Landslide identification method and device

Abstract

The invention provides a landslide identification method and device. The method comprises the following steps: acquiring a plurality of radar images; extracting target deformation time sequence data of the earth surface from the plurality of radar images; acquiring remote sensing time sequence data; generating target NDVI (normalized difference vegetation index) time sequence data according to a plurality of remote sensing data in the remote sensing time sequence data; and according to the target deformation time sequence data and the target NDVI time sequence data, landslide prediction is carried out on the earth surface to obtain a landslide prediction result, so that the dynamic change of the earth surface can be more comprehensively reflected through fusion analysis of the multi-source data in combination with the change of the earth surface deformation and the vegetation index, and the accuracy of landslide prediction is improved. And radar images and remote sensing data are periodically acquired, so that the monitoring system can acquire earth surface change information in real time, timely discover potential landslide risks and provide precious time for emergency response, and loss caused by landslide disasters can be reduced to the greatest extent.

Description

Technical Field

[0001] The present invention relates to the field of geological disaster monitoring and prediction, and in particular to a landslide identification method and device. Background Art

[0002] With global climate change, frequent extreme weather events and accelerated urbanization, geological structures such as mountains and slopes in many areas are facing greater pressure, resulting in an increasing frequency and scale of landslides. Landslides not only pose a serious threat to the safety of life and property of local residents, but may also damage infrastructure such as transportation and water conservancy, leading to serious economic losses and social problems. By predicting or identifying landslides on the surface, issuing warnings in a timely manner based on accurate landslide prediction results, and making landslide decisions, the losses caused by landslide disasters can be minimized. Therefore, it is very important to accurately predict or identify landslides on the surface. Summary of the invention

[0003] The purpose of the present disclosure is to solve one of the above technical problems at least to some extent.

[0004] To this end, the first purpose of the present invention is to propose a landslide identification method, by extracting target deformation time series data and target Normalized Difference Vegetation Index (NDVI) time series data of the surface based on multiple radar images and remote sensing time series data, and predicting landslides on the surface. Therefore, through the fusion analysis of multi-source data, combined with the changes in surface deformation and vegetation index, it is possible to more comprehensively reflect the dynamic changes of the surface and improve the accuracy of landslide prediction. In addition, the periodic collection of radar images and remote sensing data enables the monitoring system to obtain surface change information in real time, timely discover potential landslide risks, and provide valuable time for emergency response, thereby minimizing the losses caused by landslide disasters.

[0005] The second objective of the present disclosure is to provide a landslide identification device.

[0006] A third objective of the present disclosure is to provide an electronic device.

[0007] A fourth objective of the present disclosure is to provide a computer-readable storage medium.

[0008] To achieve the above-mentioned purpose, the first aspect of the present disclosure proposes a landslide identification method, comprising: acquiring multiple radar images; wherein the radar images are obtained by collecting the surface through a radar satellite in response to reaching a set period; extracting target deformation time series data of the surface from the multiple radar images; acquiring remote sensing time series data; wherein the remote sensing time data sequence includes remote sensing data obtained by collecting the surface through a remote sensing satellite at multiple time points; generating target normalized difference vegetation index NDVI time series data based on multiple remote sensing data in the remote sensing time series data; and predicting a landslide on the surface based on the target deformation time series data and the target NDVI time series data to obtain a landslide prediction result.

[0009] To achieve the above-mentioned purpose, the second aspect of the present disclosure proposes a landslide identification device, including: a first acquisition module, used to acquire multiple radar images; wherein the radar images are obtained by collecting the surface through a radar satellite in response to reaching a set period; a first extraction module, used to extract target deformation time series data of the surface from the multiple radar images; a second acquisition module, used to acquire remote sensing time series data; wherein the remote sensing time data sequence includes remote sensing data obtained by collecting the surface at multiple time points by a remote sensing satellite; a generation module, used to generate target normalized difference vegetation index NDVI time series data based on multiple remote sensing data in the remote sensing time series data; a prediction module, used to predict the landslide on the surface based on the target deformation time series data and the target NDVI time series data, so as to obtain a landslide prediction result.

[0010] To achieve the above-mentioned purpose, the third aspect embodiment of the present disclosure proposes an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that when the processor executes the program, the landslide identification method described in the first aspect embodiment is implemented.

[0011] In order to achieve the above-mentioned purpose, a fourth aspect of the present disclosure provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the landslide identification method as described in the first aspect of the present disclosure.

[0012] According to a fifth aspect of the embodiments of the present disclosure, a computer program product is provided, including: a computer program, wherein when the computer program is executed by a processor, the landslide identification method as described in the embodiments of the first aspect of the present disclosure is implemented.

[0013] Additional aspects and advantages of the present disclosure will be given in part in the following description and in part will be obvious from the following description or learned through practice of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The above and / or additional aspects and advantages of the present disclosure will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0015] Figure 1 A schematic diagram of a flow chart of a landslide identification method provided in an embodiment of the present disclosure;

[0016] Figure 2 A schematic flow chart of another landslide identification method provided by an embodiment of the present disclosure;

[0017] Figure 3 A schematic flow chart of another landslide identification method provided by an embodiment of the present disclosure;

[0018] Figure 4 A schematic flow chart of another landslide identification method provided by an embodiment of the present disclosure;

[0019] Figure 5 A schematic diagram of the principle of a landslide identification method provided by an embodiment of the present disclosure;

[0020] Figure 6 A schematic diagram of a radar image processing process provided by an embodiment of the present disclosure;

[0021] Figure 7 A schematic diagram of a remote sensing data processing process provided by an embodiment of the present disclosure;

[0022] Figure 8 A schematic diagram of a convolution operation flow provided by an embodiment of the present disclosure;

[0023] Fig. 9 A schematic diagram of an adaptive learning mechanism flow provided by an embodiment of the present disclosure;

[0024] Fig.10 A schematic diagram of the structure of a landslide identification device provided in an embodiment of the present disclosure;

[0025] Fig.11 A schematic diagram of the structure of an electronic device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0026] Embodiments of the present disclosure are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below by the candidate drawings are exemplary and are intended to be used to explain the present disclosure, and should not be construed as limiting the present disclosure.

