Intelligent identification method for high-position collapse geological disasters

Abstract

The invention discloses an intelligent identification method for high-position collapse geological disasters. The method comprises the steps that INSAR monitoring data, a radar topographic map and a remote sensing image map of a to-be-identified area are acquired; obtaining surface deformation information according to the INSAR monitoring data, and determining a hidden danger area according to the surface deformation information and a radar topographic map; registering the remote sensing image map of the hidden danger area based on the radar topographic map; and inputting the registered remote sensing image into a high-position collapse and slip identification model constructed based on CNN and RNN networks, and carrying out accurate identification on the high-position collapse and slip geological disaster point. According to the method, information of different data sources is fully utilized, the comprehensive capability and accuracy of geological disaster monitoring are improved, geological disaster monitoring is carried out on high-altitude collapse through multi-level and multi-angle analysis in combination with a deep learning network, the geological disaster occurrence place in a high-altitude area can be accurately recognized with high precision, and the geological disaster monitoring efficiency is improved. And scientific support is provided for geological disaster prevention and control work in high-altitude areas.

Description

Technical Field

[0001] The present invention relates to the technical field of geological disaster monitoring, and in particular to a method for intelligently identifying high-position landslide geological disasters. Background Art

[0002] Early identification of geological disasters is an important prerequisite for monitoring, assessment, prevention and control. High-level landslide disasters generally refer to the shearing out of the rock and soil in the upper part of a high and steep slope, which has the characteristics of high center of gravity, large potential energy, fast speed and great harm. Therefore, how to discover and effectively identify the potential hidden dangers of high-level landslide geological disasters in advance and take active prevention and control measures has become the focus and difficulty of the recent field of geological disaster prevention and control.

[0003] There are several ways to identify high-level landslides at home and abroad. Geological exploration and geophysical detection technology are used to obtain underground geological information, such as groundwater level, groundwater permeability and other parameters, and combined with geological structure and topography and geomorphology to analyze and infer potential high-level landslides. Although this method has a high recognition accuracy, it requires a large investment of time and manpower costs, and is highly restricted by the terrain, which is not conducive to large-scale identification. The method of modeling and simulating geological bodies using numerical models and geomechanical models can simulate the occurrence process of geological disasters under different conditions, and predict and evaluate high-level landslides. This method can predict the possibility and scale of landslides by simulating the influence of different factors. However, numerical simulation and simulation require the setting of various model parameters, such as material parameters, boundary conditions, etc. Different parameter selections may lead to different prediction results, and the rationality of parameter selection has a great impact on the prediction results. Limited by traditional technologies, the accuracy and timeliness of existing geological disaster identification are difficult to meet the needs of disaster prevention and mitigation in the southeastern margin of the Qinghai-Tibet Plateau.

[0004] Therefore, it is necessary to propose an intelligent identification method for high-altitude landslide geological disasters, which can comprehensively utilize SAR data, optical remote sensing satellite data, geological disaster survey results, basic geographic information and other materials to quickly and accurately identify high-altitude landslide geological disasters in high-altitude areas. Summary of the invention

[0005] In view of this, the present invention provides an intelligent identification method for high-level landslide geological disasters, which is used to solve the technical problems in the prior art that due to the difficulty in data acquisition and slow update speed, the identification of high-level geological disasters is too dependent on expert experience or specific rules, and it is difficult to achieve timely and accurate high-level landslide monitoring and early warning.

[0006] In order to achieve the above technical objectives, the present invention adopts the following technical solutions:

[0007] In a first aspect, the present invention provides a method for intelligently identifying high-level collapse and landslide geological disasters, comprising:

[0008] Obtain INSAR monitoring data, radar topographic maps and remote sensing images of the area to be identified;

[0009] Obtaining surface deformation information according to the INSAR monitoring data, and determining a potential hazard area according to the surface deformation information and a radar topographic map;

[0010] Registering the remote sensing image of the hidden danger area based on the radar topographic map;

[0011] The registered remote sensing image is input into the high-position landslide recognition model constructed based on CNN and RNN network to accurately identify the high-position landslide geological disaster points.

