High slope monitoring and early warning system based on UAV data verification

Abstract

The present invention provides a high slope monitoring and early warning system based on UAV data verification, including a UAV set up to collect slope images, a pull-wire displacement sensor array pre-buried in the slope, a fixed inclinometer and a seepage monitoring device; the UAV is used to monitor the image data of the slope from the air; the pull-wire displacement sensor array is used to monitor the displacement data of the slope surface; the fixed inclinometer and the seepage monitoring device are used to monitor the internal slip data of the slope; the data of the UAV, the pull-wire displacement sensor, the fixed inclinometer and the seepage monitoring device are used to perform real-time slope safety monitoring. By adopting the above scheme, cracks can be effectively identified, and sensitive information can be further enhanced; secondly, cracks in dangerous areas, subtle movements of dangerous rocks and subtle deformations on the monitoring section can be accurately identified and captured, future data can be predicted, and the real-time status of the slope can be simulated and displayed, so that the monitoring results of the slope are intuitive and vivid.

Description

Technical Field

[0001] The invention relates to the field of high slope monitoring and early warning, in particular to a high slope monitoring and early warning system based on unmanned aerial vehicle data verification. Background Art

[0002] In human activities, such as the construction of railways and highways, slope slip disasters caused by construction disturbances or the impact of natural weather on high slopes often occur. For example, some railways and highways, due to heavy blasting and forced excavation during construction, lost their slope support without monitoring, early warning and timely management, which posed a hazard to traffic and personal safety, caused huge economic losses and adverse social impacts, and even interrupted traffic.

[0003] So far, most slope monitoring and early warning still remain at the level of a single monitoring method and a single-factor static analysis and early warning model, and have not formed a multi-level slope monitoring system and an intelligent early warning model based on dynamic multiple factors. As for the monitoring methods of general slope projects, the relevant data obtained by a single monitoring method often lack accuracy and reference value, which will greatly affect the reliability of the early warning of the entire slope monitoring system and lead to great hidden dangers in the monitoring and safety early warning work of the slope.

[0004] Since the existing landslide monitoring and surveillance technologies focus on single factor monitoring, there is no efficient and advanced multi-level monitoring system to achieve batch monitoring of a large number of high slopes due to the wide range of landslides, inconvenient transportation, high monitoring costs, and difficult analysis. With the emergence of low-cost and highly reliable flying equipment such as drones, traditional monitoring methods have gradually been replaced, and rapid and efficient monitoring of a large area has become a reality; in other words, the use of drone equipment to conduct large-scale monitoring and report and summarize monitoring information through wireless transmission equipment can greatly improve the efficiency of slope landslide monitoring and avoid the occurrence of landslide disasters. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a high slope monitoring and early warning system based on drone data verification, which can combine the multi-dimensional monitoring of current construction needs, adopt multi-level collaborative monitoring in the process of data collection, effectively avoid the inherent errors and accidental errors caused by a single monitoring method, and review each monitoring data through the linear regression relationship of each level of monitoring data to ensure the reliability of the monitoring data.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is: a high slope monitoring and early warning system based on drone data verification, including a drone set to collect slope images, a pull-wire displacement sensor array pre-buried in the slope, a fixed inclinometer and seepage monitoring equipment;

[0007] Drones are used to monitor image data of slopes from the air;

[0008] The pull-wire displacement sensor array is used to monitor the displacement data of the slope surface;

[0009] Fixed inclinometers and seepage monitoring equipment are used to monitor internal slope slip data;

[0010] Real-time slope safety monitoring is carried out using data from drones, wire-type displacement sensors, fixed inclinometers and seepage monitoring equipment.

[0011] In the preferred solution, a rainwater collection device is also provided to collect precipitation data to assist in slope safety monitoring.

[0012] In a preferred solution, the pull-wire displacement sensor array, the fixed inclinometer, the seepage monitoring device and the rainwater collection device are electrically connected to the data acquisition device to collect monitoring data;

[0013] The data acquisition device is connected to the monitoring cloud platform via wireless means, and the drone is connected to the monitoring cloud platform via wireless means.