[0027] At present, landslide monitoring mainly relies on ground monitoring, historical data analysis and remote sensing technology. Although traditional ground monitoring methods can provide real-time data, their monitoring range is limited and they are greatly affected by factors such as terrain and climate, making it difficult to fully reflect the risk of landslides. With the development of aerospace remote sensing technology, high-precision, high-resolution and richer remote sensing data have become a huge amount of data that needs to be mined, especially the data mining technology of earth observation needs to be further improved.

[0028] Monitoring methods based on remote sensing technology, such as synthetic aperture radar interferometry (InSAR), can provide large-scale, high-precision surface deformation information. Small baseline set interferometry (SBAS-INSAR) technology can effectively obtain surface deformation information, but it relies on linear models and statistical methods, and is limited by the satellite's revisit period. It is mainly used for post-analysis and lacks real-time monitoring and early warning capabilities. Optical remote sensing satellites interpret the content of the surface and land cover in more detail, but there is a problem of relatively low image resolution. In related technologies, landslide monitoring is carried out in two fields: optical normalized vegetation index and radar SBAS-INSAR. There is a lack of fusion application of the two, and the potential information in the data cannot be fully explored.

[0029] Therefore, in order to solve the above problems, the present disclosure proposes a landslide identification method and device.

[0030] The landslide identification method and device according to the embodiments of the present disclosure will be described below with reference to the accompanying drawings.

[0031] Figure 1 A schematic flow chart of a landslide identification method provided in an embodiment of the present disclosure.

[0032] like Figure 1 As shown, the landslide identification method may include the following steps:

[0033] Step 101, acquiring a plurality of radar images.

[0034] The radar image is obtained by collecting the surface of the earth through a radar satellite in response to reaching a set period.

[0035] In order to obtain surface change information in a timely manner, in the disclosed embodiment, when a set period is reached, radar images are acquired by radar satellites. For example, radar images are acquired by radar satellites every 12 days. The surface may be a mountainous or hilly area with a sloping landform.

[0036] Step 102: extracting time series data of target deformation on the ground surface from multiple radar images.

[0037] In order to effectively understand the subtle changes in the ground surface, as an example, target deformation time series data of the ground surface is extracted from multiple radar images, wherein the target deformation time series data has high-resolution ground deformation information and can reflect the subtle changes in the ground surface.

[0038] Step 103: Acquire remote sensing time series data.

[0039] The remote sensing time data sequence includes remote sensing data collected by remote sensing satellites at multiple time points on the earth's surface.

[0040] In order to monitor the changes in the earth's surface, in the embodiments of the present disclosure, remote sensing satellites are used to continuously observe the same surface area at multiple time points, remote sensing data of the area at different time points are collected, and remote sensing time series data are generated based on the remote sensing data at different time points. The remote sensing time series data can be used to monitor the changes in the earth's surface. The remote sensing satellites include the Landsat series, Sentinel series, MODIS, ASTER, etc.

[0041] Step 104: Generate target Normalized Difference Vegetation Index NDVI time series data according to the multiple remote sensing data in the remote sensing time series data.

[0042] In order to improve the accuracy of landslide prediction, in the embodiment of the present disclosure, the NDVI value is obtained by performing specific mathematical operations on the red light band and the near-infrared band in the remote sensing data, wherein the range of the NDVI value is usually between -1 and 1, wherein a positive value indicates vegetation coverage, and a larger value indicates a higher vegetation coverage and more luxuriant growth; a negative value indicates that the surface may be covered with non-vegetation elements such as water, ice and snow or bare soil; a zero value usually indicates an area where the surface is rocky or bare soil and has no vegetation coverage.

[0043] Then, based on the NDVI value at each time point, the target NDVI time series data is constructed, wherein the target NDVI time series data can reflect the growth status and change trend of surface vegetation at different time points.

[0044] Step 105: predicting landslides on the ground surface according to the target deformation time series data and the target NDVI time series data to obtain landslide prediction results.

[0045] In order to improve the accuracy of landslide prediction, in the embodiment of the present disclosure, landslide prediction is performed on the surface through fusion analysis of multi-source data combined with target deformation time series data and target NDVI time series data.

[0046] In summary, by extracting target deformation time series data and target normalized difference vegetation index (NDVI) time series data of the surface based on multiple radar images and remote sensing time series data, landslide prediction is carried out on the surface. Therefore, through the fusion analysis of multi-source data, combined with the changes in surface deformation and vegetation index, the dynamic changes of the surface can be more comprehensively reflected, and the accuracy of landslide prediction can be improved. In addition, the periodic collection of radar images and remote sensing data enables the monitoring system to obtain surface change information in real time, timely discover potential landslide risks, and provide valuable time for emergency response, thereby minimizing the losses caused by landslide disasters.

[0047] In order to clearly explain how the landslide prediction is performed on the ground surface according to the target deformation time series data and the target NDVI time series data in the above embodiment to obtain the landslide prediction result, the present disclosure proposes another landslide identification method.