[0012] Furthermore, the surface deformation information is obtained according to the INSAR monitoring data, including:

[0013] Dividing the area to be identified into a plurality of sub-areas, and setting a target observation object in each sub-area;

[0014] Acquire the SAR monitoring data of the target observation object in each sub-area at multiple different time points, and perform coherent registration and differential interferometry processing on every two SAR monitoring data separated by a preset time;

[0015] The data after differential interferometry processing is subjected to noise filtering, and the phase unwrapping of the noise-filtered differential interferometry data is performed to obtain the surface deformation information.

[0016] Furthermore, the noise filtering of the data after differential interferometry processing includes:

[0017] The INSAR terrain simulation phase is synthesized by using the prior DEM data, and the phase change caused by the terrain change in the data after the differential interferometry processing is removed according to the terrain simulation phase.

[0018] Furthermore, the basic framework of the high-level landslide recognition model is an RNN network, including an input layer, a hidden layer and an output layer connected in sequence, and a recurrent layer connected to the hidden layer;

[0019] A CNN network is embedded in the circulation layer, and the CNN network includes four convolutional layers, four batch normalization layers and four pooling layers, and a batch normalization layer, a pooling layer and a ReLu activation function are connected between every two convolutional layers;

[0020] The output layer of the RNN network is connected to the fully connected network, and the output of the fully connected network is the output result of the high-level landslide recognition model.

[0021] Furthermore, an attention module is also included between the RNN network and the fully connected layer;

[0022] The attention module is used to calculate the attention scores of multiple output values ​​of the RNN network and perform weighted summation, and input the summation result into the fully connected layer.

[0023] Further, determining the potential danger area according to the surface deformation information and the radar topographic map includes:

[0024] Superimposing the landmark deformation information and the radar topographic map to obtain a superimposed deformation map;

[0025] Calculate the ratio of each pixel value of the superimposed deformation map to the corresponding pixel value in the radar topographic map;

[0026] When the ratio of any area is greater than a preset threshold, the area is marked as a potential hazard area.

[0027] Furthermore, the remote sensing image of the hidden danger area is registered based on the radar topographic map and the surface deformation information, including:

[0028] Taking the target observation object in the hidden danger area as a registration point, determining a plurality of feature points related to the registration point in the radar topographic map;

[0029] The remote sensing image and the radar topographic map of the hidden danger area are registered according to the plurality of feature points.

[0030] In a second aspect, the present invention further provides an intelligent identification device for high-position landslide geological disasters, comprising:

[0031] Data acquisition module, used to obtain INSAR monitoring data, radar topographic maps and remote sensing images of the area to be identified;

[0032] A comprehensive analysis module, used to obtain surface deformation information according to the INSAR monitoring data, and determine the potential danger area according to the surface deformation information and the radar topographic map;

[0033] A registration module, used for registering the remote sensing image of the hidden danger area based on the radar topographic map;

[0034] The output module is used to input the registered remote sensing image into the high-position landslide recognition model built based on CNN and RNN networks to accurately identify the high-position landslide geological disaster points.

[0035] In a third aspect, the present invention further provides an electronic device, comprising a processor and a memory, wherein the memory stores a computer program, and when the computer program is executed by the processor, the intelligent identification method for high-level landslide geological disasters described in any of the above technical solutions is implemented.

[0036] In a fourth aspect, the present invention also provides a computer-readable storage medium for storing computer-readable programs or instructions. When the programs or instructions are executed by a processor, the steps of the intelligent identification method of high-level landslide geological disasters in any of the above-mentioned implementation methods can be implemented.

