Early warning method and system for side slope geological disaster based on convolutional neural network

Abstract

The invention relates to the technical field of geological disasters, in particular to a slope geological disaster early warning method and system based on a convolutional neural network. Firstly, image data, meteorological data and seismic data are acquired; secondly, inputting the image data into a slope environment feature recognition model for recognition and evaluation to obtain slope features, underground water level features and environment distribution features; then, evaluating the slope characteristics, the underground water level characteristics and the environment distribution characteristics, and calculating to obtain a first early warning value and a second early warning value; weighting the first early warning value and the second early warning value to obtain a comprehensive early warning value; and finally, comparing the comprehensive early warning value with a comprehensive early warning threshold value, if the comprehensive early warning value exceeds the threshold value, outputting early warning prompt information, otherwise, inputting real-time monitoring data into the side slope state prediction model, and generating side slope state early warning information by using a data result of the side slope state prediction model.

Description

Technical Field

[0001] The present invention relates to the technical field of geological disasters, and in particular to an early warning method and system for slope geological disasters based on a convolutional neural network. Background Art

[0002] Slope geological disasters are geological disasters caused by changes in the natural environment and human activities. Common types include landslides, mudslides and collapses. These geological disasters usually occur in areas with complex geological environments, changeable climate conditions or frequent human activities. Slope geological disasters not only pose a threat to human life and economic development, but also have long-term impacts on the surrounding ecological environment. Therefore, it is crucial to effectively monitor and warn of slope geological disasters.

[0003] Slope geological disaster early warning has always been an important research object, especially for areas with high and steep slopes and areas with frequent natural disasters. Traditional geological disaster early warning methods and systems mostly focus on the monitoring of a single indicator, but in the face of complex and changeable natural and geological environments, these technologies cannot deeply explore the correlation between data, and often cannot comprehensively evaluate and warn of slope geological disaster risks, which will reduce the accuracy and reliability of slope geological disaster early warning.

[0004] Therefore, an early warning method and system for slope geological disasters based on convolutional neural network is proposed. Summary of the invention

[0005] The purpose of the present invention is to provide a convolutional neural network-based early warning method and system for slope geological disasters; first, image data, meteorological data and seismic data are obtained; secondly, the image data is input into a slope environmental feature recognition model for identification and evaluation to obtain slope characteristics, groundwater level characteristics and environmental distribution characteristics; then, the slope characteristics, the groundwater level characteristics and the environmental distribution characteristics are evaluated to calculate a first early warning value and a second early warning value; the first early warning value and the second early warning value are weighted to obtain a comprehensive early warning value; finally, the comprehensive early warning value is compared with a comprehensive early warning threshold value, if the threshold value is exceeded, an early warning prompt message is output, otherwise the real-time monitoring data is input into the slope state prediction model, and the slope state early warning information is generated using the data results of the slope state prediction model.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A convolutional neural network-based early warning method for slope geological disasters, comprising:

[0008] Acquire slope image data, environmental image data, meteorological data and seismic data;

[0009] Construct a slope environmental feature recognition model and slope state prediction model based on convolutional neural network;

[0010] Inputting the slope image data and the environmental image data into the slope environmental feature recognition model for recognition, and obtaining slope deformation features, slope soil features, groundwater level features, environmental soil distribution features, environmental vegetation distribution features, and environmental building distribution features;

[0011] Evaluate the environmental soil distribution characteristics, the environmental vegetation distribution characteristics, and the environmental building distribution characteristics to obtain a first warning value;

[0012] Evaluate the meteorological data, the seismic data, the slope deformation characteristics, the slope soil characteristics and the groundwater level characteristics to obtain a second warning value;

[0013] The first warning value and the second warning value are weighted with the first warning weight and the second warning weight respectively to obtain a comprehensive warning value;

[0014] The comprehensive warning value is compared with the comprehensive warning threshold. If it exceeds the comprehensive warning threshold, a warning message is sent to the staff; otherwise, the real-time monitoring data is input into the slope state prediction model, and the slope state warning message is generated using the data results of the slope state prediction model.

[0015] Furthermore, the slope image data and the environment image data are input into the slope environment feature recognition model for recognition, and the specific implementation process of obtaining slope deformation features, slope soil features, groundwater level features, environmental soil distribution features, environmental vegetation distribution features and environmental building distribution features includes:

[0016] Acquire slope image data and environment image data;

[0017] Further, the slope image data and the environment image data are subjected to denoising processing to obtain denoised image data;

[0018] Further, the contrast and brightness of the denoised image data are adjusted using histogram equalization to obtain enhanced image data;

[0019] Furthermore, the enhanced image data is input into a slope environmental feature recognition model for processing to obtain slope deformation features, slope soil features, groundwater level features, environmental soil distribution features, environmental vegetation distribution features and environmental building distribution features.

[0020] Furthermore, the specific implementation process of evaluating the environmental soil distribution characteristics, the environmental vegetation distribution characteristics and the environmental building distribution characteristics to obtain the first warning value includes:

[0021] Obtain environmental soil distribution characteristics, environmental vegetation distribution characteristics and environmental building distribution characteristics output by the slope environmental characteristic recognition model;

[0022] Further, the environmental soil distribution area, the environmental vegetation distribution area and the environmental building distribution area are obtained from the environmental soil distribution characteristics, the environmental vegetation distribution characteristics and the environmental building distribution characteristics;

[0023] Further, the environmental soil distribution area and the environmental vegetation distribution area are respectively subjected to difference calculation with the initial environmental soil distribution area and the initial environmental vegetation distribution area to obtain an environmental soil distribution difference and an environmental vegetation distribution difference;

[0024] Furthermore, a weighted evaluation is performed on the environmental soil distribution difference, the environmental vegetation distribution difference and the environmental building distribution area to obtain a first warning value.

