Mountain area earthquake hidden danger prediction method and system based on unmanned aerial vehicle and deep learning

Abstract

The invention relates to the technical field of disaster prevention and reduction, and discloses a mountain area earthquake hidden danger prediction method and system based on an unmanned aerial vehicle and deep learning, and the method comprises the steps: carrying out the image collection and data recording of a to-be-predicted mountain area through employing the unmanned aerial vehicle, obtaining the image data of the to-be-predicted mountain area, and constructing a three-dimensional to-be-predicted terrain model according to the image data; extracting terrain information of a to-be-predicted mountainous area from the three-dimensional to-be-predicted terrain model, collecting geological information of the to-be-predicted mountainous area, and performing fundamental frequency prediction on the to-be-predicted mountainous area to obtain mountainous area fundamental frequency information; and performing seismic oscillation amplification effect prediction on the to-be-predicted mountain area according to the mountain area fundamental frequency information, the geological information and the topographic information, obtaining a seismic oscillation amplification coefficient envelope value of each feature point in the to-be-predicted mountain area under the action of seismic waves, performing earthquake hidden danger grade classification, and drawing an earthquake hidden danger distribution diagram of the to-be-predicted mountain area. According to the invention, the problem that the pre-earthquake mountain earthquake disaster hidden danger is difficult to find is solved, and more effective reference is provided for prevention and treatment of mountain earthquake disasters.

Description

Technical Field

[0001] The present application relates to the field of disaster prevention and mitigation technology, and in particular to a method and system for predicting earthquake hazards in mountainous areas based on drones and deep learning. Background Art

[0002] According to existing research, complex terrain in mountainous areas can cause seismic waves to focus, reflect and scatter, resulting in a ground motion amplification effect, which intensifies the intensity of ground motion in local areas, causing rocks inside geological bodies or on the surface of mountains to break and collapse, causing originally stable slopes to lose balance, and ultimately inducing landslides and slope collapse. It can be seen that if areas with more obvious ground motion amplification effects can be detected in advance and relevant prevention and control measures are taken, the secondary disasters caused by earthquakes can be greatly reduced.

[0003] Deep learning technology can process and analyze various complex patterns and data by establishing neural network models, and has shown significant advantages in earthquake early warning. Deep learning can be applied to the processing and analysis of seismic wave data, earthquake damage prediction, and earthquake disaster impact assessment. On the other hand, drone technology plays an important role in real terrain modeling due to its efficiency, safety, and accuracy. Equipped with high-resolution cameras, infrared thermal imaging, and ground-based lidar technologies, drones can quickly enter mountainous areas, capture a wide range of terrain images, detect geological changes, and achieve rapid modeling of real terrain.

[0004] Combining the high-altitude aerial photography capabilities of drones with the data analysis and prediction capabilities of deep learning can form a new method for predicting potential earthquake disasters in mountainous areas. The images and data of disaster areas obtained by drones can be used as inputs to deep learning models. Through the analysis and prediction of the model, the extent and scope of the disaster can be more accurately determined, providing support for post-disaster emergency response.

[0005] In summary, the earthquake disaster hazard prediction system in mountainous areas based on drones and deep learning can obtain image data in real time and quickly analyze and predict through deep learning models. The system has the advantages of strong real-time performance, flexibility, convenience, little impact from the external environment, and low cost, providing reliable technical support for geological disaster prediction, disaster emergency response, and risk zoning. Summary of the invention

[0006] To achieve the above objectives, this application provides the following technical solutions: According to the first aspect of the present invention, the present invention claims a method for predicting earthquake hazards in mountainous areas based on drones and deep learning, comprising: Using a drone to collect images and record data of the mountainous area to be predicted, obtaining image data of the mountainous area to be predicted, and constructing a three-dimensional terrain model to be measured based on the image data; Extracting terrain information of the mountainous area to be predicted from the three-dimensional terrain model to be measured, collecting geological information of the mountainous area to be predicted, and performing fundamental frequency prediction on the mountainous area to be predicted to obtain fundamental frequency information of the mountainous area; Predicting the seismic amplification effect of the mountainous area to be predicted based on the base frequency information, geological information and topographic information of the mountainous area, and obtaining the envelope value of the seismic amplification coefficient of each characteristic point in the mountainous area to be predicted under the action of seismic waves; The earthquake hazard level is classified by the earthquake amplification coefficient envelope value, and a distribution map of earthquake disaster hazards in the mountainous area to be predicted is drawn according to the earthquake hazard level classification.