[0037] The present invention provides a method for intelligently identifying high-position landslide geological disasters. The method firstly comprehensively acquires INSAR data, radar topographic maps and remote sensing images. Secondly, by combining INSAR monitoring data and radar topographic maps, potential high-position landslide hazard areas can be more accurately determined, the observation range can be narrowed, and the accuracy of early warning and identification can be improved. Finally, the remote sensing image of the hazard area is registered based on the radar topographic map, and the registered remote sensing image is input into a high-position landslide identification model constructed based on CNN and RNN networks to locate the high-position landslide geological disaster points, and the deep learning technology is used to perform intelligent identification and positioning of high-position landslides, thereby improving the automation and accuracy of identification. This method comprehensively utilizes multiple data sources such as INSAR monitoring data, radar topographic maps and remote sensing images, and can obtain surface information from different angles and scales, thereby improving the comprehensiveness and accuracy of geological disaster identification. By obtaining surface deformation information through INSAR monitoring data and registering remote sensing images using radar topographic maps, the spatial deviation between different data sources can be eliminated, and the precise positioning of geological disaster risk areas can be achieved, thereby improving the precision and accuracy of geological disaster identification. The high-position landslide identification model constructed based on CNN and RNN networks can learn the characteristics of geological disasters from a large amount of remote sensing image data, realize automated geological disaster identification and positioning, and improve identification efficiency and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 A schematic diagram of the process of the intelligent identification method of high-level collapse and landslide geological disasters provided by the present invention;

[0039] Figure 2 A flow chart of an embodiment of obtaining surface deformation information based on INSAR monitoring data provided by the present invention;

[0040] Figure 3 A schematic diagram of the structure of an embodiment of a high-level collapse and landslide identification model provided by the present invention;

[0041] Figure 4 A schematic diagram of the structure of an embodiment of an intelligent identification device for high-level landslide geological disasters provided by the present invention;

[0042] Figure 5 A schematic structural diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0043] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not used to limit the scope of the present invention.

[0044] Before introducing the embodiments of the present invention, the relevant terms of the present application are first explained.

[0045] SAR: Synthetic Aperture Radar is an active radar system that is used to obtain ground information by emitting microwave signals and receiving their reflected signals. Unlike traditional optical remote sensing technology, SAR can conduct surface observations in any weather conditions (such as clouds, rain and fog), and has a certain penetration ability, which can penetrate vegetation and soil surface to obtain information below the surface.

[0046] InSAR: Interferometric Synthetic Aperture Radar uses satellite-borne radar sensors to continuously observe the earth's surface. By measuring the phase difference of radar signals at different times at the same location, it can monitor and measure changes in surface deformation and subsidence. This technology has important application value in the fields of geological disaster monitoring, crustal movement research, and urban subsidence monitoring.

[0047] DEM: Digital Elevation Model is a digital surface elevation data model used to describe the height and shape of the earth's surface. DEM data is usually stored in raster or vector form, containing surface elevation information, and can be used in various geographic information system (GIS) analysis, terrain modeling, hydrological simulation, geological exploration and other fields.

[0048] The present invention provides a method, device, electronic device and computer-readable storage device for intelligent identification of high-level collapse and landslide geological disasters, which are described below respectively.

[0049] Combination Figure 1 As shown, a specific embodiment of the present invention discloses a method for intelligently identifying high-position landslide geological disasters, including:

[0050] Step S101, obtaining INSAR monitoring data, radar topographic map and remote sensing image map of the area to be identified;

[0051] Step S102, obtaining surface deformation information according to the INSAR monitoring data, and determining a potential hazard area according to the surface deformation information and a radar topographic map;

[0052] Step S103, registering the remote sensing image of the hidden danger area based on the radar topographic map;

[0053] Step S104: input the registered remote sensing image into a high-position landslide recognition model constructed based on CNN and RNN networks to accurately identify high-position landslide geological disaster points.