[0025] Furthermore, the specific implementation process of evaluating the meteorological data, the seismic data, the slope deformation characteristics, the slope soil characteristics and the groundwater level characteristics to obtain the second warning value includes:

[0026] Collect meteorological and seismic data;

[0027] Further, the slope deformation characteristics, slope soil characteristics and groundwater level characteristics output by the slope environmental characteristics recognition model are obtained;

[0028] Further, precipitation information, historical earthquake frequency information, slope deformation information, slope soil moisture information and groundwater level information are obtained from the meteorological data, the earthquake data, the slope deformation characteristics, the slope soil characteristics and the groundwater level characteristics respectively;

[0029] Furthermore, a weighted evaluation is performed on the precipitation information, the historical earthquake frequency information, the slope deformation information, the slope soil moisture information and the groundwater level information to calculate a second warning value.

[0030] Furthermore, the specific implementation process of inputting the real-time monitoring data into the slope state prediction model and using the data results of the slope state prediction model to generate slope state warning information includes:

[0031] Obtain historical monitoring data from the historical database;

[0032] Furthermore, real-time monitoring data is obtained; wherein the real-time monitoring data includes: real-time slope monitoring data and real-time environmental monitoring data;

[0033] Further, the historical monitoring data is input into a slope state prediction model for training to obtain state impact prediction parameters;

[0034] Further, the state influence prediction parameter, the real-time slope monitoring data and the real-time environment monitoring data are input into the slope state prediction model, the model parameters are updated, and the slope state influence parameter is output; wherein the slope state influence parameter includes: a first slope influence parameter and a second environment influence parameter;

[0035] Furthermore, the real-time monitoring information and the slope state influencing parameters are comprehensively evaluated to obtain slope state warning information.

[0036] A convolutional neural network-based early warning system for slope geological disasters, comprising: a system control module, a data acquisition module, a data processing module and an early warning output module;

[0037] Wherein, the system control module is used to control the start, pause and stop of the device;

[0038] The data acquisition module is used to obtain slope image data, environmental image data, meteorological data, seismic data and historical monitoring data in the historical database;

[0039] The data processing module is used to process the slope image data, the environmental image data, the meteorological data and the seismic data to calculate a comprehensive warning value; wherein the data processing module includes: a first processing unit and a second processing unit;

[0040] The warning output module is used to evaluate the real-time monitoring data, output the warning results and provide warning feedback; wherein, the warning output module includes: a warning judgment unit, a warning prompt unit and a slope status warning unit.

[0041] Furthermore, the data processing module inputs the slope image data and the environmental image data into the slope environmental feature recognition model for recognition, and obtains the slope deformation feature, slope soil feature, groundwater level feature, environmental soil distribution feature, environmental vegetation distribution feature and environmental building distribution feature. The specific implementation process includes:

[0042] Acquire slope image data and environment image data;

[0043] Further, the slope image data and the environment image data are subjected to denoising processing to obtain denoised image data;

[0044] Further, the contrast and brightness of the denoised image data are adjusted using histogram equalization to obtain enhanced image data;

[0045] Furthermore, the enhanced image data is input into a slope environmental feature recognition model for processing to obtain slope deformation features, slope soil features, groundwater level features, environmental soil distribution features, environmental vegetation distribution features and environmental building distribution features.

[0046] Furthermore, the data processing module evaluates the environmental soil distribution characteristics, environmental vegetation distribution characteristics and environmental building distribution characteristics, and the specific implementation process of obtaining the first warning value includes:

[0047] Obtain environmental soil distribution characteristics, environmental vegetation distribution characteristics and environmental building distribution characteristics output by the slope environmental characteristic recognition model;

[0048] Further, the environmental soil distribution area, the environmental vegetation distribution area and the environmental building distribution area are obtained from the environmental soil distribution characteristics, the environmental vegetation distribution characteristics and the environmental building distribution characteristics;

[0049] Further, the environmental soil distribution area and the environmental vegetation distribution area are respectively subjected to difference calculation with the initial environmental soil distribution area and the initial environmental vegetation distribution area to obtain an environmental soil distribution difference and an environmental vegetation distribution difference;

[0050] Furthermore, a weighted evaluation is performed on the environmental soil distribution difference, the environmental vegetation distribution difference and the environmental building distribution area to obtain a first warning value.

[0051] Furthermore, the data processing module evaluates the meteorological data, the seismic data, the slope deformation characteristics, the slope soil characteristics and the groundwater level characteristics to obtain the second warning value. The specific implementation process includes:

[0052] Collect meteorological and seismic data;

[0053] Further, the slope deformation characteristics, slope soil characteristics and groundwater level characteristics output by the slope environmental characteristics recognition model are obtained;

[0054] Further, precipitation information, historical earthquake frequency information, slope deformation information, slope soil moisture information and groundwater level information are obtained from the meteorological data, the earthquake data, the slope deformation characteristics, the slope soil characteristics and the groundwater level characteristics respectively;

[0055] Furthermore, a weighted evaluation is performed on the precipitation information, the historical earthquake frequency information, the slope deformation information, the slope soil moisture information and the groundwater level information to calculate a second warning value.

[0056] Furthermore, the slope state warning unit inputs the real-time monitoring data into the slope state prediction model, and the specific implementation process of generating slope state warning information using the data results of the slope state prediction model includes:

[0057] Obtain historical monitoring data from the historical database;

[0058] Furthermore, real-time monitoring data is obtained; wherein the real-time monitoring data includes: real-time slope monitoring data and real-time environmental monitoring data;

[0059] Further, the historical monitoring data is input into a slope state prediction model for training to obtain state impact prediction parameters;

[0060] Further, the state influence prediction parameter, the real-time slope monitoring data and the real-time environment monitoring data are input into the slope state prediction model, the model parameters are updated, and the slope state influence parameter is output; wherein the slope state influence parameter includes: a first slope influence parameter and a second environment influence parameter;

[0061] Furthermore, the real-time monitoring information and the slope state influencing parameters are comprehensively evaluated to obtain slope state warning information.