[0007] Furthermore, the method of using a drone to collect images and record data of the mountainous area to be predicted, obtaining image data of the mountainous area to be predicted, and constructing a three-dimensional terrain model to be measured based on the image data, further includes: Acquire the characteristics of the mountainous terrain and earthquake hazards to be predicted, and plan the flight route and flight range of the UAV; Through the high-resolution camera sensor carried by the drone, all-round image acquisition and data recording of the mountainous area to be predicted are performed to ensure that image data of a preset quantity and quality are obtained; The image data taken by the drone is imported into the corresponding three-dimensional terrain construction software, submitted to the aerial triangulation calculation, and a three-dimensional terrain model to be measured is constructed.

[0008] Furthermore, the terrain information of the mountainous area to be predicted includes at least: Terrain type, height, slope, local features.

[0009] Furthermore, the method of predicting the seismic amplification effect of the mountainous area to be predicted based on the base frequency information, geological information and terrain information of the mountainous area, and obtaining the envelope value of the seismic amplification coefficient of each characteristic point in the mountainous area to be predicted under the action of seismic waves, further includes: Input the acquired base frequency information, terrain information and geological information of the mountain area into the seismic amplification effect prediction module to realize the prediction of the amplification effect of simple harmonic waves of different frequencies, and output the first seismic amplification coefficient of each characteristic point in the mountain area to be predicted; The first seismic amplification coefficients of simple harmonic waves of different frequencies are input into the data integration module. The second seismic amplification coefficients under different seismic waves are obtained through parameter combination. Based on the second seismic amplification coefficients of each characteristic point under each seismic wave, the corresponding seismic amplification coefficient envelope value is calculated.

[0010] Furthermore, the earthquake hazard level is classified by the earthquake amplification coefficient envelope value, and a distribution map of earthquake disaster hazards in the mountainous area to be predicted is drawn according to the earthquake hazard level classification, which also includes: Input the obtained envelope value of the seismic motion amplification coefficient of each test point into the regional earthquake disaster hazard level classification module, and output the earthquake hazard level of the corresponding test point by judging the seismic motion amplification coefficient of the test point point by point; According to the terrain information, the seismic amplification coefficients and disaster risks of different test points are graded using a post-processing imaging module, and plotted point by point on the three-dimensional test terrain model according to the data distribution points to obtain a regional earthquake disaster risk distribution map.

[0011] Furthermore, when constructing a three-dimensional terrain model to be measured based on the image data, it also includes: Dividing the image data into aerial images and ground point cloud data; Performing image preprocessing on the aerial images and ground point cloud data; The image preprocessing at least includes: Dedistortion processing, point cloud registration, 3D reconstruction, texture mapping, and data fusion.

[0012] Furthermore, when performing fundamental frequency prediction on the mountainous area to be predicted to obtain fundamental frequency information of the mountainous area, the method further includes: Collecting the measured earthquake acceleration time history curve, determining the geological feature label of the mountainous area to be predicted, and constructing a site fundamental frequency database; Determine an initial network, select fundamental frequency sample data from the site fundamental frequency database, train and optimize the initial network based on the fundamental frequency sample data, and obtain a site fundamental frequency prediction optimization neural network model; The required data of the mountainous area to be predicted is input into the site fundamental frequency prediction optimization neural network, and the site dominant frequencies of different sites are output.

[0013] Furthermore, the obtained base frequency information, terrain information and geological information of the mountain area are input into the seismic motion amplification effect prediction module to realize the prediction of the amplification effect of simple harmonic waves of different frequencies, and the first seismic motion amplification coefficient of each characteristic point in the mountain area to be predicted is output, which also includes: Collecting measured earthquake acceleration time history curves, determining geological feature labels of the mountainous area to be predicted, and constructing a seismic motion amplification effect database; Determine an initial network, select fundamental frequency sample data from the earthquake amplification effect database, train and optimize the initial network according to the fundamental frequency sample data, and obtain an earthquake amplification effect optimized neural network model; The required data of the mountainous area to be predicted is input into the seismic amplification effect optimization neural network model, and the seismic amplification coefficients of different sites under the action of simple harmonic waves of various frequencies are output.