[0054] Compared with the prior art, the method of this embodiment firstly comprehensively acquires INSAR data, radar topographic maps and remote sensing images; secondly, by combining INSAR monitoring data and radar topographic maps, potential high-level collapse and landslide hazard areas can be more accurately determined, the observation range can be narrowed, and the accuracy of early warning and identification can be improved; finally, the remote sensing image map of the hazard area is aligned based on the radar topographic map, and the aligned remote sensing image map is input into the high-level collapse and landslide identification model constructed based on CNN and RNN networks to locate the high-level collapse and landslide geological disaster points, and the deep learning technology is used for intelligent identification and positioning of high-level collapse and landslide, thereby improving the automation and accuracy of identification. The method of this embodiment comprehensively utilizes multiple data sources such as INSAR monitoring data, radar topographic maps and remote sensing images, and can obtain surface information from different angles and scales, thereby improving the comprehensiveness and accuracy of geological disaster identification; obtaining surface deformation information through INSAR monitoring data and registering remote sensing images using radar topographic maps can eliminate the spatial deviation between different data sources, and can achieve accurate positioning of geological disaster risk areas, thereby improving the precision and accuracy of geological disaster identification; the high-level collapse and landslide identification model constructed based on CNN and RNN networks can learn the characteristics of high-level collapse and landslide geological disasters from a large amount of remote sensing image data, realize automated geological disaster identification and positioning, and improve identification efficiency and accuracy.

[0055] As a specific embodiment, the remote sensing image is a high-definition remote sensing image obtained by a drone. The drone can complete low-altitude flight, avoiding interference with the image caused by weather and clouds, and is convenient, fast, and highly maneuverable, and can provide effective technical support for geological disaster monitoring.

[0056] As a preferred embodiment, obtaining the surface deformation information according to the INSAR monitoring data includes:

[0057] Dividing the area to be identified into a plurality of sub-areas, and setting a target observation object in each sub-area;

[0058] Acquire the SAR monitoring data of the target observation object in each sub-area at multiple different time points, and perform coherent registration and differential interferometry processing on every two SAR monitoring data separated by a preset time;

[0059] The data after differential interferometry processing is subjected to noise filtering, and the phase unwrapping of the noise-filtered differential interferometry data is performed to obtain the surface deformation information.

[0060] By dividing the monitoring area into multiple sub-areas and setting target observations in each sub-area, the computational complexity is reduced, the processing efficiency is improved, the monitoring target can be located more accurately, and the impact of external interference factors on the monitoring results can be reduced. By coherently registering and differential interferometry processing every two SAR monitoring data separated by a preset time, surface deformation information with higher temporal and spatial resolution can be obtained, so as to more accurately monitor and analyze the temporal and spatial evolution of surface deformation. Noise filtering of the data after differential interferometry processing can reduce the impact of noise on surface deformation information and improve the reliability and accuracy of the data.

[0061] As a specific example, Figure 2 As shown, Figure 2 The flowchart of obtaining surface deformation information based on INSAR monitoring data is shown.

[0062] The original deformation information obtained by differential interferometry processing generally includes deformation phase, reference ellipsoid phase, terrain phase, atmospheric delay phase, noise phase, etc. To extract the deformation phase from the original deformation information, the interfering reference ellipsoid phase and terrain phase must be removed from the original phase, that is, secondary differential interferometry is required.

[0063] Specifically, the secondary differential interferometry is performed by a two-track method. As a preferred embodiment, the noise filtering of the data after the differential interferometry processing includes:

[0064] The INSAR terrain simulation phase is synthesized by using the prior DEM data, and the phase change caused by the terrain change in the data after the differential interferometry processing is removed according to the terrain simulation phase.

[0065] By combining prior DEM data with INSAR data, the accuracy and precision of terrain phase can be improved. Externally obtained prior DEM data may have higher spatial resolution and accuracy than INSAR data, which can further improve the spatiotemporal resolution of surface deformation monitoring.