[0062] Compared with the prior art, the present invention has the following beneficial effects:

[0063] 1. The present invention proposes a man-made environment early warning function for early warning of man-made environmental changes in slope geological disasters; this function identifies and processes image data through a slope environmental feature recognition model to obtain environmental soil distribution features, environmental vegetation distribution features and environmental building distribution features; this function evaluates and calculates different environmental distribution data to obtain a first early warning value; the first early warning value reflects the degree of influence of the changing environment caused by human activities on the occurrence of slope geological disasters; this function can effectively improve the accuracy and reliability of slope geological disaster early warnings.

[0064] 2. The present invention proposes a natural environment early warning function for early warning of natural environment changes for slope geological disasters; this function identifies and processes image data through a slope environment feature recognition model to obtain slope features and groundwater level features; this function evaluates and calculates meteorological data, seismic data, the slope features and the groundwater level features to obtain a second early warning value; the second early warning value reflects the degree of influence of the changing environment caused by natural activities on the occurrence of slope geological disasters; this function can effectively improve the accuracy and reliability of early warning of slope geological disasters.

[0065] 3. The present invention proposes a slope status warning function for warning the real-time status of the slope; this function uses historical monitoring data to train the slope status prediction model to obtain status impact prediction parameters; the model parameters are updated using the status impact prediction parameters and real-time monitoring data to obtain slope status impact parameters; this function conducts a comprehensive evaluation of the real-time monitoring information and the slope status impact parameters to obtain slope status warning information; the slope status warning information reflects the degree of impact of the real-time environment on the slope status; this function can effectively improve the accuracy and reliability of slope geological disaster warnings. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 A schematic diagram of a flow chart of a method for early warning of geological disasters on slopes based on a convolutional neural network according to the present invention;

[0067] Figure 2 It is a structural schematic diagram of the slope environmental feature recognition model of the present invention;

[0068] Figure 3 It is a structural schematic diagram of the slope state prediction model of the present invention;

[0069] Figure 4 The present invention is a structural schematic diagram of an early warning system for slope geological disasters based on a convolutional neural network. DETAILED DESCRIPTION

[0070] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0071] Slope geological disasters are geological disasters caused by changes in the natural environment and human activities. Common types include landslides, mudslides and collapses. These geological disasters usually occur in areas with complex geological environments, changeable climate conditions or frequent human activities. Slope geological disasters not only pose a threat to human life and economic development, but also have long-term impacts on the surrounding ecological environment. Therefore, it is crucial to effectively monitor and warn of slope geological disasters.

[0072] Slope geological disaster early warning has always been an important research object, especially for areas with high and steep slopes and areas with frequent natural disasters. Traditional geological disaster early warning methods and systems mostly focus on the monitoring of a single indicator, but in the face of complex and changeable natural and geological environments, these technologies cannot deeply explore the correlation between data, and often cannot comprehensively evaluate and warn of slope geological disaster risks, which will reduce the accuracy and reliability of slope geological disaster early warning.

[0073] Embodiment 1

[0074] The specific implementation process in the embodiment of the present application will be realized by a slope geological disaster early warning method based on a convolutional neural network of the present invention, see Figure 1 The process of the method proposed in the present invention is described in the following; a method for early warning of slope geological disasters based on a convolutional neural network comprises:

[0075] S10. Collect slope image data, environmental image data, meteorological data and seismic data;

[0076] S20. Inputting the slope image data and the environmental image data into a slope environmental feature recognition model to obtain slope deformation characteristics, slope soil characteristics, groundwater level characteristics and environmental distribution characteristics;

[0077] S30. Comprehensively evaluate the environmental distribution characteristics, the slope deformation characteristics, the slope soil characteristics and the groundwater level characteristics, and calculate the first warning value and the second warning value respectively;

[0078] S40. Perform weighted calculation on the first warning value and the second warning value to obtain a comprehensive warning value;

[0079] S50. Compare the comprehensive warning value with the comprehensive warning threshold. If it exceeds the threshold, send warning information to the staff; otherwise, input the real-time monitoring data into the slope state prediction model, and generate slope state warning information using the data results of the slope state prediction model.

[0080] Furthermore, the specific implementation process of a slope geological disaster early warning method based on convolutional neural network is as follows:

[0081] Acquire slope image data, environmental image data, meteorological data and seismic data;

[0082] Construct a slope environmental feature recognition model and slope state prediction model based on convolutional neural network;

[0083] Inputting the slope image data and the environmental image data into the slope environmental feature recognition model for recognition, and obtaining slope deformation features, slope soil features, groundwater level features, environmental soil distribution features, environmental vegetation distribution features, and environmental building distribution features;

[0084] Evaluate the environmental soil distribution characteristics, the environmental vegetation distribution characteristics, and the environmental building distribution characteristics to obtain a first warning value;

[0085] Evaluate the meteorological data, the seismic data, the slope deformation characteristics, the slope soil characteristics and the groundwater level characteristics to obtain a second warning value;

[0086] The first warning value and the second warning value are weighted with the first warning weight and the second warning weight respectively to obtain a comprehensive warning value;

[0087] The comprehensive warning value is compared with the comprehensive warning threshold. If it exceeds the comprehensive warning threshold, a warning message is sent to the staff; otherwise, the real-time monitoring data is input into the slope state prediction model, and the slope state warning message is generated using the data results of the slope state prediction model.

[0088] In this embodiment, a method for early warning of slope geological disasters based on a convolutional neural network is proposed; first, image data, meteorological data and seismic data are obtained; secondly, the image data is input into a slope environmental feature recognition model for identification and evaluation to obtain slope characteristics, groundwater level characteristics and environmental distribution characteristics; then, the slope characteristics, the groundwater level characteristics and the environmental distribution characteristics are evaluated to calculate a first warning value and a second warning value; the first warning value and the second warning value are weighted to obtain a comprehensive warning value; finally, the comprehensive warning value is compared with a comprehensive warning threshold, and if the threshold is exceeded, a warning prompt message is output, otherwise the real-time monitoring data is input into the slope state prediction model, and the slope state warning information is generated using the data results of the slope state prediction model; this method can effectively improve the accuracy and reliability of early warning of slope geological disasters.