[0014] In the preferred solution, the image data of the drone aerial photography is subjected to a one-dimensional wavelet transform on the pixel data of the image row by row, and is decomposed into two component outputs of a low-pass filter L and a high-pass filter H;

[0015] Then, the pixel data of the image is subjected to one-dimensional wavelet transform column by column, and decomposed into four component outputs: LL, LH, HL, and HH, to highlight sensitive data;

[0016] The sensitive data are image data related to cracks in the slope.

[0017] In a preferred solution, the pull-wire displacement sensor array is subjected to Kriging interpolation operation to obtain ground monitoring data.

[0018] In the preferred solution, the ground monitoring data and the image data of the drone are tested by the F test method. If the test result meets the requirements, the optimal solution is output; if not, the surface function in the spatial interpolation is replaced by the AI ​​learning method until the F test method is met and the optimal solution is output;

[0019] The optimal solution is the optimized surface data.

[0020] In the preferred solution, the data obtained by the rainwater collection device, the seepage monitoring equipment and the fixed inclinometer are used as the internal data of the optimized slope;

[0021] The optimized surface data and the optimized internal data of the slope are visualized to establish a monitoring and early warning model.

[0022] In the preferred solution, the monitoring and early warning model includes a prediction module and an evaluation module;

[0023] The prediction module has the following steps:

[0024] S01. Establish monitoring of factors affecting slope stability;

[0025] S02, preprocessing the monitored data;

[0026] S03, establish BP neural network prediction model;

[0027] S04. Predict future data;

[0028] The evaluation module has the following steps:

[0029] S11, Grey correlation analysis;

[0030] S12, establish fuzzy clustering iterative model;

[0031] S13, introducing predicted future data into the fuzzy clustering iteration model;

[0032] S14. Output the evaluated slope status.

[0033] In a preferred embodiment, the factors of the fuzzy clustering iterative model include bulk density, cohesion, internal friction angle, slope angle, slope height and pore pressure ratio.

[0034] In the preferred scheme, in step S14, historical monitoring data of the slope and basic hydrological and geological data are used as sample data, and corresponding eigenvalue matrix samples and relative membership matrices are constructed according to the eigenvalues ​​of each factor in the sample. The weights of each influencing factor are analyzed using grey correlation analysis, and the optimal fuzzy clustering matrix and the optimal fuzzy clustering center matrix are solved through iterative operations. After sorting, each relative membership is multiplied by the corresponding evaluation level value to obtain the level of stability evaluation.

[0035] The present invention provides a high slope monitoring and early warning system based on drone data verification. By adopting the above scheme, firstly, cracks can be effectively identified, and sensitive information can be further strengthened; secondly, cracks in dangerous areas, subtle movements of dangerous rocks and subtle deformations on the monitoring section can be accurately identified and captured, and future data can be predicted through visualization technology combined with a prediction model based on neural network prediction, and the real-time status of the slope can be simulated and displayed, so that the monitoring results of the slope are intuitive and vivid. Thirdly, an intelligent monitoring and early warning model based on multi-factor coupling is used for early warning reporting; this system can not only effectively solve the problem of low data accuracy, but also monitor and report on slope displacement changes in real time, effectively avoid landslide hazards, and greatly reduce economic losses and personnel safety risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The present invention will be further described below in conjunction with the accompanying drawings and embodiments:

[0037] Figure 1 It is a schematic diagram of the on-site arrangement of the slope of the present invention.

[0038] Figure 2 It is a schematic diagram of the process structure of the present invention.

[0039] Figure 3 It is a relationship diagram of the collected data, internal factors and external factors of the present invention.

[0040] Figure 4 This is the flow chart of the BP neural network prediction model.