[0066] like Figure 3 As shown, Figure 3 The figure is a possible structural diagram of the high-level landslide recognition model, in which the batch normalization layer and the ReLu activation function are omitted. As a preferred embodiment, the basic framework of the high-level landslide recognition model is an RNN network, including an input layer, a hidden layer and an output layer connected in sequence, and a recurrent layer connected to the hidden layer;

[0067] A CNN network is embedded in the circulation layer, and the CNN network includes four convolutional layers, four batch normalization layers and four pooling layers, and a batch normalization layer, a pooling layer and a ReLu activation function are connected between every two convolutional layers;

[0068] The output layer of the RNN network is connected to the fully connected network, and the output of the fully connected network is the output result of the high-level landslide recognition model.

[0069] The CNN network extracts surface features such as texture and shape at different scales, and the RNN framework is used to capture geological change information in the time series. This model can fully extract the characteristics of geological hazards at different scales, making the identification model more comprehensive and accurate, making full use of spatial and temporal information, and more comprehensively understanding and analyzing data.

[0070] The above process is explained below using a specific example.

[0071] Assume that the input remote sensing image is X, and the image after registration is X_r.

[0072] Input the registered remote sensing image Xr to the input layer. Assume that the four convolutional layers in the CNN network are C1, C2, C3, and C4, the four batch normalization layers are BN1, BN2, BN3, and BN4, and the four pooling layers are P1, P2, P3, and P4. Assume that the output of each convolutional layer is H1, H2, H3, and H4.

[0073] Then: the output from the input layer to the C1 layer: H1 = ReLU (C1 (X_r));

[0074] Output from C1 layer to BN1 layer: H1-nor = BatchNormalization(H1);

[0075] Output from BN1 layer to P1 layer: H1-pooled = Pooling (H1-nor);

[0076] Output from P1 layer to C2 layer: H2 = ReLU(C2(H1_pooled))

[0077] And so on, we get H2, H3, and H4.

[0078] RNN network: The input of the RNN network is the output sequence H1, H2, H3, H4 of the CNN network. Assume that the hidden layer output sequence of the RNN network is R1, R2, R3, R4.

[0079] Output from the input layer to the R1 layer: R1 = RNN (H1)

[0080] Output from layer R1 to layer R2: R2 = RNN (H2, R1)

[0081] And so on, we get R2, R3, and R4.

[0082] Fully connected network: Assuming the output sequence of the RNN network is R_output, the output of the fully connected network is Y. Y = FC (R_output).

[0083] In some cases, the input sequence of the RNN network also includes the precipitation V corresponding to the registered remote sensing image, as well as the slope T and slope M to determine the corresponding influence weights. At this time, in the RNN network, the input sequence includes the output feature sequence of the CNN network, as well as precipitation (V), slope (T) and slope (M). At each time step, consider the hidden state (R i-1 ) and the input features of the current time step, and calculate the hidden state of the current time step. The calculation formula of the RNN hidden layer is: R i =RNN(H i R i-1 ,V,T,M), where (i=1,2,3,4).

[0084] Precipitation is one of the main causes of landslides. Large amounts of precipitation increase the moisture content in the soil, making it soft and increasing the risk of landslides. Slope aspect and gradient affect soil stability. For the same amount of precipitation, steep slopes and slope aspects that are not conducive to water drainage increase the risk of landslides. Generally speaking, the steeper the slope and the less conducive the slope is to drainage, the higher the risk of landslides.

[0085] In addition, the surface state and soil properties are also important state quantities. The surface state includes factors such as soil type, vegetation coverage, and groundwater level.

[0086] As a preferred embodiment, the RNN network and the fully connected layer further include an attention module;

[0087] The attention module is used to calculate the attention scores of multiple output values ​​of the RNN network and perform weighted summation, and input the summation result into the fully connected layer.

[0088] The attention mechanism can assign different attention weights to different parts of the input data, allowing the model to focus more on the information that is most important to the current task. By performing weighted summation of multiple output values ​​of the RNN network, the model can more effectively capture key features and improve the model's representation ability and prediction accuracy.

[0089] It should be noted that in order to improve the robustness and generalization ability of the network and reduce overfitting of the training data, a Dropout layer can be inserted after the fully connected layer or convolutional layer and before the activation function.