[0048] Figure 2 A schematic flow chart of another landslide identification method provided in an embodiment of the present disclosure.

[0049] like Figure 2 As shown, the landslide identification method may include the following steps:

[0050] Step 201, acquiring a plurality of radar images.

[0051] The radar image is obtained by collecting the surface of the earth through a radar satellite in response to reaching a set period.

[0052] Step 202: extracting time series data of target deformation on the ground surface from multiple radar images.

[0053] Step 203: Acquire remote sensing time series data.

[0054] The remote sensing time data sequence includes remote sensing data collected by remote sensing satellites at multiple time points on the earth's surface.

[0055] Step 204: Generate target Normalized Difference Vegetation Index NDVI time series data according to the multiple remote sensing data in the remote sensing time series data.

[0056] Step 205 , encoding the target deformation time series data and the target NDVI time series data in the same geographic coordinate system to obtain deformation time series data and NDVI time series data carrying the same geographic code.

[0057] In order to make the deformation information and vegetation index information accurately correspond in space, in the embodiment of the present disclosure, the target deformation time series data and the target NDVI time series data are integrated into a unified geographic coordinate system to ensure that each set of deformation data and NDVI data is accompanied by the same geographic location identifier, i.e., geocoding.

[0058] Step 206 , feature extraction is performed on the deformation time series data and the NDVI time series data carrying the same geographic code to obtain a first feature vector and a second feature vector.

[0059] In the disclosed embodiment, feature extraction operations are performed on deformation time series data and NDVI time series data carrying the same geographic code, respectively, wherein the deformation time series data is processed to extract key deformation features to form a first feature vector; the NDVI time series data is processed through a corresponding feature extraction method to obtain features such as vegetation growth conditions and change trends to form a second feature vector. The first feature vector and the second feature vector respectively represent the deformation features and vegetation coverage features of the same position in the geographic space.

[0060] Step 207: fuse the first feature vector and the second feature vector to obtain a first fused feature vector.

[0061] In order to improve the comprehensiveness and accuracy of landslide prediction, as a possible implementation method, the first eigenvector and the second eigenvector are fused to generate a first fused eigenvector.

[0062] As an example, the first feature vector and the second feature vector are concatenated to obtain a first fused feature vector.

[0063] As another example, by setting weights, a weighted sum is performed on the first eigenvector and the second eigenvector to obtain a first fused eigenvector.

[0064] Step 208: Based on the first fused feature vector, landslide prediction is performed on the ground surface to obtain a landslide prediction result.

[0065] In order to improve the accuracy of landslide prediction, as an example, a relevant prediction algorithm is used to perform landslide prediction on the surface based on the first fused feature vector to obtain a landslide prediction result.

[0066] As another example, the first fused feature vector is input into a trained long short-term memory (LSTM) model to obtain a first landslide prediction probability of the surface output by the LSTM model; based on the first landslide prediction probability, a landslide prediction result is generated to indicate whether a landslide occurs.

[0067] That is to say, after the LSTM model receives the first fused feature vector, the LSTM model processes and analyzes the feature vector, and finally outputs a prediction value, which represents the first prediction probability of a landslide on the surface; wherein, the first prediction probability is a value between 0 and 1, indicating the possibility of a landslide on the surface in a certain time period in the future according to the currently input feature vector; then, based on this prediction probability, a landslide prediction result is generated. The landslide prediction result includes "yes" or "no", which is used to indicate whether the landslide is predicted. For example, if the first prediction probability is greater than 0.5, the landslide prediction result is "yes", indicating that a landslide is predicted to occur in a certain time period in the future; for another example, if the first prediction probability is less than 0.5, the landslide prediction result is "no", indicating that a landslide is not predicted to occur in a certain time period in the future.

[0068] It should be noted that the LSTM model is trained using the following steps:

[0069] 1. Obtain an initial LSTM model and obtain sample data; the sample data includes sample deformation time series data and sample NDVI time series data of the sample surface;

[0070] 2. Extract the third eigenvector from the sample deformation time series data, and extract the fourth eigenvector from the sample NDVI time series data;

[0071] 3. Fusing the third eigenvector and the fourth eigenvector to obtain a second fused eigenvector;

[0072] 4. Input the second fused feature vector into the LSTM model to obtain the second landslide prediction probability of the sample surface;

[0073] 5. Train the LSTM model based on the second landslide prediction probability and the landslide annotation probability of the sample surface.

[0074] In the disclosed embodiment, a loss function is generated according to the difference between the second landslide prediction probability and the landslide annotation probability of the sample surface, and the coefficients of the LSTM model are adjusted according to the value of the loss function to minimize the value of the loss function.

[0075] It should be noted that the above only uses the termination condition of LSTM model training as the minimization of the value of the loss function as an example. In actual application, other termination conditions can also be set. For example, the termination condition can also be that the number of training times reaches a set threshold, the training time is greater than a set time threshold, etc. The present disclosure does not limit this.

[0076] In addition, it should be noted that in order to reduce the model training efficiency, in the embodiment of the present disclosure, after acquiring new sample data, the LSTM model can be incrementally learned to update some parameters of the model, thereby improving the training efficiency.

[0077] It should be noted that the execution process of steps 201 to 204 can be implemented in any manner in the embodiments of the present disclosure, and the embodiments of the present disclosure do not limit this and will not be described in detail.