[0041] In the figure: UAV 1, wire-type displacement sensor 2, rainwater collection device 3, seepage monitoring equipment 4, fixed inclinometer 5, data acquisition device 6, slope 7. DETAILED DESCRIPTION

[0042] like Figure 1 In the invention, a high slope monitoring and early warning system based on UAV data verification includes a UAV 1 configured to collect images of a slope 7, an array of pull-wire displacement sensors 2 pre-buried in the slope 7, a fixed inclinometer 5, and a seepage monitoring device 4;

[0043] According to the needs of the project, the wire displacement meter and the fixed inclinometer are evenly distributed in the slope area, and the remote monitoring platform is arranged in the area with convenient transportation network. The layout of the wire displacement meter is as follows: Vertical displacement installation method: Fix the wire displacement sensor 2 above the slope surface, pull the wire rope from top to bottom, control the pre-tension length to about 50% of the range, and fix the pull-out end by means of expansion bolts. If the wire pulling distance is too long, the wire rope will fall seriously. You can add a pulley every 5 meters or so to pull the wire rope, tighten the wire rope, and reduce the measurement error. Horizontal displacement installation method: Fix the sensor on the slope surface, pull the wire rope horizontally, control the pre-tension length to about 50% of the range, and fix the pull-out end by means of expansion bolts. If the wire pulling distance is too long, the wire rope will fall seriously. You can add a pulley every 5 meters or so to pull the wire rope, tighten the wire rope, and reduce the measurement error.

[0044] The installation method of the fixed inclinometer 5 is:

[0045] Installation of inclinometer pipe: First install the bottom cover of the inclinometer pipe and fix it with screws or glue. The inclinometer pipes are connected with pipe joints. The inclinometer pipes and pipe joints must be fixed with screws and then sealed with glue. When installing the inclinometer pipe, attention should be paid to the direction of the guide groove, which must be consistent with the direction required by the design. When confirming that the inclinometer pipe is installed well, backfilling can be carried out. Backfilling generally uses bentonite balls or original soil sand. When backfilling, water should be injected every 3 to 5 meters. The purpose of water injection is to make the bentonite balls or original soil sand firmly bonded to the hole wall after encountering water, and this method is used until the hole mouth. The inclinometer pipe exposed on the surface should be well protected and covered with a pipe cover to prevent objects from falling in. The surface pipe mouth section of the inclinometer pipe should be poured with concrete and made into a concrete pier to protect the stability of the pipe mouth and its corner. Surveying and mapping points should be set on the pier.

[0046] Installation of the instrument

[0047] First, check whether the guide wheel of the inclinometer rotates flexibly and whether the torsion spring is strong. Check whether the sensor components are working properly, cut the connecting rod according to the design elevation, and connect the fixed inclinometer end to end with a wire rope. After confirming that it is intact, prepare for installation. When assembling the fixed inclinometer, it should be assembled into measuring units according to the number required by the construction drawings. After checking and confirming that it is intact, prepare for hoisting. When hoisting, put the measuring units into the inclinometer tube in order. The connection between each measuring unit is firmly connected with a connecting rod, and all the guide wheels of each measuring unit are in the same direction. Each set of fixed inclinometers should be numbered and recorded in order. When installing one by one, the cables should be straightened one by one. All cables should be loose and not tightened. The last connecting rod should be tied to the horizontal axis of the orifice device and locked with a lock. Finally, protective facilities should be set up at the orifice. The observation cable is fixed and buried according to the specified direction.

[0048] The seepage monitoring device 4 is arranged inside the slope 7 .

[0049] In order to ensure the effectiveness of the data connection between the drone and the remote monitoring cloud platform, and considering the complex environment of the construction area and the weak network card signal, the drone is equipped with a USB wireless network card with a high-power wifi module. The drone lithium battery is a high-performance lithium battery with a capacity of no less than 40,000 mAh.

[0050] The daily data collection frequency of the entire monitoring system is once a day. Under adverse conditions such as rainfall, the data collection can be encrypted to once an hour. The mobile network uses 4G or above to ensure the smooth flow of real-time data. The frequency of automatic monitoring by drones is 2 times a week.