[0090] Assume that the input sequence is X=(X 1 ,X 2 ,…,X t ), after RNN processing, the hidden state sequence H = (H 1 ,H 2 ,…,H t ), where t is the sequence length.

[0091] Then the attention weight is:

[0092]

[0093] Among them, e t is the hidden state h t The scores obtained by attention scoring can be normalized by softmax using different scoring functions such as dot product, addition or other forms of functions to obtain attention weights.

[0094] Use the attention weights to perform a weighted sum of the hidden states:

[0095]

[0096] The attention-weighted hidden state (c) is used as the input of the fully connected layer to continue the subsequent prediction or classification task. In this way, it is possible to dynamically focus on different parts of the input sequence, increase the focus on key features, and thus improve the expressiveness and performance of the model.

[0097] In practice, the key features of interest include at least terrain elevation, slope and aspect, rock properties, groundwater level, meteorological data, etc. Among them, terrain elevation can provide information about terrain, such as mountains, canyons, plains, etc.; rock properties include density, porosity, toughness, strength and other parameters; meteorological data include temperature, humidity, wind speed, rainfall and other information; groundwater level can provide information about hydrogeology, such as whether groundwater accumulates and the direction of flow.

[0098] As a preferred embodiment, determining the potential hazard area according to the surface deformation information and the radar topographic map includes:

[0099] Superimposing the landmark deformation information and the radar topographic map to obtain a superimposed deformation map;

[0100] Calculate the ratio of each pixel value of the superimposed deformation map to the corresponding pixel value in the radar topographic map;

[0101] When the ratio of any area is greater than a preset threshold, the area is marked as a potential hazard area.

[0102] Specifically, the image registration technology can be used to align the landmark deformation information and the radar topographic map to the same coordinate system to achieve the registration operation and ensure that the spatial positions of the two are consistent. For each pixel point, the ratio of the pixel value in the superimposed deformation map to the corresponding pixel value in the radar topographic map is calculated. Based on historical data and geological analysis, a preset threshold is defined for each terrain and region. When the ratio of a certain area exceeds the threshold, the area is marked as a hidden danger area. These areas may be affected by geological disasters, surface subsidence, etc., and require further investigation and monitoring.

[0103] As a preferred embodiment, registering the remote sensing image of the hidden danger area based on the radar topographic map and the surface deformation information includes:

[0104] Taking the target observation object in the hidden danger area as a registration point, determining a plurality of feature points related to the registration point in the radar topographic map;

[0105] The remote sensing image and the radar topographic map of the hidden danger area are registered according to the plurality of feature points.

[0106] Furthermore, according to the position information of the selected registration points in different images, a feature point matching algorithm (such as SIFT, SURF, ORB, etc.) is used to automatically identify the corresponding feature points, and registration is performed according to the matching relationship of these feature points; or a geographic information system (GIS) tool or image processing software is used to extract the geographic coordinates (latitude and longitude or projection coordinates) of the selected registration points in each image.

[0107] In some embodiments, an image cross-correlation method may be used to calculate the similarity between different images, and the registration transformation parameters may be optimized by maximizing the similarity.

[0108] In order to better implement the intelligent identification method of high-level collapse and landslide geological disasters in the embodiment of the present invention, based on the intelligent identification method of high-level collapse and landslide geological disasters, correspondingly, please refer to Figure 4 , Figure 4 The structure diagram of an embodiment of the intelligent identification device for high-position landslide geological disasters provided by the present invention is as follows. The intelligent identification device for high-position landslide geological disasters 400 provided by the embodiment of the present invention includes:

[0109] The model building module 401 is used to obtain INSAR monitoring data, radar topographic maps and remote sensing image maps of the area to be identified;

[0110] A data acquisition module 402 is used to obtain surface deformation information according to the INSAR monitoring data, and determine a potential hazard area according to the surface deformation information and a radar topographic map;

[0111] A training module 403 is used to register the remote sensing image of the hidden danger area based on the radar topographic map;

[0112] The output module 404 is used to input the registered remote sensing image into the high-position landslide recognition model constructed based on the CNN and RNN networks to accurately identify the high-position landslide geological disaster points.

