-2) Clustering is performed on the deformation amount, deformation rate and two-dimensional data composed of the deformation amount and deformation rate at different points in the same period, wherein the deformation amount, deformation rate and two-dimensional data composed of the deformation amount and deformation rate are defined as objects, and the objects are composed of sample data; 3-2-1) Randomly assign the sample data of an object at different locations in the same period to form several random clusters; 3-2-2) For each point of sample data, calculate the distance between this sample data and the center point of each cluster, and assign this sample data to the cluster with the closest distance to it to generate a new cluster; 3-2-3) Calculate the center point of the new cluster by calculating the average value of all sample data in the new cluster; 3-2-4) Repeat steps 3-2-2) to 3-2-3) until the center point of the cluster no longer changes significantly or the preset number of iterations is reached; 3-2-5) Generate sample clusters for this object at different locations in the same period; 3-3) Based on the sample clusters generated for each object, the cluster distribution of the basic combination sequence of each point in the specified area of ​​the slope with different periods is obtained.

5. The slope deformation prediction method according to claim 4, characterized in that: The step 4) specifically includes: 4-1) For the same period, all points in the designated area of ​​the slope are divided into multiple partitions according to the cluster distribution of the basic combination sequence of each point; 4-2) Based on the deformation curves of each point in the same partition, the cumulative curve of each point is obtained: the deformation of a point in the partition obtained at different times is cumulatively calculated to obtain the cumulative curve; 4-3) According to the weights of each point in the same partition at different times, the cumulative curve is converted into a growth curve; 4-4) Based on the growth curves of each point in different periods, the growth curves of the same point in different periods are connected end to end to form a cumulative growth curve; 4-5) Perform polynomial fitting on the cumulative growth curve of the same point to fit multiple polynomial curves; 4-6) Calculate the mean square error of each polynomial curve, and select a polynomial curve with the smallest mean square error from all polynomial curves as the prediction curve; 4-7) According to the time interval, the prediction curve of each point is predicted for multiple time periods backward to obtain a cumulative prediction curve, wherein the time of data prediction is recorded as the prediction time; 4-8) According to the weight of each point at the prediction time obtained in step 4-3), the cumulative deformation of the prediction curve of each point at each prediction time is restored, and then the deformation at each prediction time is obtained by a cumulative subtraction calculation; 4-9) The deformation of different points in the specified area of ​​the slope at the prediction time is obtained, so as to realize the deformation prediction of each point of the slope based on the cluster distribution and complete the slope deformation prediction.

6. A multi-source heterogeneous data acquisition system, used to acquire multi-source heterogeneous data in the slope deformation prediction method according to any one of claims 1 to 5, characterized in that: It comprises a base arranged on the slope, a column is arranged on the base, a positioning receiver is installed on the column, a reflecting pyramid is installed on the positioning receiver, a flexible displacement meter is arranged under the base, the flexible displacement meter extends into the slope to a set depth, the flexible displacement meter and the positioning receiver are electrically connected to a collector arranged on the base, wherein: the positioning receiver is used to obtain the Beidou GNSS data of the fixed point on the slope surface; the flexible displacement meter is used to obtain the MEMS data of the fixed point on the slope surface, and the MEMS data is the internal horizontal displacement data of the slope at different depths in the same vertical direction; the reflecting pyramid is used to assist the satellite remote sensing InSAR to obtain the InSAR data of the permanent scattering points on the slope surface.

7. The multi-source heterogeneous data acquisition system according to claim 6, characterized in that: In addition to the location of the reflection pyramid, the permanent scattering points on the slope surface also include the locations of other objects located on the slope.

8. The multi-source heterogeneous data acquisition system according to claim 6, characterized in that: The base and the columns thereon, the positioning receiver, the reflective pyramid and the collector, and the flexible displacement meter under the base constitute a group of acquisition devices, and multiple groups of the acquisition devices are arranged at intervals in a specified area of ​​the slope, wherein: for one group of acquisition devices, the flexible displacement meter and the column are arranged vertically, and the flexible displacement meter, the column, the positioning receiver and the reflective pyramid are arranged coaxially; the slope refers to the top and surface of the slope.

9. The multi-source heterogeneous data acquisition system according to claim 6, characterized in that: The flexible displacement meter is a bidirectional flexible displacement meter, which is used to measure the horizontal displacement inside the slope in the left and right bank directions and the upstream and downstream directions.

Images