[0014] Furthermore, the first seismic amplification coefficients of simple harmonic waves of different frequencies are input into the data integration module, and the second seismic amplification coefficients under different seismic waves are obtained by parameter combination, and the corresponding seismic amplification coefficient envelope value is calculated according to the second seismic amplification coefficients of each characteristic point under each seismic wave, and further includes: ; After determining the range of n, randomly generate it through the random() function, and randomly select n first seismic amplification systems from the first seismic amplification systems of simple harmonic waves of different frequencies for combination, and then repeat the work N times, N>10000, sort the calculated second seismic amplification coefficient results from small to large, and determine the maximum data of the first 95% as the final second seismic amplification coefficient; Evaluate specific seismic waves, obtain the seismic amplification coefficients of the predicted mountainous area under the action of simple harmonic waves of various frequencies, combine them, and generate the seismic amplification coefficients under the action of specific seismic waves as the quantitative coefficients of the disaster hazards under the action of specific seismic waves in the predicted mountainous area; Or, the seismic amplification coefficient is evaluated based on Monte Carlo, and the envelope value of the seismic amplification coefficient of the mountain area to be predicted under the action of multiple groups of random seismic waves is calculated as the quantitative coefficient of the final disaster hidden danger of the mountain area to be predicted.

[0015] According to the second aspect of the present invention, the present invention claims protection for a system for predicting earthquake hazards in mountainous areas based on drones and deep learning, comprising: one or more processors; A memory having one or more programs stored thereon, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method for predicting earthquake hazards in mountainous areas based on drones and deep learning.

[0016] Compared with the prior art, the beneficial technical effects of this application are: The method and system for predicting earthquake hazards in mountainous areas based on drones and deep learning of the present application adopt drones to collect images and record data of the mountainous areas to be predicted, obtain image data of the mountainous areas to be predicted, and construct a three-dimensional terrain model to be tested based on the image data; extract terrain information of the mountainous areas to be predicted from the three-dimensional terrain model to be tested, collect geological information of the mountainous areas to be predicted, perform fundamental frequency prediction on the mountainous areas to be predicted, and obtain fundamental frequency information of the mountainous areas; predict the seismic amplification effect of the mountainous areas to be predicted based on the fundamental frequency information, geological information and terrain information of the mountainous areas, obtain the envelope value of the seismic amplification coefficient of each characteristic point in the mountainous areas to be predicted under the action of seismic waves, classify the seismic hazard level, and draw a distribution map of earthquake disaster hazards in the mountainous areas to be predicted. The present invention improves the problem of difficulty in discovering earthquake disaster hazards in mountainous areas before an earthquake, and provides a more effective reference for the prevention and control of earthquake disasters in mountainous areas. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A workflow diagram of a method for predicting earthquake hazards in mountainous areas based on drones and deep learning, as claimed in an embodiment of the present invention. DETAILED DESCRIPTION

[0018] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0019] The terms "first", "second" and "third" in this application are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features. Therefore, the features defined as "first", "second" and "third" can explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "multiple" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined. All directional indications (such as up, down, left, right, front, back...) in the embodiments of this application are only used to explain the relative position relationship, movement, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication also changes accordingly. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally also includes steps or units that are not listed, or optionally also includes other steps or units inherent to these processes, methods, products or devices.