[0113] It should be noted here that the corresponding device 400 provided in the above embodiment can implement the technical solutions described in the above method embodiments. The specific implementation principles of the above modules or units can refer to the corresponding contents in the above method embodiments, which will not be repeated here.

[0114] like Figure 5 As shown, the above-mentioned intelligent identification method of high-level collapse and landslide geological disasters, the present invention also provides an electronic device 500, which can be a computing device such as a mobile terminal, a desktop computer, a notebook, a palm computer and a server. The electronic device includes a processor 501, a memory 502 and a display 503.

[0115] In some embodiments, the memory 502 may be an internal storage unit of a computer device, such as a hard disk or memory of a computer device. In other embodiments, the memory 502 may also be an external storage device of a computer device, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), etc. equipped on the computer device. Further, the memory 502 may also include both an internal storage unit of a computer device and an external storage device. The memory 502 is used to store application software and various types of data installed in the computer device, such as program codes for installing the computer device. The memory 502 may also be used to temporarily store data that has been output or is to be output. In one embodiment, a high-level collapse and sliding geological disaster intelligent identification method program 804 is stored on the memory 502, and the high-level collapse and sliding geological disaster intelligent identification method program 804 can be executed by the processor 501, thereby realizing a high-level collapse and sliding geological disaster intelligent identification method of each embodiment of the present invention.

[0116] In some embodiments, the processor 501 can be a central processing unit (CPU), a microprocessor or other data processing chip, which is used to run the program code or process data stored in the memory 502, such as executing a method program for intelligent identification of high-level landslide geological disasters.

[0117] In some embodiments, the display 503 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode) touch device, etc. The display 503 is used to display information on the computer device and to display a visual user interface. The components 501-503 of the computer device communicate with each other through a system bus.

[0118] This embodiment also provides a computer-readable storage medium, on which a high-level collapse and landslide geological disaster intelligent identification program is stored. When the high-level collapse and landslide geological disaster intelligent identification program is executed by a processor, the steps in the above embodiment can be implemented.

[0119] The method of the present invention firstly obtains INSAR data, radar topographic maps and remote sensing images comprehensively; secondly, by combining INSAR monitoring data and radar topographic maps, potential high-position collapse and landslide hazard areas can be more accurately determined, the observation range can be narrowed, and the accuracy of early warning and identification can be improved; finally, the remote sensing image map of the hazard area is registered based on the radar topographic map, and the registered remote sensing image map is input into a high-position collapse and landslide identification model constructed based on CNN and RNN networks, the high-position collapse and landslide geological disaster points are located, and the deep learning technology is used to perform intelligent identification and positioning of high-position collapse and landslide, thereby improving the automation and accuracy of identification.

[0120] This method comprehensively utilizes multiple data sources such as INSAR monitoring data, radar topographic maps and remote sensing images, and can obtain surface information from different angles and scales, thereby improving the comprehensiveness and accuracy of geological disaster identification. By obtaining surface deformation information through INSAR monitoring data and registering remote sensing images using radar topographic maps, the spatial deviation between different data sources can be eliminated, and the precise positioning of geological disaster risk areas can be achieved, thereby improving the precision and accuracy of geological disaster identification. The high-position landslide identification model constructed based on CNN and RNN networks can learn the characteristics of geological disasters from a large amount of remote sensing image data, realize automated geological disaster identification and positioning, and improve identification efficiency and accuracy.