[0051] The drone 1 is used to monitor the image data of the slope 7 from the air;

[0052] The array of wire-type displacement sensors 2 is used to monitor the displacement data of the surface of the slope 7;

[0053] The fixed inclinometer 5 and the seepage monitoring device 4 are used to monitor the internal slip data of the slope 7;

[0054] Real-time slope safety monitoring is performed using data from the drone 1 , the wire-type displacement sensor 2 , the fixed inclinometer 5 , and the seepage monitoring equipment 4 .

[0055] In the preferred solution, a rainwater collection device 3 is also provided to collect precipitation data to assist in slope safety monitoring. This structure improves the full coverage of monitoring data, by deploying drones above the high slope, deploying wire displacement meters and rainwater collectors on the surface of the high slope, and deploying fixed inclinometers inside, to coordinate monitoring from three dimensions, thereby ensuring the comprehensiveness of the data and effectively avoiding the inherent errors and accidental errors caused by a single monitoring method.

[0056] In the preferred solution, the array of wire displacement sensors 2, the fixed inclinometer 5, the seepage monitoring device 4 and the rainwater collection device 3 are electrically connected to the data acquisition device 6 to collect monitoring data;

[0057] The data acquisition device 6 is connected to the monitoring cloud platform via wireless, and the drone 1 is connected to the monitoring cloud platform via wireless.

[0058] In the preferred solution, the image data of the aerial photography of the drone 1 is subjected to a one-dimensional wavelet transform on the pixel data of the image row by row, and is decomposed into two component outputs of a low-pass filter L and a high-pass filter H;

[0059] Then, the pixel data of the image is subjected to one-dimensional wavelet transform column by column, and decomposed into four component outputs: LL, LH, HL, and HH, to highlight sensitive data;

[0060] The sensitive data are image data related to the cracks of the slope 7 .

[0061] In the preferred solution, the array of the wire-type displacement sensors 2 is subjected to Kriging interpolation operation to convert the surface point displacement data into surface displacement data to obtain ground monitoring data.

[0062] In the preferred solution, the ground monitoring data and the image data of the UAV 1 are tested by the F test method. If the test result meets the requirements, the optimal solution is output; if not, the surface function in the spatial interpolation is replaced by the AI ​​learning method until the F test method is met and the optimal solution is output; to ensure the accuracy of the above-mentioned spatial interpolation data and the UAV data calculation, it is necessary to use the AI ​​deep learning algorithm to continuously optimize and reconstruct the random surface function in the Kriging interpolation method. Furthermore, to ensure the accuracy of the data, it is necessary to use the surface monitoring data and the UAV monitoring data of a fixed time period as samples, and the historical monitoring data of the slope as a statistical sample. Through correlation analysis and the F test method, a regression model for collaborative monitoring review is established, and each monitoring data is reviewed and corrected based on this model.

[0063] The optimal solution is the optimized surface data.

[0064] In the preferred solution, the data obtained by the rainwater collection device 3, the seepage monitoring device 4 and the fixed inclinometer 5 are used as the internal data of the optimized slope 7;

[0065] The optimized surface data and the optimized internal data of slope 7 are visualized to establish a monitoring and early warning model. To ensure that the monitoring results are intuitive and vivid, the processed data is presented in the form of three-dimensional visualization modeling, with the following beneficial effects: first, cracks can be effectively identified; second, cracks in dangerous areas, subtle movements of dangerous rocks, and subtle deformations on the monitoring section can be accurately identified and captured, and the real-time status of the slope is simulated and displayed through visualization technology, making the monitoring results of the slope clearer. In order to improve the accuracy of image visualization, the convolutional neural network model is combined with the Faster-R-CNN algorithm to effectively improve the image recognition speed and accuracy.