[0089] For the purpose of specific explanation, the following embodiments are described as follows:

[0090] Acquire slope image data, environmental image data, meteorological data and seismic data;

[0091] In this embodiment, the slope image data and environmental image data are collected by using a drone equipped with a camera and remote sensing equipment, which is conducive to obtaining specific information on the groundwater level; meteorological data and earthquake data can be called from relevant agency centers.

[0092] Furthermore, a slope environmental feature recognition model and a slope state prediction model based on convolutional neural network are constructed;

[0093] Furthermore, the slope image data and the environment image data are input into the slope environment feature recognition model for recognition, and the specific implementation process of obtaining slope deformation features, slope soil features, groundwater level features, environmental soil distribution features, environmental vegetation distribution features and environmental building distribution features includes:

[0094] Acquire slope image data and environment image data;

[0095] Further, the slope image data and the environment image data are subjected to denoising processing to obtain denoised image data;

[0096] Further, the contrast and brightness of the denoised image data are adjusted using histogram equalization to obtain enhanced image data;

[0097] Furthermore, the enhanced image data is input into a slope environmental feature recognition model for processing to obtain slope features, groundwater level features and environmental distribution features; wherein the slope features include: slope deformation features and slope soil features; the environmental distribution features include: environmental soil distribution features, environmental vegetation distribution features and environmental building distribution features;

[0098] The slope environmental feature recognition model is a network model based on a convolutional neural network, and its structure can be referred to Figure 2 , mainly including: input layer, slope environment feature extraction layer, slope environment feature marking layer, slope environment feature enhancement layer, slope environment feature recognition layer and output layer; the specific implementation process of the slope environment feature recognition model includes:

[0099] Inputting the slope environment image data into the input layer to obtain a slope environment feature map;

[0100] Further, the slope environment feature map is input into the slope environment feature extraction layer to obtain the slope environment convolution feature;

[0101] Further, the slope environment convolution feature is input into the slope environment feature labeling layer to obtain an anchor frame of the slope environment convolution feature; wherein the anchor frame is set to five different sizes;

[0102] Further, extracting the features in the anchor frame and inputting them into the slope environment feature enhancement layer to obtain enhanced slope environment convolution features; wherein the slope environment feature enhancement layer realizes feature enhancement by fusing the output features of the VGG model with the features in the anchor frame;

[0103] Further, the enhanced slope environment convolution feature is input into the slope environment feature recognition layer to obtain a slope environment feature recognition result;

[0104] Furthermore, the slope environmental feature recognition result is input into the output layer to obtain slope features, groundwater level features and environmental distribution features.

[0105] In this embodiment, the preprocessing process of slope image data and environmental image data includes: Gaussian denoising, histogram equalization and edge detection; the Gaussian denoising is used to reduce interference noise generated from the outside to improve the recognition accuracy of the feature recognition model; the histogram equalization is used to enhance the contrast and brightness of the image to improve the detection capability of edge image information; this process can effectively improve the accuracy and reliability of online production monitoring of cross-linked polyethylene.

[0106] Furthermore, the specific implementation process of evaluating the environmental soil distribution characteristics, the environmental vegetation distribution characteristics and the environmental building distribution characteristics to obtain the first warning value includes:

[0107] Obtain environmental soil distribution characteristics, environmental vegetation distribution characteristics and environmental building distribution characteristics output by the slope environmental characteristic recognition model;

[0108] Further, the environmental soil distribution area, the environmental vegetation distribution area and the environmental building distribution area are obtained from the environmental soil distribution characteristics, the environmental vegetation distribution characteristics and the environmental building distribution characteristics;

[0109] Further, the environmental soil distribution area and the environmental vegetation distribution area are respectively subjected to difference calculation with the initial environmental soil distribution area and the initial environmental vegetation distribution area to obtain an environmental soil distribution difference and an environmental vegetation distribution difference;

[0110] Furthermore, a weighted evaluation is performed on the environmental soil distribution difference, the environmental vegetation distribution difference and the environmental building distribution area to obtain a first warning value; the calculation formula of the first warning value is:

[0111]

[0112] α1+α2+α3=1.0;

[0113] Wherein, DYYJ represents the first warning value. The larger the value, the greater the impact of environmental changes caused by human activities on slope geological hazards. M represents the number of environmental data samples. α1 represents the environmental soil distribution coefficient. Δtr i It is expressed as the difference of the environmental soil distribution of the i-th sample; tr th It is represented by the environmental soil distribution threshold; α2 is represented by the environmental vegetation distribution coefficient; Δzb i It is represented as the difference of the environmental vegetation distribution of the i-th sample; zb th It is represented by the environmental vegetation distribution threshold; α3 is represented by the environmental building distribution coefficient; A jz It is expressed as the distribution area of ​​the environmental buildings; A hj Expressed as the area of ​​the monitored environment.

[0114] In this embodiment, the environmental soil distribution threshold, the environmental vegetation distribution threshold and the area of ​​the monitored environment vary with geographical location and ecology, so the set values ​​need to be set by relevant personnel according to actual conditions.

[0115] The distribution coefficient in this embodiment represents the weight of the influence of different environmental characteristics on the slope geological disaster. The larger the coefficient, the higher the influence. In this embodiment, the environmental soil distribution coefficient, the environmental vegetation distribution coefficient and the environmental building distribution coefficient are set to 0.35, 0.35 and 0.3 respectively. Of course, the distribution coefficient can also be adjusted according to the actual situation.