[0121] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. A method for intelligent identification of high-level landslide geological disasters, characterized in that: include: Obtain INSAR monitoring data, radar topographic maps and remote sensing images of the area to be identified; Obtaining surface deformation information according to the INSAR monitoring data, and determining a potential hazard area according to the surface deformation information and a radar topographic map; Registering the remote sensing image of the hidden danger area based on the radar topographic map; The registered remote sensing image is input into the high-position landslide recognition model constructed based on CNN and RNN network to accurately identify the high-position landslide geological disaster points.

2. The intelligent identification method of high-level collapse and landslide geological disasters according to claim 1 is characterized in that: The surface deformation information obtained according to the INSAR monitoring data includes: Dividing the area to be identified into a plurality of sub-areas, and setting a target observation object in each sub-area; Acquire the SAR monitoring data of the target observation object in each sub-area at multiple different time points, and perform coherent registration and differential interferometry processing on every two SAR monitoring data separated by a preset time; The data after differential interferometry processing is subjected to noise filtering, and the phase unwrapping of the noise-filtered differential interferometry data is performed to obtain the surface deformation information.

3. The intelligent identification method of high-level landslide geological disasters according to claim 2 is characterized in that: Noise filtering of data after differential interferometry processing includes: The INSAR terrain simulation phase is synthesized by using the prior DEM data, and the phase change caused by the terrain change in the data after the differential interferometry processing is removed according to the terrain simulation phase.

4. The intelligent identification method of high-level collapse and landslide geological disasters according to claim 1 is characterized in that: The basic framework of the high-level collapse and landslide recognition model is an RNN network, which includes an input layer, a hidden layer and an output layer connected in sequence, and a circulation layer connected to the hidden layer; A CNN network is embedded in the circulation layer, and the CNN network includes four convolutional layers, four batch normalization layers and four pooling layers, and a batch normalization layer, a pooling layer and a ReLu activation function are connected between every two convolutional layers; The output layer of the RNN network is connected to the fully connected network, and the output of the fully connected network is the output result of the high-level landslide recognition model.

5. The intelligent identification method of high-level landslide geological disasters according to claim 4 is characterized in that: An attention module is also included between the RNN network and the fully connected layer; The attention module is used to calculate the attention scores of multiple output values ​​of the RNN network and perform weighted summation, and input the summation result into the fully connected layer.

6. The intelligent identification method of high-level collapse and landslide geological disasters according to claim 1 is characterized in that: Determine the potential hazard area based on the surface deformation information and radar topographic map, including: Superimposing the landmark deformation information and the radar topographic map to obtain a superimposed deformation map; Calculate the ratio of each pixel value of the superimposed deformation map to the corresponding pixel value in the radar topographic map; When the ratio of any area is greater than a preset threshold, the area is marked as a potential hazard area.

7. The intelligent identification method of high-level collapse and landslide geological disasters according to claim 2 is characterized in that: The remote sensing image of the hidden danger area is registered based on the radar topographic map and the surface deformation information, including: Taking the target observation object in the hidden danger area as a registration point, determining a plurality of feature points related to the registration point in the radar topographic map; The remote sensing image and the radar topographic map of the hidden danger area are registered according to the plurality of feature points.

8. An intelligent identification device for high-level landslide geological disasters, characterized in that: include: Data acquisition module, used to obtain INSAR monitoring data, radar topographic maps and remote sensing images of the area to be identified; A comprehensive analysis module, used to obtain surface deformation information according to the INSAR monitoring data, and determine the potential danger area according to the surface deformation information and the radar topographic map; A registration module, used for registering the remote sensing image of the hidden danger area based on the radar topographic map; The output module is used to input the registered remote sensing image into the high-position landslide recognition model built based on CNN and RNN networks to accurately identify the high-position landslide geological disaster points.

9. An electronic device, characterized in that: It includes a processor and a memory, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the intelligent identification method for high-level collapse and landslide geological disasters as described in any one of claims 1-7 is implemented.

10. A computer-readable storage medium, characterized in that: Used to store computer-readable programs or instructions, which, when executed by a processor, can implement the steps of the method for intelligent identification of high-level landslide geological disasters as described in any one of claims 1 to 7.

Images