[0078] In summary, the target deformation time series data and the target NDVI time series data are encoded in the same geographic coordinate system to obtain the deformation time series data and the NDVI time series data carrying the same geographic code; the deformation time series data and the NDVI time series data carrying the same geographic code are respectively extracted to obtain the first feature vector and the second feature vector; the first feature vector and the second feature vector are fused to obtain the first fused feature vector; based on the first fused feature vector, the landslide is predicted on the surface to obtain the landslide prediction result. Thus, the target deformation time series data and the target NDVI time series data are encoded in a unified geographic coordinate system and the features are respectively extracted, and then fused into the first fused feature vector. The first feature vector contains the comprehensive information of surface deformation and vegetation change, which can more accurately reflect the dynamic changes of the surface and improve the accuracy of landslide prediction.

[0079] In order to clearly illustrate how to extract the deformation time series data of the ground surface from multiple radar images in the above embodiment, the present disclosure proposes another landslide identification method.

[0080] Figure 3 A schematic flow chart of another landslide identification method provided in an embodiment of the present disclosure.

[0081] like Figure 3 As shown, the landslide identification method may include the following steps:

[0082] Step 301, acquiring multiple radar images.

[0083] The radar image is obtained by collecting the surface of the earth through a radar satellite in response to reaching a set period.

[0084] Step 302 , obtaining the temporal baseline and the spatial baseline of each radar image, and grouping the multiple radar images according to the temporal baseline and the spatial baseline of each radar image to obtain multiple small baseline sets.

[0085] In an embodiment of the present disclosure, the temporal baseline and spatial baseline of each pair of images are calculated based on the temporal information of each radar image and the satellite orbit parameters to generate a temporal-spatial baseline map; according to the temporal-spatial baseline map, a reference image is determined from each interferometric image pair; wherein the reference image is an image that satisfies the following conditions: the shortest temporal baseline and spatial baseline, the largest number of connections, and the image is located in the middle of the temporal-spatial baseline; the reference image is used as the main image for registration, and multiple radar images are grouped based on the temporal baseline threshold and the spatial baseline threshold set by the baseline length.

[0086] Step 303 , according to any small baseline set, performing differential interference processing on any interference image pair in any small baseline set to obtain a second interference pattern.

[0087] In order to improve the accuracy of surface deformation, in the embodiment of the present disclosure, for any interferometric image pair in any small baseline set, the phase difference of the any interferometric image pair is calculated to generate a corresponding interferogram; and deflating, adaptive filtering, coherence generation, phase unwrapping and other processing are performed to generate a series of unwrapped phase maps; the coherence maps and unwrapping result maps of each generated image pair are checked one by one, and the image pairs with low coherence are removed to obtain a second interferogram.

[0088] Step 304: Determine the deformation time series data of the ground surface according to the second interference pattern.

[0089] In order to further improve the accuracy of surface deformation information, as a possible implementation method, the deformation sequence data was obtained by performing orbit refinement, flattening, residual phase removal, deformation rate and residual terrain estimation, and atmospheric residual phase removal on the second interferogram, and converting the elevation information of the final interferogram into the geographic coordinate system.

[0090] In the disclosed embodiment, the second interference map is sequentially subjected to orbit refinement and flattening processing to obtain a third interference map; the residual phase is identified from the third interference map, and the residual phase is eliminated from the third interference map to obtain a fourth interference map; the deformation rate and residual terrain are estimated on the fourth interference map to obtain a first estimation result; based on the first estimation result, the atmospheric residual phase is identified from the fourth interference map, and the atmospheric residual phase is eliminated from the fourth interference map to obtain a fifth interference map; according to the conversion relationship between the radar coordinate system and the geographic coordinate system, the elevation information in the fifth interference map is converted to the geographic coordinate system to obtain target deformation sequence data.

[0091] That is to say, based on the coherence and phase information of each control point (interference point) in the second interference map, the target control point is determined from each control point. For example, a point with high coherence and good phase is selected as the target control point, that is, a control point that is closer to the time-space baseline (temporal baseline and spatial baseline) and relatively stable and has a strong reference is selected as the target control point; then, based on the target control point, the second interference map is successively subjected to orbit refinement and flattening processing to obtain a third interference map, and the residual phase is removed from the third interference map to obtain a fourth interference map; wherein the residual phase may come from various error sources, such as atmospheric interference, surface coverage, etc. cover changes, etc.; then, the deformation rate and residual terrain of the fourth interference map are estimated to obtain a first estimation result (including the deformation rate and residual terrain); on the basis of the first estimation result, atmospheric filtering is performed to estimate and remove the atmospheric phase (the phase error caused by atmospheric interference) to reduce the influence of the atmospheric residual phase on the extraction of deformation information; finally, according to the conversion relationship between the radar coordinate system and the geographic coordinate system, the elevation information in the fifth interference map is converted to the geographic coordinate system to obtain the target deformation sequence data; wherein, the deformation time series data records the deformation of the surface at different time points.

[0092] Step 305: Acquire remote sensing time series data.

[0093] The remote sensing time data sequence includes remote sensing data collected by remote sensing satellites at multiple time points on the earth's surface.

[0094] Step 306: Generate target Normalized Difference Vegetation Index NDVI time series data according to the multiple remote sensing data in the remote sensing time series data.

[0095] Step 307: predicting landslides on the ground surface according to the target deformation time series data and the target NDVI time series data to obtain landslide prediction results.

[0096] It should be noted that the execution process of step 301 and steps 305 to 307 can be implemented in any manner in the embodiments of the present disclosure, and the embodiments of the present disclosure do not limit this and will not be described in detail.