[0116] In this embodiment, a man-made environment early warning function is proposed for early warning of man-made environmental changes in slope geological disasters; this function identifies and processes image data through a slope environmental feature recognition model to obtain environmental soil distribution features, environmental vegetation distribution features and environmental building distribution features; this function evaluates and calculates different environmental distribution data to obtain a first early warning value; the first early warning value reflects the degree of influence of the changing environment caused by human activities on the occurrence of slope geological disasters; this function can effectively improve the accuracy and reliability of slope geological disaster early warnings.

[0117] Furthermore, the specific implementation process of evaluating the meteorological data, the seismic data, the slope deformation characteristics, the slope soil characteristics and the groundwater level characteristics to obtain the second warning value includes:

[0118] Collect meteorological and seismic data;

[0119] Further, the slope deformation characteristics, slope soil characteristics and groundwater level characteristics output by the slope environmental characteristics recognition model are obtained;

[0120] Further, precipitation information, historical earthquake frequency information, slope deformation information, slope soil moisture information and groundwater level information are obtained from the meteorological data, the earthquake data, the slope deformation characteristics, the slope soil characteristics and the groundwater level characteristics respectively; wherein the historical earthquake frequency information is obtained by collecting the amplitude information of the seismic wave and performing energy calculation;

[0121] Furthermore, a weighted evaluation is performed on the precipitation information, the historical earthquake frequency information, the slope deformation information, the slope soil moisture information and the groundwater level information to calculate a second warning value; the calculation formula of the second warning value is:

[0122]

[0123] β1+β2+β3+β4+β5=1.0;

[0124] Wherein, DEYJ represents the second warning value. The larger the value, the greater the impact of environmental changes caused by natural activities on slope geological hazards. N represents the number of data samples. arctan() represents the inverse tangent function. β1 represents the precipitation impact weight. jsl i Represents the precipitation information of the i-th sample; jsl th It is represented by the standard value of precipitation; β2 is represented by the impact weight of historical earthquake frequency; nl i nl represents the historical earthquake frequency information of the i-th sample; th It is represented by the standard value of historical earthquake frequency; β3 is represented by the influence weight of slope deformation; xb i xb represents the slope deformation information of the i-th sample; th It is represented by the standard value of slope shape; β4 is represented by the influence weight of slope soil moisture; sd i Represented as the slope soil moisture information of the i-th sample; sd th It is expressed as the standard value of slope soil moisture; β5 is expressed as the influence weight of groundwater level; sw i Represents the groundwater level information of the i-th sample; sw th Expressed as the standard value of groundwater level.

[0125] The standard value of precipitation, the standard value of historical earthquake frequency, the standard value of slope deformation, the standard value of slope soil moisture and the standard value of groundwater level in this embodiment will be affected by geographical location and ecology, so the set values ​​need to be set by relevant personnel according to actual conditions.

[0126] The influence weights in this embodiment represent the influence weights of different natural factors on slope geological disasters. The larger the coefficient, the higher the degree of influence. In this embodiment, the influence weights of precipitation, historical earthquake frequency, slope shape, slope soil moisture and groundwater level are all set to 0.2. Of course, these influence weights can also be adjusted according to actual conditions.

[0127] In this embodiment, a natural environment early warning function is proposed for early warning of natural environment changes for slope geological disasters; this function identifies and processes image data through a slope environment feature recognition model to obtain slope features and groundwater level features; this function evaluates and calculates meteorological data, seismic data, the slope features and the groundwater level features to obtain a second early warning value; the second early warning value reflects the degree of influence of the changing environment caused by natural activities on the occurrence of slope geological disasters; this function can effectively improve the accuracy and reliability of slope geological disaster early warnings.

[0128] Furthermore, the first warning value and the second warning value are weighted with the first warning weight and the second warning weight respectively to obtain a comprehensive warning value; the calculation formula of the comprehensive warning value is:

[0129] ZHYJ=ω*DYYJ+(1-ω)*DEYJ;

[0130] Among them, ZHYJ represents the comprehensive warning value; ω represents the set weight value, which is 0.5; DYYJ represents the first warning value; DEYJ represents the second warning value.

[0131] Furthermore, the specific implementation process of inputting the real-time monitoring data into the slope state prediction model and using the data results of the slope state prediction model to generate slope state warning information includes:

[0132] Obtain historical monitoring data from the historical database;

[0133] Furthermore, real-time monitoring data is obtained; wherein the real-time monitoring data includes: real-time slope monitoring data and real-time environmental monitoring data;

[0134] Furthermore, the historical monitoring data is input into the slope state prediction model for training to obtain the state impact prediction parameters; the structure of the slope state prediction model is as follows: Figure 3 As shown, the main implementation process includes:

[0135] Inputting historical monitoring data into the convolutional layer to obtain a monitoring feature map;

[0136] Further, the monitoring feature map is input into a convolutional coding layer to obtain a monitoring convolutional coding feature;

[0137] Further, the monitoring convolutional coding features are input into a convolutional feature enhancement layer to obtain enhanced monitoring convolutional coding features; wherein the convolutional feature enhancement layer includes: a channel attention layer and a spatial attention layer;

[0138] Further, the enhanced monitoring convolutional coding feature is input into a convolutional decoding layer to obtain a monitoring convolutional decoding feature;

[0139] Furthermore, the monitoring convolution decoding features are input into the deconvolution layer to obtain the state impact prediction parameters.

[0140] Further, the state influence prediction parameter, the real-time slope monitoring data and the real-time environment monitoring data are input into the slope state prediction model, the model parameters are updated, and the slope state influence parameter is output; wherein the slope state influence parameter includes: a first slope influence parameter and a second environment influence parameter;

[0141] Furthermore, the real-time monitoring information and the slope state influencing parameters are comprehensively evaluated to obtain slope state warning information; the calculation formula of the slope state warning information is:

[0142] ZTYJ=η bp *INF bp +η hj *INF hj ;

[0143] Wherein, ZTYJ represents the slope status warning information; η bp It is represented as the first slope influence parameter; INF bp Represents real-time slope monitoring information; η hj Expressed as the second environmental impact parameter; INF hj Represents real-time environmental monitoring information.