[0020] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0021] According to the first embodiment of the present invention, the present invention claims a method for predicting earthquake hazards in mountainous areas based on drones and deep learning, referring to Figure 1 ,include: Using a drone to collect images and record data of the mountainous area to be predicted, obtaining image data of the mountainous area to be predicted, and constructing a three-dimensional terrain model to be measured based on the image data; Extracting terrain information of the mountainous area to be predicted from the three-dimensional terrain model to be measured, collecting geological information of the mountainous area to be predicted, and performing fundamental frequency prediction on the mountainous area to be predicted to obtain fundamental frequency information of the mountainous area; Predicting the seismic amplification effect of the mountainous area to be predicted based on the base frequency information, geological information and topographic information of the mountainous area, and obtaining the envelope value of the seismic amplification coefficient of each characteristic point in the mountainous area to be predicted under the action of seismic waves; The earthquake hazard level is classified by the earthquake amplification coefficient envelope value, and a distribution map of earthquake disaster hazards in the mountainous area to be predicted is drawn according to the earthquake hazard level classification.

[0022] Furthermore, the method of using a drone to collect images and record data of the mountainous area to be predicted, obtaining image data of the mountainous area to be predicted, and constructing a three-dimensional terrain model to be measured based on the image data, further includes: Acquire the characteristics of the mountainous terrain and earthquake hazards to be predicted, and plan the flight route and flight range of the UAV; Through the high-resolution camera sensor carried by the drone, all-round image acquisition and data recording of the mountainous area to be predicted are performed to ensure that image data of a preset quantity and quality are obtained; The image data taken by the drone is imported into the corresponding three-dimensional terrain construction software, submitted to the aerial triangulation calculation, and a three-dimensional terrain model to be measured is constructed.

[0023] Furthermore, the terrain information of the mountainous area to be predicted includes at least: Terrain type, height, slope, local features.

[0024] Furthermore, the method of predicting the seismic amplification effect of the mountainous area to be predicted based on the base frequency information, geological information and terrain information of the mountainous area, and obtaining the envelope value of the seismic amplification coefficient of each characteristic point in the mountainous area to be predicted under the action of seismic waves, further includes: Input the acquired base frequency information, terrain information and geological information of the mountain area into the seismic amplification effect prediction module to realize the prediction of the amplification effect of simple harmonic waves of different frequencies, and output the first seismic amplification coefficient of each characteristic point in the mountain area to be predicted; The first seismic amplification coefficients of simple harmonic waves of different frequencies are input into the data integration module. The second seismic amplification coefficients under different seismic waves are obtained through parameter combination. Based on the second seismic amplification coefficients of each characteristic point under each seismic wave, the corresponding seismic amplification coefficient envelope value is calculated.

[0025] Furthermore, the earthquake hazard level is classified by the earthquake amplification coefficient envelope value, and a distribution map of earthquake disaster hazards in the mountainous area to be predicted is drawn according to the earthquake hazard level classification, which also includes: Input the obtained envelope value of the seismic motion amplification coefficient of each test point into the regional earthquake disaster hazard level classification module, and output the earthquake hazard level of the corresponding test point by judging the seismic motion amplification coefficient of the test point point by point; According to the terrain information, the seismic amplification coefficients and disaster risks of different test points are graded using a post-processing imaging module, and plotted point by point on the three-dimensional test terrain model according to the data distribution points to obtain a regional earthquake disaster risk distribution map.

[0026] Among them, in this embodiment, the seismic motion amplification coefficient envelope values ​​of 1.0, 1.5, 2.0, and 2.5 are taken as critical, and the earthquake disaster hidden danger levels are divided into: low hidden danger, general hidden danger, major hidden danger, serious hidden danger, and extremely serious hidden danger.

[0027] Furthermore, when constructing a three-dimensional terrain model to be measured based on the image data, it also includes: Dividing the image data into aerial images and ground point cloud data; Performing image preprocessing on the aerial images and ground point cloud data; The image preprocessing at least includes: Dedistortion processing, point cloud registration, 3D reconstruction, texture mapping, and data fusion.

[0028] In this embodiment, it also includes: Data collection: UAVs use the sensors and cameras they carry to obtain high-resolution aerial images and ground point cloud data. Aerial images provide rich texture and color information, while ground point cloud data provides ground geometry information.

[0029] Image processing: De-distort the collected aerial images to eliminate the effects caused by camera lens distortion. Use image matching algorithms to match images from different perspectives to find the same feature points. Through these feature points, calculate the camera's attitude parameters and the three-dimensional structure of the scene.