[0097] In summary, by obtaining the temporal baseline and spatial baseline of each radar image, and grouping multiple radar images according to the temporal baseline and spatial baseline of each radar image, a plurality of small baseline sets are obtained; according to any small baseline set, any interference image pair in any small baseline set is subjected to differential interference processing to obtain a second interference map; according to the second interference map, the deformation time series data of the surface is determined. Thus, by accurately obtaining and considering the temporal baseline and spatial baseline of the radar image, the radar images can be more effectively grouped to form a plurality of small baseline sets, which helps to reduce the phase error and decoherence problems in the interference processing; secondly, differential interference processing is performed on the interference image pairs in each small baseline set to generate a second interference map with higher quality; finally, based on the second interference map, more accurate and reliable surface deformation time series data can be determined, so that landslide prediction can be performed based on accurate deformation time series data, thereby improving the accuracy of landslide prediction.

[0098] In order to clearly explain how to generate target Normalized Difference Vegetation Index NDVI time series data according to multiple remote sensing data in the remote sensing time series data in the above embodiment, the present disclosure proposes another landslide identification method.

[0099] Figure 4 A schematic flow chart of another landslide identification method provided in an embodiment of the present disclosure.

[0100] like Figure 4 As shown, the landslide identification method may include the following steps:

[0101] Step 401, acquiring multiple radar images.

[0102] The radar image is obtained by collecting the surface of the earth through a radar satellite in response to reaching a set period.

[0103] Step 402: extracting time series data of target deformation on the ground surface from multiple radar images.

[0104] Step 403: Acquire remote sensing time series data.

[0105] The remote sensing time data sequence includes remote sensing data collected by remote sensing satellites at multiple time points on the earth's surface.

[0106] Step 404 , extracting NDVI data in a first format from a plurality of remote sensing data, and performing format conversion on the NDVI data in the first format to obtain NDVI data in a second format.

[0107] In the embodiment of the present disclosure, MRT software (Medium Resolution Imaging Spectrometer) is used to extract and format the NDVI data of the remote sensing data to obtain NDVI data in the second format.

[0108] Step 405 , according to the conversion relationship between the sinusoidal projection coordinate system and the UTM coordinate system, the NDVI data in the second format is converted to the UTM coordinate system to obtain projection data in the UTM coordinate system.

[0109] Then, the projection of the NDVI data in the second format is converted from the sinusoidal projection coordinate system to the UTM coordinate system (eg, UTM-WGS84, a global geographic coordinate system) to obtain projection data in the UTM coordinate system.

[0110] Step 406: convert the format of the projection data to obtain converted data.

[0111] In order to improve the versatility of NDVI time series data, in the embodiment of the present disclosure, the projection data is converted into conversion data in a standard geographic spatial data format (eg, Geo-tif).

[0112] Step 407, generating target NDVI time series data based on the converted data.

[0113] In order to improve the quality of NDVI time series data, in the embodiment of the present disclosure, pixel outliers in the converted data are eliminated to obtain converted data after the outliers are removed; the converted data after the outliers are removed are seamlessly mosaicked and cultivated land mask extracted to obtain initial NDVI time series data; the initial NDVI time series data are smoothed to obtain target NDVI time series data.

[0114] That is to say, pixel outliers will interfere with the analysis and interpretation of the data. The converted data is input into ENVI5.3 to remove pixel outliers. For example, pixel values ​​less than 0 are assigned 0, and pixel values ​​greater than 0 are multiplied by 0.0001. Then, the converted data after removing outliers are seamlessly mosaicked and cultivated land mask extracted to obtain the initial NDVI time series data. Due to the influence of various factors (such as weather changes, sensor noise, etc.), the initial NDVI time series data may fluctuate or noisy. Smoothing the initial NDVI time series data can reduce these fluctuations and make the data smoother and more stable.

[0115] It should be noted that, in the embodiment of the present disclosure, the Savitzky-Golay filtering method is used to smooth the initial NDVI time series data. This filtering method can simulate the time series data of the vegetation index within a certain growth period, obtain its long-term change trend, highlight the local abnormal noise information, and reduce the random interference of the data time, spatial scale and sensor.

[0116] Step 408: predicting landslides on the ground surface based on the target deformation time series data and the target NDVI time series data to obtain landslide prediction results.

[0117] In summary, NDVI data in a first format are extracted from multiple remote sensing data, and the format of the NDVI data in the first format is converted to obtain NDVI data in a second format; according to the conversion relationship between the sinusoidal projection coordinate system and the UTM coordinate system, the NDVI data in the second format are converted to the UTM coordinate system to obtain projection data in the UTM coordinate system; the projection data is format converted to obtain conversion data; and target NDVI time series data is generated according to the conversion data. Therefore, by extracting and converting NDVI data from multiple remote sensing data to a unified UTM coordinate system, data from different sources can be standardized, projection differences can be eliminated, and the comparability and interoperability of data can be improved, thereby generating accurate and reliable target NDVI time series data.

[0118] Based on any embodiment of the present disclosure, the implementation process of the present disclosure can be as follows: Figure 5 As shown, it mainly includes the following steps:

[0119] 1. Radar image processing: collect radar images, generate connection maps, differential interferometry processing, track refinement and deflating, SBAS inversion, geocoding, etc. Obtain surface deformation data through SBAS-InSAR technology, geocode the data and construct surface deformation time series data; Figure 6 As shown, the following steps are included:

[0120] (1) Collect radar images: Use synthetic aperture radar (SAR) satellites (such as Sentinel1) to obtain SBAS-INSAR images. The radar images contain high-resolution surface deformation information and can reflect small changes in the surface. According to the needs of landslide monitoring, the time interval for image acquisition is set (such as every 12 days) to ensure timely capture of dynamic changes in landslides;