[0066] In the preferred solution, the monitoring and early warning model includes a prediction module and an evaluation module;

[0067] The prediction module has the following steps:

[0068] S01. Establish monitoring of factors affecting slope stability;

[0069] The slope is a complex system, and slope stability analysis is also a complex problem. At present, slope stability research generally studies the slope stability indirectly by studying the factors that affect the slope stability. Since there are many factors that affect the stability of the slope, we should first determine which factors are the main factors affecting the slope stability. Due to the complexity of landslides, the development and occurrence of landslides are affected by various factors. Generally speaking, they are mainly divided into internal factors and external factors. The characteristics of internal factors are slow changes; the characteristics of external factors are significant changes. Internal factors are also called geological and geomorphological background factors, and external factors are also called non-geological factors. They are the impact of changes in the external environment on slope stability. Internal factors include stratum lithology, geological structure, etc., which are the inherent geological characteristics of the slope, and they constitute the main body of the entire slope. Internal factors are the basic conditions for the formation and development of landslides, and are characterized by slow and few changes, while external factors are catalysts for the formation of landslides. Some external factors may accelerate the formation of landslides, and some external factors may inhibit the formation of slopes. The characteristics of external factors are rapid changes and many mutations. For example Figure 3 as shown in .

[0070] At the current stage, the research generally converts the internal and external factors affecting the slope into the study of bulk density, cohesion, internal friction angle, slope angle, slope height, and pore pressure ratio. Therefore, the main factors in the fuzzy clustering iteration model are determined as bulk density, cohesion, internal friction angle, slope angle, slope height, and pore pressure ratio to study the stability of the slope, which can essentially predict the stability of the slope.

[0071] S02, preprocessing the monitored data;

[0072] Starting from the composite model, the main factors affecting the stability of the slope are determined to be bulk density, cohesion, internal friction angle, slope angle, slope height, and pore pressure ratio. The data monitored by the A-section pull-wire displacement meter is selected as the displacement as the monitoring result as the research object, so the soil sampling point is also near the A-section pull-wire displacement meter, and the soil sampling depth is 25 meters. The soil at 4 meters, 10 meters, 16 meters, and 22 meters is selected as the sample soil sample, and four groups of monitoring values ​​of bulk density, cohesion, internal friction angle, slope angle, and pore pressure ratio are obtained. The average value of the four is taken as the final monitoring value. The data of the six influencing factors of the slope are collected, and the initial data is denoised using the MATLAB wavelet analysis toolbox.

[0073] S03, establish BP neural network prediction model;

[0074] The BP neural network is roughly divided into three steps, namely neuron input, neuron training and neuron output. By inputting the initial data of the original factors affecting slope stability as the neuron input and using the model for training, the trend of future changes in various factors can be predicted, and each prediction result is used as the neuron output.

[0075] By using MATLAB software and programming according to the above steps, data prediction of various factors affecting slope stability can be achieved, such as Figure 4 as shown in .

[0076] S04. Predict future data;

[0077] After obtaining the noise reduction data of the six factors affecting slope stability, the next step is to predict the future changes of each factor. According to the BP neural network programming principles, seven predictions are made for bulk density, cohesion, internal friction angle, slope angle, slope height, and pore pressure ratio.

[0078] The weights of the main factors affecting slope stability are calculated, and the arc sliding data of the soil slope in section A is used as a sample to calculate the weights of the main factors affecting the slope. The main influencing parameters are bulk density, cohesion, internal friction angle, slope angle, slope height, and pore pressure ratio.

[0079] The evaluation module has the following steps:

[0080] S11, Grey correlation analysis;

[0081] Grey correlation analysis theory is used to study the relationship between two or more factors in a system, which changes with time or different objects. This measure used to describe the size of the correlation is called correlation. The slope evaluation error caused by subjectivity requires the introduction of a new weight determination method. Grey correlation analysis theory is to study the size of the correlation between each influencing factor and the subject. Obviously, the greater the correlation between the factor and the subject, the greater the influence of the factor on the subject. Therefore, the grey correlation theory can be introduced to determine the weight of each factor.