[0030] Point cloud processing: Filter the collected ground point cloud data to remove noise points and outliers. Use the point cloud registration algorithm to register point cloud data from different perspectives to obtain a globally consistent point cloud model. Use the point cloud segmentation algorithm to segment the point cloud data into different objects or ground.

[0031] 3D reconstruction: 3D reconstruction based on aerial images can calculate the 3D positions of feature points through multi-view geometry and triangulation methods, and generate a dense 3D point cloud model. 3D reconstruction based on ground point cloud data can generate a globally consistent ground model through point cloud registration and point cloud fusion methods.

[0032] Texture mapping: Mapping the texture of aerial images onto 3D models to enhance the realism of the models. This requires matching the aerial images with the 3D models and projecting the color and texture information in the images onto the surface of the models.

[0033] Data fusion: Aerial images and ground point cloud data are registered and fused to obtain a more accurate and complete 3D model.

[0034] Furthermore, when performing fundamental frequency prediction on the mountainous area to be predicted to obtain fundamental frequency information of the mountainous area, the method further includes: Collecting the measured earthquake acceleration time history curve, determining the geological feature label of the mountainous area to be predicted, and constructing a site fundamental frequency database; Determine an initial network, select fundamental frequency sample data from the site fundamental frequency database, train and optimize the initial network based on the fundamental frequency sample data, and obtain a site fundamental frequency prediction optimization neural network model; The required data of the mountainous area to be predicted is input into the site fundamental frequency prediction optimization neural network, and the site dominant frequencies of different sites are output.

[0035] In this embodiment, it also includes: According to the type of terrain, the measured earthquake acceleration time history curves are collected and Fourier analysis is performed to determine the site's dominant period. Based on this, the finite element model is verified and parameter expansion is performed to obtain the simulated earthquake acceleration time history curves, and Fourier analysis is performed to determine the site's dominant frequency. The terrain and geological feature labels are determined, and the site base frequency database is built with the measured data and simulated data as the first preset ratio; Construct a neural network model, determine the initial network model, randomly select sample data from the site base frequency database, and divide the training data set and the test data set according to the second preset ratio. Define the loss function as ,in, is the total number of samples in the test data set, To predict the results, For real results; Input all feature labels in the site base frequency database into the neural network, perform neural network training, and wait for the loss function to If the weight is less than the set value, the training ends. Output the final weight value of each feature label, judge the importance of each terrain geological feature label in the neural network, and define the feature label with a weight value greater than 1% as the main label; Re-establish the neural network model, use the main topographic and geological labels as the input features of the neural network, and perform training again. If it is less than the set value, the training ends and the training results are saved. The final model is the field fundamental frequency optimization neural network model. The data required for the terrain to be tested is input into the site fundamental frequency prediction optimization neural network, and the site dominant frequencies of different sites are output to achieve the prediction of the dominant frequencies of different sites.

[0036] Furthermore, the obtained base frequency information, terrain information and geological information of the mountain area are input into the seismic motion amplification effect prediction module to realize the prediction of the amplification effect of simple harmonic waves of different frequencies, and the first seismic motion amplification coefficient of each characteristic point in the mountain area to be predicted is output, which also includes: Collecting measured earthquake acceleration time history curves, determining geological feature labels of the mountainous area to be predicted, and constructing a seismic motion amplification effect database; Determine an initial network, select fundamental frequency sample data from the earthquake amplification effect database, train and optimize the initial network according to the fundamental frequency sample data, and obtain an earthquake amplification effect optimized neural network model; The required data of the mountainous area to be predicted is input into the seismic amplification effect optimization neural network model, and the seismic amplification coefficients of different sites under the action of simple harmonic waves of various frequencies are output.