[0144] In this embodiment, a slope state warning function is proposed for warning the real-time state of the slope; this function uses historical monitoring data to train the slope state prediction model to obtain state impact prediction parameters; the model parameters are updated using the state impact prediction parameters and real-time monitoring data to obtain slope state impact parameters; this function comprehensively evaluates the real-time monitoring information and the slope state impact parameters to obtain slope state warning information; the slope state warning information reflects the degree of influence of the real-time environment on the slope state; this function can effectively improve the accuracy and reliability of slope geological disaster warnings.

[0145] In order to verify the actual effect of the early warning method for slope geological disasters based on convolutional neural network proposed in the present invention, several groups of comparative experiments were designed; among them, method one applied the slope environmental feature recognition model and slope state prediction model based on convolutional neural network proposed in the present invention; method two only used the slope state prediction model to warn of geological disasters, omitting the slope environmental feature recognition model; method three used traditional manual recording parameters to warn of slope geological disasters;

[0146] As shown in Table 1, the slope geological disaster early warning method (method one) combining the slope environmental feature recognition model and the slope state prediction model performed best, indicating that the method proposed in the present invention is the most effective in slope geological disaster early warning.

[0147] Table 1. Comparison of the effects of different early warning schemes

[0148]

[0149] Embodiment 2

[0150] As an embodiment of the present invention, refer to Figure 4, an early warning system for slope geological disasters based on convolutional neural network, comprising: a system control module, a data acquisition module, a data processing module and an early warning output module;

[0151] Wherein, the system control module is used to control the start, pause and stop of the device;

[0152] The data acquisition module is used to obtain slope image data, environmental image data, meteorological data, seismic data and historical monitoring data in the historical database;

[0153] The data processing module is used to process the slope image data, the environmental image data, the meteorological data and the seismic data to calculate a comprehensive warning value; wherein the data processing module includes: a first processing unit and a second processing unit;

[0154] Furthermore, the data processing module inputs the slope image data and the environmental image data into the slope environmental feature recognition model for recognition, and obtains the slope deformation feature, slope soil feature, groundwater level feature, environmental soil cover feature, environmental vegetation cover feature and environmental building cover feature. The specific implementation process includes:

[0155] Acquire slope image data and environment image data;

[0156] Further, the slope image data and the environment image data are subjected to denoising processing to obtain denoised image data;

[0157] Further, the contrast and brightness of the denoised image data are adjusted using histogram equalization to obtain enhanced image data;

[0158] Furthermore, the enhanced image data is input into a slope environmental feature recognition model for processing to obtain slope deformation features, slope soil features, groundwater level features, environmental soil cover features, environmental vegetation cover features and environmental building cover features.

[0159] Furthermore, the first processing unit evaluates the environmental soil distribution characteristics, the environmental vegetation distribution characteristics and the environmental building distribution characteristics to obtain the first warning value, and the specific implementation process includes:

[0160] Obtain environmental soil distribution characteristics, environmental vegetation distribution characteristics and environmental building distribution characteristics output by the slope environmental characteristic recognition model;

[0161] Further, the environmental soil distribution area, the environmental vegetation distribution area and the environmental building distribution area are obtained from the environmental soil distribution characteristics, the environmental vegetation distribution characteristics and the environmental building distribution characteristics;

[0162] Further, the environmental soil distribution area and the environmental vegetation distribution area are respectively subjected to difference calculation with the initial environmental soil distribution area and the initial environmental vegetation distribution area to obtain an environmental soil distribution difference and an environmental vegetation distribution difference;

[0163] Furthermore, a weighted evaluation is performed on the environmental soil distribution difference, the environmental vegetation distribution difference and the environmental building distribution area to obtain a first warning value; the calculation formula of the first warning value is:

[0164]

[0165] α1+α2+α3=1.0;

[0166] Wherein, DYYJ represents the first warning value. The larger the value, the greater the impact of environmental changes caused by human activities on slope geological hazards. M represents the number of environmental data samples. α1 represents the environmental soil distribution coefficient. Δtr i It is expressed as the difference of the environmental soil distribution of the i-th sample; tr th It is represented by the environmental soil distribution threshold; α2 is represented by the environmental vegetation distribution coefficient; Δzb i It is represented as the difference of the environmental vegetation distribution of the i-th sample; zb th It is represented by the environmental vegetation distribution threshold; α3 is represented by the environmental building distribution coefficient; A jz It is expressed as the distribution area of ​​the environmental buildings; A hj Expressed as the area of ​​the monitored environment.

[0167] Furthermore, the second processing unit evaluates the meteorological data, the seismic data, the slope deformation characteristics, the slope soil characteristics and the groundwater level characteristics to obtain the second warning value, and the specific implementation process includes:

[0168] Collect meteorological and seismic data;

[0169] Further, the slope deformation characteristics, slope soil characteristics and groundwater level characteristics output by the slope environmental characteristics recognition model are obtained;

[0170] Further, precipitation information, historical earthquake frequency information, slope deformation information, slope soil moisture information and groundwater level information are obtained from the meteorological data, the earthquake data, the slope deformation characteristics, the slope soil characteristics and the groundwater level characteristics respectively;

[0171] Furthermore, a weighted evaluation is performed on the precipitation information, the historical earthquake frequency information, the slope deformation information, the slope soil moisture information and the groundwater level information to calculate a second warning value; the calculation formula of the second warning value is:

[0172]

[0173] β1+β2+β3+β4+β5=1.0;

[0174] Wherein, DEYJ represents the second warning value. The larger the value, the greater the impact of environmental changes caused by natural activities on slope geological hazards. N represents the number of data samples. arctan() represents the inverse tangent function. β1 represents the precipitation impact weight. jsl i Represents the precipitation information of the i-th sample; jsl th It is represented by the standard value of precipitation; β2 is represented by the impact weight of historical earthquake frequency; nl i nl represents the historical earthquake frequency information of the i-th sample; th It is represented by the standard value of historical earthquake frequency; β3 is represented by the influence weight of slope deformation; xb i xb represents the slope deformation information of the i-th sample; th It is represented by the standard value of slope shape; β4 is represented by the influence weight of slope soil moisture; sd i Represented as the slope soil moisture information of the i-th sample; sd th It is expressed as the standard value of slope soil moisture; β5 is expressed as the influence weight of groundwater level; sw i Represents the groundwater level information of the i-th sample; sw th Expressed as the standard value of groundwater level.