[0121] (2) Generate a connection map (space-time baseline map): Calculate the time-series baseline and spatial baseline of the interferometric image pair based on the time-series information and the satellite orbit parameters of the image to generate a space-time baseline map; Based on the space-time baseline map, select the image with the shortest space-time baseline, the largest number of connections, and the image located in the middle of the space-time baseline from each interferometric image pair as the reference image; Group the SAR images based on the threshold set by the baseline length to form multiple small baseline sets;

[0122] (3) Differential interferometry processing: Calculate the phase difference of each interferometric image pair to generate the corresponding interferogram; then perform deflating, adaptive filtering, coherence generation, phase unwrapping and other processing to generate a series of unwrapped phase maps; check the coherence maps and unwrapping result maps of each image pair generated one by one, and remove image pairs with low coherence;

[0123] (4) Track refinement and re-flattening: Select control points on the filtered interferogram, and select points with high coherence and good phase as control points; combine with DEM, perform track refinement and re-flattening operations, and remove residual phase;

[0124] (5) SBAS conversion: The deformation rate and residual terrain are first estimated to flatten the synthesized interferogram; phase unwrapping and refinement are performed again to generate more optimized results; atmospheric filtering is performed based on the deformation rate estimated in the first time, and the atmospheric phase is estimated and removed to obtain more accurate deformation results in the time series;

[0125] (6) Geocoding: Since the elevation information obtained in the previous step is along the radar line of sight, it is necessary to convert the elevation information from the radar coordinate system to the geographic coordinate system; finally, the average deformation rate data, deformation information at each time point, and three-dimensional position information are obtained;

[0126] 2. Remote sensing data processing: unify geographic coordinates through projection and format conversion, extract NDVI values ​​and reconstruct time series to construct target NDVI time series data;

[0127] (1) MODIS data preprocessing: The 250m resolution NDVI and its quality control data synthesized based on the MVC method for 16 days are geometrically corrected and atmospherically corrected, and then converted into projection and format. Figure 7 As shown in the figure, MRT software is used to extract and convert the NDVI data of MODIS data (remote sensing data), and the projection is converted from the original sinusoidal projection to UTM-WGS84, and then converted to Geo-tif data format. Then the converted data is input into ENVI5.3 to remove pixel outliers, assign 0 to pixel values ​​less than 0, and multiply pixel values ​​greater than 0 by 0.0001, and perform seamless mosaicking and farmland mask extraction on the data after removing outliers to form NDVI time series data;

[0128] (2) MODIS NDVI time series reconstruction: The selected data is the 16-day maximum synthetic data. Although the influence of clouds is reduced to a certain extent, there are still random interference factors such as water vapor, aerosol and sensor accuracy changes, which make the NDVI time series data irregular and unable to accurately reflect the landmark information. Therefore, it is necessary to reconstruct the NDVI time series to reduce the noise level of the time series data; TIMESAT software can visualize and selectively reconstruct the vegetation index time series; the Savitzky-Golay filtering method is selected, which can simulate the time series data of the vegetation index within a certain growth period, obtain its long-term change trend, highlight the local abnormal noise information, and reduce the random interference of the data time, spatial scale and sensor;

[0129] 3. Feature vector calculation: Input the two types of data into the CNN layer respectively, and calculate the feature vector and geographic code of the deformation data and the feature vector and geographic code of the NDVI data through the feature calculation of the first convolution layer respectively;

[0130] It should be understood that since the surface deformation data monitored by the SBAS-InSAR technology is a typical time series data, and the NDVI data is generated by the Savitzky-Golay filtering method, the proposed convolutional neural network (CNN) model is used to effectively extract the spatial features in the landslide time series data. In the embodiment of the present disclosure, the geographic coordinate system of the two data is unified for unified geographic coding to achieve a one-to-one correspondence between the deformation data and the NDVI data, which is convenient for subsequent feature extraction and prediction, such as Figure 8 As shown, the following modules are used for implementation:

[0131] (1) The first convolution module is used to input the temporal deformation data with geocoding into the convolution layer, and output the feature vector of the deformation data and the corresponding geocoding through feature operation of the temporal deformation data;

[0132] (2) The second convolution module is used to input the NDVI time series data with geocoding into the convolution layer, and output the feature vector of the NDVI data and the corresponding geocoding through feature operation of the NDVI time series data;

[0133] (3) Sampling module: used to fuse the feature vector and corresponding geographic code of the deformation data formed by the first convolution operation and the feature vector and corresponding geographic code of the NDVI data formed by the second convolution operation, and form a fused feature vector of the deformation data and the NDVI data through a unified geographic code.

[0134] 4. Feature fusion: According to the unified geographic coding, the feature vectors of the two data are fused to form a fused feature vector;

[0135] 5. Time series modeling: The landslide fusion feature vector obtained by convolution operation and fusion is input into the LSTM prediction module. LSTM automatically predicts the landslide results based on the time series features in the feature vector.

[0136] 6. Adaptive learning: e.g. Fig. 9 As shown in the figure, set indicators such as learning rate, dynamically adjust the parameters of the deep learning model through incremental learning method, and optimize the model performance.

[0137] (1) Regular data update: The system regularly receives new remote sensing data, calculates SBAS-INSAR deformation data and NDVI time series data, and dynamically updates the model input to ensure the timeliness of the prediction results;

[0138] (2) Adaptive learning mechanism: Based on the real-time monitoring results and the actual landslide occurrence, an online learning algorithm with incremental learning is used to dynamically adjust the parameters of the deep learning model and optimize the model performance.

[0139] In order to implement the above embodiments, the present disclosure proposes a landslide identification device.