[0037] In this embodiment, it also includes: According to the type of terrain, the measured earthquake acceleration time history curves are collected and Fourier analysis is performed to determine the seismic amplification coefficients of different site terrains under the action of simple harmonic waves of different frequencies. Based on this, the finite element model is verified, the parameters are expanded, the simulated earthquake acceleration time history curves are obtained, and Fourier analysis is performed to determine the seismic amplification coefficients of different site terrains under the action of simple harmonic waves of different frequencies. The site feature labels (including the site fundamental frequency) are determined, and a seismic amplification effect database is built with the measured data and the simulated data as the first preset ratio; Construct a neural network model, determine the initial network model, randomly select seismic data from the seismic amplification effect database, and divide the training data set and the test data set according to the second preset ratio. Define the loss function as ,in, is the total number of samples in the test data set, To predict the results, For real results; Input all feature labels in the seismic amplification effect database into the neural network, perform neural network training, and wait for the loss function to If the weight is less than the set value, the training ends. Output the final weight value of each feature label, judge the importance of each terrain geological feature label in the neural network, and define the feature label with a weight value greater than 1% as the main label; Re-establish the neural network model, use the main topographic and geological labels as the input features of the neural network, and perform training again. If it is less than the set value, the training ends and the training results are saved. The final model is the seismic amplification effect optimization neural network model. According to the measured seismic time history curve and the site fundamental frequency obtained in step (4.5), the data required for the terrain to be measured is input into the seismic amplification effect prediction optimization neural network, and the seismic amplification coefficients of different sites under the action of simple harmonic waves of various frequencies are output.

[0038] Furthermore, the first seismic amplification coefficients of simple harmonic waves of different frequencies are input into the data integration module, and the second seismic amplification coefficients under different seismic waves are obtained by parameter combination, and the corresponding seismic amplification coefficient envelope value is calculated according to the second seismic amplification coefficients of each characteristic point under each seismic wave, and further includes: ; After determining the range of n, randomly generate it through the random() function, and randomly select n first seismic amplification systems from the first seismic amplification systems of simple harmonic waves of different frequencies for combination, and then repeat the work N times, N>10000, sort the calculated second seismic amplification coefficient results from small to large, and determine the maximum data of the first 95% as the final second seismic amplification coefficient; Evaluate specific seismic waves, obtain the seismic amplification coefficients of the predicted mountainous area under the action of simple harmonic waves of various frequencies, combine them, and generate the seismic amplification coefficients under the action of specific seismic waves as the quantitative coefficients of the disaster hazards under the action of specific seismic waves in the predicted mountainous area; Or, the seismic amplification coefficient is evaluated based on Monte Carlo, and the envelope value of the seismic amplification coefficient of the mountain area to be predicted under the action of multiple groups of random seismic waves is calculated as the quantitative coefficient of the final disaster hidden danger of the mountain area to be predicted.

[0039] In this embodiment, evaluating a specific seismic wave includes: Based on the seismic amplification effect optimization neural network model, the seismic amplification coefficients of different data points of the terrain model under the action of simple harmonic waves of various frequencies are generated in batches; Select a specific seismic wave for quasi-Fourier transform. The formula of Fourier series is as follows: f(t)=a0 / 2+Σ(an*cos(nω0*t)+bn*sin(nω0*t)) Where f(t) is a periodic function, ω0 is the angular frequency, a0, an and bn are the Fourier coefficients respectively, and n is a positive integer.

[0040] Then, the seismic amplification coefficients of different sites under the action of simple harmonic waves of different frequencies obtained in the terrain model are combined to generate the seismic amplification coefficient under the action of specific seismic waves, which is used as the quantitative coefficient of the disaster potential under the action of the specific seismic waves of the site; The Monte Carlo-based ground motion amplification factor assessment includes: According to the assessment requirements, the Monte Carlo random method is used to generate random seismic harmonic wave related parameters such as wave number, frequency and amplitude. These parameters should cover the expected range of seismic wave characteristics.

[0041] Based on the seismic amplification effect optimization neural network model, the seismic amplification coefficients of different data points of the terrain model are obtained under the action of simple harmonic waves with random parameters. According to the generated random wave number, the seismic amplification coefficients under the action of simple harmonic waves with different parameters are combined to obtain the average seismic amplification coefficient of the terrain model, which is used as the quantitative coefficient of the disaster potential of the site under the action of a random seismic wave.

[0042] The prediction and composition of the seismic amplification coefficients are repeated many times to obtain multiple groups of randomly simulated seismic amplification coefficients of the terrain to be tested.

[0043] All simulation results are analyzed, and the envelope value of the seismic amplification factor of each site under the action of multiple groups of random seismic waves is calculated as the quantitative coefficient of the final disaster risk of the site.