[0175] Furthermore, the first warning value and the second warning value are weighted with the first warning weight and the second warning weight respectively to obtain a comprehensive warning value; the calculation formula of the comprehensive warning value is:

[0176] ZHYJ=ω*DYYJ+(1-ω)*DEYJ;

[0177] Among them, ZHYJ represents the comprehensive warning value; ω represents the set weight value, which is 0.5; DYYJ represents the first warning value; DEYJ represents the second warning value.

[0178] The warning output module is used to evaluate the real-time monitoring data, output the warning results and provide warning feedback; wherein, the warning output module includes: a warning judgment unit, a warning prompt unit and a slope status warning unit.

[0179] The warning judgment unit is used to judge the numerical relationship between the comprehensive warning value and the comprehensive warning threshold. If the value exceeds the threshold, a signal is sent to the warning prompt unit to output the disaster warning prompt information; otherwise, the slope state warning unit continues to process; the warning prompt unit is used to output the slope geological disaster warning information in the form of voice, video, etc.

[0180] Furthermore, the slope state warning unit inputs the real-time monitoring data into the slope state prediction model, and the specific implementation process of generating slope state warning information using the data results of the slope state prediction model includes:

[0181] Obtain historical monitoring data from the historical database;

[0182] Furthermore, real-time monitoring data is obtained; wherein the real-time monitoring data includes: real-time slope monitoring data and real-time environmental monitoring data;

[0183] Further, the historical monitoring data is input into a slope state prediction model for training to obtain state impact prediction parameters;

[0184] Further, the state influence prediction parameter, the real-time slope monitoring data and the real-time environment monitoring data are input into the slope state prediction model, the model parameters are updated, and the slope state influence parameter is output; wherein the slope state influence parameter includes: a first slope influence parameter and a second environment influence parameter;

[0185] Furthermore, the real-time monitoring information and the slope state influencing parameters are comprehensively evaluated to obtain slope state warning information; the calculation formula of the slope state warning information is:

[0186] ZTYJ=η bp *INF bp +η hj *INF hj ;

[0187] Wherein, ZTYJ represents the slope status warning information; η bp It is represented as the first slope influence parameter; INF bp Represents real-time slope monitoring information; η hj Expressed as the second environmental impact parameter; INF hj Represents real-time environmental monitoring information.

[0188] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for early warning of slope geological disasters based on convolutional neural network, characterized in that: include: Acquire slope image data, environmental image data, meteorological data and seismic data; Construct a slope environmental feature recognition model and slope state prediction model based on convolutional neural network; Inputting the slope image data and the environmental image data into the slope environmental feature recognition model for recognition, and obtaining slope deformation features, slope soil features, groundwater level features, environmental soil distribution features, environmental vegetation distribution features, and environmental building distribution features; Evaluate the environmental soil distribution characteristics, the environmental vegetation distribution characteristics, and the environmental building distribution characteristics to obtain a first warning value; Evaluate the meteorological data, the seismic data, the slope deformation characteristics, the slope soil characteristics and the groundwater level characteristics to obtain a second warning value; The first warning value and the second warning value are weighted with the first warning weight and the second warning weight respectively to obtain a comprehensive warning value; Comparing the comprehensive warning value with the comprehensive warning threshold, and sending a warning message to the staff if the comprehensive warning threshold is exceeded; Otherwise, the real-time monitoring data is input into the slope state prediction model, and the slope state warning information is generated using the data results of the slope state prediction model.

2. The early warning method for slope geological disasters based on convolutional neural network according to claim 1 is characterized in that: The specific implementation process of inputting the slope image data and the environmental image data into the slope environmental feature recognition model for recognition to obtain slope deformation features, slope soil features, groundwater level features, environmental soil distribution features, environmental vegetation distribution features and environmental building distribution features includes: Acquire slope image data and environment image data; Performing denoising processing on the slope image data and the environment image data to obtain denoised image data; Using histogram equalization to adjust the contrast and brightness of the denoised image data to obtain enhanced image data; The enhanced image data is input into a slope environmental feature recognition model for processing to obtain slope deformation features, slope soil features, groundwater level features, environmental soil distribution features, environmental vegetation distribution features and environmental building distribution features.

3. The early warning method for slope geological disasters based on convolutional neural network according to claim 1 is characterized in that: The specific implementation process of evaluating the environmental soil distribution characteristics, the environmental vegetation distribution characteristics, and the environmental building distribution characteristics to obtain the first warning value includes: Obtain environmental soil distribution characteristics, environmental vegetation distribution characteristics and environmental building distribution characteristics output by the slope environmental characteristic recognition model; Obtaining an environmental soil distribution area, an environmental vegetation distribution area and an environmental building distribution area from the environmental soil distribution characteristics, the environmental vegetation distribution characteristics and the environmental building distribution characteristics; Performing difference calculation on the environmental soil distribution area and the environmental vegetation distribution area with the initial environmental soil distribution area and the initial environmental vegetation distribution area, respectively, to obtain an environmental soil distribution difference and an environmental vegetation distribution difference; A weighted evaluation is performed on the environmental soil distribution difference, the environmental vegetation distribution difference and the environmental building distribution area to obtain a first warning value.