[0082] S12, establishing a fuzzy clustering iterative model;

[0083] Fuzzy clustering is used to solve the problem of knowing a large amount of historical sample data. According to the eigenvalues ​​of each factor given in the sample, the corresponding eigenvalue matrix is ​​constructed using their eigenvalues, the eigenvalues ​​of the corresponding level of each factor are calculated according to the given relative membership, and the categories are divided according to their eigenvalues.

[0084] S13, introducing predicted future data into the fuzzy clustering iteration model;

[0085] The historical data of the six factors that have been determined to affect slope stability are trained through the BP neural network model, and predictions are made based on this. The prediction results are input into the evaluation module, and the fuzzy clustering iterative model is introduced.

[0086] S14. Output the evaluated slope status.

[0087] The slope stability is evaluated according to the constructed fuzzy clustering iterative model, and the calculation process is realized by self-programming using MATLAB. The optimal fuzzy clustering matrix can be obtained by the improved fuzzy clustering iterative model calculation program. After obtaining the optimal fuzzy clustering matrix, the slope stability state needs to be divided. In general, the slope stability state can be divided into three levels, the first level is stable, the second level is general, and the third level is damaged. Then the relative level characteristic vector of each cluster sample can determine the optimal fuzzy clustering matrix. According to the corresponding divided area, after obtaining the corresponding evaluation value of the sample slope, the stability of the slope can be judged.

[0088] In order to fully consider the influence of the monitored displacement and crack development as well as the original geological factors and natural conditions of the slope in the stability evaluation, it is proposed to adopt the fuzzy clustering iterative model as the basic model for stability assessment, and construct a composite model of BP neural network model and fuzzy clustering iterative model as the early warning module.

[0089] In the preferred scheme, in step S14, historical monitoring data of the slope and basic hydrological and geological data are used as sample data, and corresponding eigenvalue matrix samples and relative membership matrices are constructed according to the eigenvalues ​​of each factor in the sample. The weights of each influencing factor are analyzed using grey correlation analysis, and the optimal fuzzy clustering matrix and the optimal fuzzy clustering center matrix are solved through iterative operations. After sorting, each relative membership is multiplied by the corresponding evaluation level value to obtain the level of stability evaluation.

[0090] Landslides are related to rainfall. According to historical data, the number of landslides increases significantly when the cumulative rainfall on the same day is greater than 100mm. Therefore, when the cumulative rainfall on the same day is greater than 100mm, the monitoring cloud platform should enter an alert state and increase the frequency of slope monitoring. When the road traffic and communication conditions are met for drone monitoring on the next day or a certain day thereafter, all monitoring instruments should be checked accordingly according to the set early warning thresholds, and drone coordinated monitoring should be carried out on the day when the conditions for drone monitoring are met.

[0091] The above embodiments are only preferred technical solutions of the present invention and should not be regarded as limitations of the present invention. The technical features recorded in the present invention can be used in combination with each other without conflict. The protection scope of the present invention shall be the technical solutions recorded in the claims, including the equivalent replacement solutions of the technical features in the technical solutions recorded in the claims. That is, equivalent replacement improvements within this scope are also within the protection scope of the present invention.