[0044] According to the second embodiment of the present invention, the present invention claims a system for predicting earthquake hazards in mountainous areas based on drones and deep learning, comprising: one or more processors; A memory having one or more programs stored thereon, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method for predicting earthquake hazards in mountainous areas based on drones and deep learning.

[0045] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0046] In addition, each functional unit in each embodiment of the present application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above integrated unit can be implemented in the form of hardware or in the form of software functional units. The above is only an implementation method of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the specification and drawings of this application, or directly or indirectly used in other related technical fields, is also included in the patent protection scope of the present application.

[0047] The specific implementation methods of the invention are described in detail above, but they are only examples, and the present application is not limited to the specific implementation methods described above. For those skilled in the art, any equivalent modification or substitution of the invention is also within the scope of the present application, and therefore, the equalization, modification, and improvement made without departing from the spirit and principle of the present application should be included in the scope of the present application.

Claims

1. A method for predicting earthquake hazards in mountainous areas based on drones and deep learning, characterized in that: include: Using a drone to collect images and record data of the mountainous area to be predicted, obtaining image data of the mountainous area to be predicted, and constructing a three-dimensional terrain model to be measured based on the image data; Extracting terrain information of the mountainous area to be predicted from the three-dimensional terrain model to be measured, collecting geological information of the mountainous area to be predicted, and performing fundamental frequency prediction on the mountainous area to be predicted to obtain fundamental frequency information of the mountainous area; Predicting the seismic amplification effect of the mountainous area to be predicted based on the base frequency information, geological information and topographic information of the mountainous area, and obtaining the envelope value of the seismic amplification coefficient of each characteristic point in the mountainous area to be predicted under the action of seismic waves; The earthquake hazard level is classified by the earthquake amplification coefficient envelope value, and a distribution map of earthquake disaster hazards in the mountainous area to be predicted is drawn according to the earthquake hazard level classification.

2. The method for predicting earthquake hazards in mountainous areas based on drones and deep learning according to claim 1, characterized in that: The method of using a drone to collect images and record data of the mountainous area to be predicted, obtaining image data of the mountainous area to be predicted, and constructing a three-dimensional terrain model to be measured based on the image data, further includes: Acquire the characteristics of the mountainous terrain and earthquake hazards to be predicted, and plan the flight route and flight range of the UAV; Through the high-resolution camera sensor carried by the drone, all-round image acquisition and data recording of the mountainous area to be predicted are performed to ensure that image data of a preset quantity and quality are obtained; The image data taken by the drone is imported into the corresponding three-dimensional terrain construction software, submitted to the aerial triangulation calculation, and a three-dimensional terrain model to be measured is constructed.

3. The method for predicting earthquake hazards in mountainous areas based on drones and deep learning according to claim 1 is characterized in that: The terrain information of the mountainous area to be predicted includes at least: Terrain type, height, slope, local features.

4. The method for predicting earthquake hazards in mountainous areas based on drones and deep learning according to claim 1, characterized in that: The method of predicting the seismic amplification effect of the mountainous area to be predicted based on the base frequency information, geological information and terrain information of the mountainous area, and obtaining the envelope value of the seismic amplification coefficient of each characteristic point in the mountainous area to be predicted under the action of seismic waves, further includes: Input the acquired base frequency information, terrain information and geological information of the mountain area into the seismic amplification effect prediction module to realize the prediction of the amplification effect of simple harmonic waves of different frequencies, and output the first seismic amplification coefficient of each characteristic point in the mountain area to be predicted; The first seismic amplification coefficients of simple harmonic waves of different frequencies are input into the data integration module. The second seismic amplification coefficients under different seismic waves are obtained through parameter combination. The corresponding seismic amplification coefficient envelope value is calculated based on the second seismic amplification coefficients of each characteristic point under each seismic wave.