[0140] Fig.10 A schematic diagram of the structure of a landslide identification device provided in an embodiment of the present disclosure.

[0141] like Fig.10 As shown, the landslide identification device 1000 includes: a first acquisition module 1010 , a first extraction module 1020 , a second acquisition module 1030 , a generation module 1040 and a prediction module 1050 .

[0142] Among them, the first acquisition module 1010 is used to acquire multiple radar images; wherein the radar images are obtained by collecting the surface through radar satellites in response to reaching a set period; the first extraction module 1020 is used to extract target deformation time series data of the surface from multiple radar images; the second acquisition module 1030 is used to acquire remote sensing time series data; wherein the remote sensing time data sequence includes remote sensing data obtained by collecting the surface through remote sensing satellites at multiple time points; the generation module 1040 is used to generate target normalized difference vegetation index NDVI time series data based on multiple remote sensing data in the remote sensing time series data; the prediction module 1050 is used to predict landslides on the surface based on the target deformation time series data and the target NDVI time series data to obtain landslide prediction results.

[0143] As a possible implementation method of the embodiment of the present disclosure, the prediction module 1050 is used to encode the target deformation time series data and the target NDVI time series data in the same geographic coordinate system to obtain the deformation time series data and the NDVI time series data carrying the same geographic code; perform feature extraction on the deformation time series data and the NDVI time series data carrying the same geographic code respectively to obtain a first feature vector and a second feature vector; fuse the first feature vector and the second feature vector to obtain a first fused feature vector; and perform landslide prediction on the surface based on the first fused feature vector to obtain a landslide prediction result.

[0144] As a possible implementation method of the embodiment of the present disclosure, the prediction module 1050 is used to input the first fused feature vector into a trained long short-term memory network LSTM model to obtain a first landslide prediction probability of the surface output by the LSTM model; based on the first landslide prediction probability, generate a landslide prediction result indicating whether a landslide occurs.

[0145] As a possible implementation method of the embodiment of the present disclosure, the LSTM model is trained using the following modules: a third acquisition module, a second extraction module, a fusion module, an input module and a training module.

[0146] Among them, the third acquisition module is used to obtain the initial LSTM model and obtain sample data; wherein the sample data includes sample deformation time series data and sample NDVI time series data of the sample surface; the second extraction module is used to extract the third eigenvector from the sample deformation time series data, and extract the fourth eigenvector from the sample NDVI time series data; the fusion module is used to fuse the third eigenvector and the fourth eigenvector to obtain a second fused eigenvector; the input module is used to input the second fused eigenvector into the LSTM model to obtain a second landslide prediction probability of the sample surface; the training module is used to train the LSTM model according to the second landslide prediction probability and the landslide annotation probability of the sample surface.

[0147] As a possible implementation method of the embodiment of the present disclosure, the first extraction module 1020 is used to obtain the time baseline and space baseline of each radar image, and group multiple radar images according to the time baseline and space baseline of each radar image to obtain multiple small baseline sets; according to any small baseline set, any interference image pair in any small baseline set is differentially interfered to obtain a second interference map; according to the second interference map, the target deformation time series data of the surface is determined.

[0148] As a possible implementation method of the embodiment of the present disclosure, the first extraction module 1020 is also used to perform orbit refinement and flattening processing on the second interference map in sequence to obtain a third interference map; identify the residual phase from the third interference map, and eliminate the residual phase from the third interference map to obtain a fourth interference map; perform deformation rate and residual terrain estimation on the fourth interference map to obtain a first estimation result; based on the first estimation result, identify the atmospheric residual phase from the fourth interference map, and eliminate the atmospheric residual phase from the fourth interference map to obtain a fifth interference map; according to the conversion relationship between the radar coordinate system and the geographic coordinate system, convert the elevation information in the fifth interference map to the geographic coordinate system to obtain deformation sequence data.

[0149] As a possible implementation method of the embodiment of the present disclosure, the first extraction module 1020 is also used to determine the target interference point from each interference point based on the coherence and phase information of each control point in the second interference map; according to the target interference point, the second interference map is successively subjected to orbit refinement and flattening processing to obtain a third interference map.

[0150] As a possible implementation method of the embodiment of the present disclosure, a generation module 1040 is used to extract NDVI data in a first format from multiple remote sensing data, and perform format conversion on the NDVI data in the first format to obtain NDVI data in a second format; according to the conversion relationship between the sinusoidal projection coordinate system and the UTM coordinate system, the NDVI data in the second format is converted to the UTM coordinate system to obtain projection data in the UTM coordinate system; the projection data is format converted to obtain conversion data; and target NDVI time series data is generated based on the conversion data.

[0151] As a possible implementation method of the embodiment of the present disclosure, the generation module 1040 is also used to eliminate pixel outliers in the conversion data to obtain the conversion data after the outliers are removed; seamlessly mosaic and farmland mask extraction are performed on the conversion data after the outliers are removed to obtain initial NDVI time series data; and the initial NDVI time series data is smoothed to obtain target NDVI time series data.

[0152] The landslide identification device of the disclosed embodiment extracts target deformation time series data and target normalized difference vegetation index (NDVI) time series data of the surface based on multiple radar images and remote sensing time series data, and predicts landslides on the surface. Therefore, through the fusion analysis of multi-source data and the combination of changes in surface deformation and vegetation index, it can more comprehensively reflect the dynamic changes of the surface and improve the accuracy of landslide prediction. In addition, the periodic collection of radar images and remote sensing data enables the monitoring system to obtain surface change information in real time, timely discover potential landslide risks, and provide valuable time for emergency response, thereby minimizing the losses caused by landslide disasters.