Claims

1. A high slope monitoring and early warning system based on drone data verification, characterized by: The invention comprises an unmanned aerial vehicle (1) for collecting images of a slope (7), an array of pull-wire displacement sensors (2) pre-buried in the slope (7), a fixed inclinometer (5) and a seepage monitoring device (4); the unmanned aerial vehicle (1) is used to monitor image data of the slope (7) from the air; the array of pull-wire displacement sensors (2) is used to monitor displacement data on the surface of the slope (7); the fixed inclinometer (5) and the seepage monitoring device (4) are used to monitor internal slip data of the slope (7); and real-time slope safety monitoring is performed using data from the unmanned aerial vehicle (1), the pull-wire displacement sensors (2), the fixed inclinometer (5) and the seepage monitoring device (4); A rainwater collection device (3) is also provided to collect precipitation data to assist in slope safety monitoring; The rainwater collection device (3), the seepage monitoring device (4) and the fixed inclinometer (5) obtain data as internal data of the optimized slope (7); the optimized surface data and the internal data of the optimized slope (7) are used to establish a monitoring and early warning model in a visual manner; (1) The image data taken by the drone is subjected to one-dimensional wavelet transform row by row, and decomposed into two components: low-pass filter L and high-pass filter H. Then, one-dimensional wavelet transform is performed on the pixel data of the image column by column, and decomposed into four components: LL, LH, HL, and HH, to highlight sensitive data. The sensitive data is image data related to cracks in the slope (7); The array of the wire-type displacement sensor (2) is subjected to Kriging interpolation operation to obtain ground monitoring data; The ground monitoring data and the image data from the drone (1) are tested by the F test method. If the test result satisfies the requirement, the optimal solution is output. If it is not satisfied, the surface function in the spatial interpolation is replaced again through the AI ​​learning method until the F test method is satisfied, and the optimal solution is output; the optimal solution is the optimized surface data; The monitoring and early warning model includes a prediction module and an evaluation module; The prediction module has the following steps: S01, establish monitoring of factors affecting slope stability; S02, preprocessing the monitored data; Based on the composite model, the main factors affecting slope stability are determined to be bulk density, cohesion, internal friction angle, slope angle, slope height and pore pressure ratio. The data of the six influencing factors of the slope are collected, and the initial data are denoised using the MATLAB wavelet analysis toolbox. S03, establish BP neural network prediction model; The BP neural network is divided into three steps, namely, neuron input, neuron training and neuron output. The initial data of the factors affecting slope stability are input as neuron input, and the model is used for training to predict the changing trend of various factors in the future. Each prediction result is used as neuron output. S04. Predict future data; After obtaining the noise reduction data of the six factors affecting slope stability, the future changes of each factor were predicted. According to the programming principle of BP neural network, seven backward predictions were made for bulk density, cohesion, internal friction angle, slope angle, slope height and pore pressure ratio, and the weights of the main factors affecting slope stability were calculated. The evaluation module has the following steps: S11, Grey correlation analysis; Grey correlation analysis is used to study the correlation between each influencing factor and the subject. The greater the correlation between the factor and the subject, the greater the influence of the factor on the subject. Grey correlation theory is introduced to determine the weight of each factor. S12, establish fuzzy clustering iterative model; The factors of the fuzzy clustering iteration model include bulk density, cohesion, internal friction angle, slope angle, slope height and pore pressure ratio; Fuzzy clustering is used to solve the problem of knowing a large amount of historical sample data. According to the eigenvalues ​​of each factor given in the sample, the corresponding eigenvalue matrix is ​​constructed using their eigenvalues. According to the given relative membership, the eigenvalues ​​of the corresponding level of each factor are calculated, and the categories are divided according to their eigenvalues. S13, introducing predicted future data into the fuzzy clustering iteration model; The historical data of the six factors that have been determined to affect slope stability are trained through the BP neural network model, and predictions are made based on this. The prediction results are input into the evaluation module, and the fuzzy clustering iterative model is introduced. S14, outputting the evaluated slope status; The historical monitoring data of slopes and basic hydrological and geological data are used as sample data. The corresponding eigenvalue matrix samples and relative membership matrix are constructed according to the eigenvalues ​​of each factor in the sample. The weights of each influencing factor are analyzed using grey correlation analysis. The optimal fuzzy clustering matrix and the optimal fuzzy clustering center matrix are solved through iterative operations. After sorting, each relative membership is multiplied by the corresponding evaluation level value to get the stability evaluation grade.

2. The high slope monitoring and early warning system based on drone data verification according to claim 1 is characterized by: The array of wire-type displacement sensors (2), the fixed inclinometer (5), the seepage monitoring equipment (4) and the rainwater collection device (3) are electrically connected to the data acquisition device (6) to collect monitoring data; the data acquisition device (6) is connected to the monitoring cloud platform by wireless means, and the drone (1) is connected to the monitoring cloud platform by wireless means.

Images