5. The method for predicting earthquake hazards in mountainous areas based on drones and deep learning according to claim 1, characterized in that: The method further comprises: performing earthquake hazard level classification by using the earthquake motion amplification coefficient envelope value, and drawing an earthquake disaster hazard distribution map of the mountainous area to be predicted according to the earthquake hazard level classification. Input the obtained envelope value of the seismic motion amplification coefficient of each test point into the regional earthquake disaster hazard level classification module, and output the earthquake hazard hazard level of the corresponding test point by judging the seismic motion amplification coefficient of the test point point by point; According to the terrain information, the seismic amplification coefficients and disaster risks of different test points are graded using a post-processing imaging module, and plotted point by point on the three-dimensional test terrain model according to the data distribution points to obtain a regional earthquake disaster risk distribution map.

6. The method for predicting earthquake hazards in mountainous areas based on drones and deep learning according to claim 1, characterized in that: When constructing a three-dimensional terrain model to be measured based on the image data, it also includes: Dividing the image data into aerial images and ground point cloud data; Performing image preprocessing on the aerial images and ground point cloud data; The image preprocessing at least includes: Dedistortion processing, point cloud registration, 3D reconstruction, texture mapping, and data fusion.

7. The method for predicting earthquake hazards in mountainous areas based on drones and deep learning according to claim 1, characterized in that: When performing fundamental frequency prediction on the mountainous area to be predicted to obtain fundamental frequency information of the mountainous area, the method further includes: Collecting the measured earthquake acceleration time history curve, determining the geological feature label of the mountainous area to be predicted, and constructing a site fundamental frequency database; Determine an initial network, select fundamental frequency sample data from the site fundamental frequency database, train and optimize the initial network based on the fundamental frequency sample data, and obtain a site fundamental frequency prediction optimization neural network model; The required data of the mountainous area to be predicted is input into the site fundamental frequency prediction optimization neural network, and the site dominant frequencies of different sites are output.

8. The method for predicting earthquake hazards in mountainous areas based on drones and deep learning according to claim 4, characterized in that: The method further comprises: inputting the acquired base frequency information, terrain information and geological information of the mountain area into a seismic amplification effect prediction module to predict the amplification effect of simple harmonic waves of different frequencies, and outputting the first seismic amplification coefficient of each characteristic point in the mountain area to be predicted. Collecting measured earthquake acceleration time history curves, determining geological feature labels of the mountainous area to be predicted, and constructing a seismic motion amplification effect database; Determine an initial network, select fundamental frequency sample data from the earthquake amplification effect database, train and optimize the initial network according to the fundamental frequency sample data, and obtain an earthquake amplification effect optimized neural network model; The required data of the mountainous area to be predicted is input into the seismic amplification effect optimization neural network model, and the seismic amplification coefficients of different sites under the action of simple harmonic waves of various frequencies are output.

9. The method for predicting earthquake hazards in mountainous areas based on drones and deep learning according to claim 4, characterized in that: The first seismic amplification coefficients of simple harmonic waves of different frequencies are input into the data integration module, and the second seismic amplification coefficients under different seismic waves are obtained by parameter combination, and the corresponding seismic amplification coefficient envelope value is calculated according to the second seismic amplification coefficients of each characteristic point under each seismic wave, and further includes: ； After determining the range of n, randomly generate it through the random() function, and randomly select n first seismic amplification systems from the first seismic amplification systems of simple harmonic waves of different frequencies for combination, and then repeat the work N times, N>10000, sort the calculated second seismic amplification coefficient results from small to large, and determine the maximum data of the first 95% as the final second seismic amplification coefficient; Evaluate specific seismic waves, obtain the seismic amplification coefficients of the predicted mountainous area under the action of simple harmonic waves of various frequencies, combine them, and generate the seismic amplification coefficients under the action of specific seismic waves as the quantitative coefficients of the disaster hazards under the action of specific seismic waves in the predicted mountainous area; Or, the seismic amplification coefficient is evaluated based on Monte Carlo, and the envelope value of the seismic amplification coefficient of the mountain area to be predicted under the action of multiple groups of random seismic waves is calculated as the quantitative coefficient of the final disaster hidden danger of the mountain area to be predicted.

10. A system for predicting earthquake hazards in mountainous areas based on drones and deep learning, characterized in that: include: one or more processors; A memory having one or more programs stored thereon, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement a method for predicting earthquake hazards in mountainous areas based on drones and deep learning according to any one of claims 1 to 9.

Images