4. The early warning method for slope geological disasters based on convolutional neural network according to claim 1 is characterized in that: The specific implementation process of evaluating the meteorological data, the seismic data, the slope deformation characteristics, the slope soil characteristics and the groundwater level characteristics to obtain the second warning value includes: Collect meteorological and seismic data; Obtain slope deformation characteristics, slope soil characteristics and groundwater level characteristics output by the slope environmental characteristics recognition model; Obtaining precipitation information, historical earthquake frequency information, slope deformation information, slope soil moisture information and groundwater level information from the meteorological data, the earthquake data, the slope deformation characteristics, the slope soil characteristics and the groundwater level characteristics respectively; A weighted evaluation is performed on the precipitation information, the historical earthquake frequency information, the slope deformation information, the slope soil moisture information and the groundwater level information to calculate a second warning value.

5. The early warning method for slope geological disasters based on convolutional neural network according to claim 1 is characterized in that: The specific implementation process of inputting the real-time monitoring data into the slope state prediction model and generating the slope state warning information using the data results of the slope state prediction model includes: Obtain historical monitoring data from the historical database; Acquire real-time monitoring data; wherein the real-time monitoring data includes: real-time slope monitoring data and real-time environmental monitoring data; Inputting the historical monitoring data into a slope state prediction model for training to obtain state impact prediction parameters; Input the state influence prediction parameter, the real-time slope monitoring data and the real-time environment monitoring data into the slope state prediction model, update the model parameters, and output the slope state influence parameter; wherein the slope state influence parameter includes: a first slope influence parameter and a second environment influence parameter; The real-time monitoring information and the slope state influencing parameters are comprehensively evaluated to obtain slope state warning information.

6. A convolutional neural network-based early warning system for slope geological disasters, characterized in that: include: System control module, data acquisition module, data processing module and early warning output module; Wherein, the system control module is used to control the start, pause and stop of the device; The data acquisition module is used to obtain slope image data, environmental image data, meteorological data, seismic data and historical monitoring data in the historical database; The data processing module is used to process the slope image data, the environmental image data, the meteorological data and the seismic data to calculate a comprehensive warning value; wherein the data processing module includes: a first processing unit and a second processing unit; The warning output module is used to evaluate the real-time monitoring data, output the warning results and provide warning feedback; wherein, the warning output module includes: a warning judgment unit, a warning prompt unit and a slope status warning unit.

7. The early warning system for slope geological disasters based on convolutional neural network according to claim 6 is characterized in that: The data processing module inputs the slope image data and the environmental image data into the slope environmental feature recognition model for recognition, and obtains slope deformation features, slope soil features, groundwater level features, environmental soil distribution features, environmental vegetation distribution features, and environmental building distribution features. The specific implementation process includes: Acquire slope image data and environment image data; Performing denoising processing on the slope image data and the environment image data to obtain denoised image data; Using histogram equalization to adjust the contrast and brightness of the denoised image data to obtain enhanced image data; The enhanced image data is input into a slope environmental feature recognition model for processing to obtain slope deformation features, slope soil features, groundwater level features, environmental soil distribution features, environmental vegetation distribution features and environmental building distribution features.

8. The early warning system for slope geological disasters based on convolutional neural network according to claim 6 is characterized in that: The data processing module evaluates the environmental soil distribution characteristics, environmental vegetation distribution characteristics and environmental building distribution characteristics to obtain the first warning value. The specific implementation process includes: Obtain environmental soil distribution characteristics, environmental vegetation distribution characteristics and environmental building distribution characteristics output by the slope environmental characteristic recognition model; Obtaining an environmental soil distribution area, an environmental vegetation distribution area and an environmental building distribution area from the environmental soil distribution characteristics, the environmental vegetation distribution characteristics and the environmental building distribution characteristics; Performing difference calculation on the environmental soil distribution area and the environmental vegetation distribution area with the initial environmental soil distribution area and the initial environmental vegetation distribution area, respectively, to obtain an environmental soil distribution difference and an environmental vegetation distribution difference; A weighted evaluation is performed on the environmental soil distribution difference, the environmental vegetation distribution difference and the environmental building distribution area to obtain a first warning value.

9. The early warning system for slope geological disasters based on convolutional neural network according to claim 6, characterized in that: The specific implementation process of the data processing module evaluating the meteorological data, the seismic data, the slope deformation characteristics, the slope soil characteristics and the groundwater level characteristics to obtain the second warning value includes: Collect meteorological and seismic data; Obtain slope deformation characteristics, slope soil characteristics and groundwater level characteristics output by the slope environmental characteristics recognition model; Obtaining precipitation information, historical earthquake frequency information, slope deformation information, slope soil moisture information and groundwater level information from the meteorological data, the earthquake data, the slope deformation characteristics, the slope soil characteristics and the groundwater level characteristics respectively; A weighted evaluation is performed on the precipitation information, the historical earthquake frequency information, the slope deformation information, the slope soil moisture information and the groundwater level information to calculate a second warning value.

10. The early warning system for slope geological disasters based on convolutional neural network according to claim 6, characterized in that: The slope state warning unit inputs the real-time monitoring data into the slope state prediction model, and the specific implementation process of using the data results of the slope state prediction model to generate slope state warning information includes: Obtain historical monitoring data from the historical database; Acquire real-time monitoring data; wherein the real-time monitoring data includes: real-time slope monitoring data and real-time environmental monitoring data; Inputting the historical monitoring data into a slope state prediction model for training to obtain state impact prediction parameters; Input the state influence prediction parameter, the real-time slope monitoring data and the real-time environment monitoring data into the slope state prediction model, update the model parameters, and output the slope state influence parameter; wherein the slope state influence parameter includes: a first slope influence parameter and a second environment influence parameter; The real-time monitoring information and the slope state influencing parameters are comprehensively evaluated to obtain slope state warning information.

Images