[0153] In order to implement the above embodiment, the present disclosure also provides an electronic device, Fig.11 A schematic diagram of the structure of an electronic device provided in an embodiment of the present disclosure. The electronic device includes:

[0154] A memory 1101 , a processor 1102 , and a computer program stored in the memory 1101 and executable on the processor 1102 .

[0155] When the processor 1102 executes the program, the landslide identification method provided in the above embodiment is implemented.

[0156] Furthermore, the electronic device further comprises:

[0157] The communication interface 1103 is used for communication between the memory 1101 and the processor 1102 .

[0158] The memory 1101 is used to store computer programs that can be executed on the processor 1102 .

[0159] The memory 1101 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0160] The processor 1102 is configured to implement the landslide identification method described in the above embodiment when executing the program.

[0161] If the memory 1101, the processor 1102 and the communication interface 1103 are implemented independently, the communication interface 1103, the memory 1101 and the processor 1102 can be connected to each other through a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Fig.11 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0162] Optionally, in a specific implementation, if the memory 1101, the processor 1102 and the communication interface 1103 are integrated on a chip, the memory 1101, the processor 1102 and the communication interface 1103 can communicate with each other through an internal interface.

[0163] The processor 1102 may be a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present disclosure.

[0164] In order to implement the above embodiment, the embodiment of the present disclosure further proposes a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the landslide identification method as in the above embodiment is implemented.

[0165] In the description of this specification, the description of the candidate terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present disclosure. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0166] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of the present disclosure, "plurality" means at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0167] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present disclosure includes additional implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present disclosure belong.

[0168] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute the instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purpose of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways if necessary, and then stored in a computer memory.

[0169] It should be understood that the various parts of the present disclosure can be implemented in hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented in software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used to implement: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0170] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0171] In addition, each functional unit in each embodiment of the present disclosure may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0172] The storage medium mentioned above may be a read-only memory, a disk or an optical disk, etc. Although the embodiments of the present disclosure have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations of the present disclosure. A person of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present disclosure.

Claims

1. A landslide identification method, characterized in that: include: Acquire multiple radar images; wherein the radar images are acquired by collecting the surface of the earth through a radar satellite in response to reaching a set period; Extracting target deformation time series data of the ground surface from the plurality of radar images; Acquiring remote sensing time series data; wherein the remote sensing time series data includes remote sensing data acquired by remote sensing satellites at multiple time points on the surface of the earth; Generate target normalized difference vegetation index NDVI time series data according to multiple remote sensing data in the remote sensing time series data; According to the target deformation time series data and the target NDVI time series data, landslide prediction is performed on the surface to obtain a landslide prediction result.

2. The method according to claim 1, characterized in that The step of performing landslide prediction on the surface according to the target deformation time series data and the target NDVI time series data to obtain a landslide prediction result includes: Encoding the target deformation time series data and the target NDVI time series data in the same geographic coordinate system to obtain deformation time series data and NDVI time series data carrying the same geographic code; Performing feature extraction on the deformation time series data and the NDVI time series data carrying the same geographic code respectively to obtain a first feature vector and a second feature vector; Fusing the first feature vector and the second feature vector to obtain a first fused feature vector; Based on the first fused feature vector, a landslide prediction is performed on the ground surface to obtain a landslide prediction result.

3. The method according to claim 2, characterized in that The performing landslide prediction on the ground surface based on the first fused feature vector to obtain a landslide prediction result includes: Inputting the first fused feature vector into a trained long short-term memory network (LSTM) model to obtain a first landslide prediction probability of the surface output by the LSTM model; A landslide prediction result indicating whether a landslide occurs is generated according to the first landslide prediction probability.

4. The method according to claim 3, characterized in that The LSTM model is trained using the following steps: Obtaining an initial LSTM model and obtaining sample data; wherein the sample data includes sample deformation time series data and sample NDVI time series data of the sample surface; Extracting a third eigenvector from the sample deformation time series data, and extracting a fourth eigenvector from the sample NDVI time series data; Fusing the third eigenvector and the fourth eigenvector to obtain a second fused eigenvector; Inputting the second fused feature vector into the LSTM model to obtain a second landslide prediction probability of the sample surface; The LSTM model is trained according to the second landslide prediction probability and the landslide annotation probability of the sample surface.

5. The method according to claim 1, characterized in that The step of extracting the target deformation time series data of the ground surface from the plurality of radar images comprises: Acquire a temporal baseline and a spatial baseline of each radar image, and group the multiple radar images according to the temporal baseline and the spatial baseline of each radar image to obtain multiple small baseline sets; According to any small baseline set, performing differential interference processing on any interference image pair in any small baseline set to obtain a second interference image; According to the second interference pattern, target deformation time series data of the ground surface is determined.

6. The method according to claim 5, characterized in that Determining the target deformation time series data of the surface according to the second interference pattern includes: performing orbital refining and flattening processing on the second interference graph in sequence to obtain a third interference graph; Identifying a residual phase from the third interference pattern, and removing the residual phase from the third interference pattern to obtain a fourth interference pattern; performing deformation rate and residual terrain estimation on the fourth interference pattern to obtain a first estimation result; Based on the first estimation result, identifying the atmospheric residual phase from the fourth interferogram, and removing the atmospheric residual phase from the fourth interferogram to obtain a fifth interferogram; According to the conversion relationship between the radar coordinate system and the geographic coordinate system, the elevation information in the fifth interference map is converted to the geographic coordinate system to obtain the target deformation sequence data.

7. The method according to claim 6, characterized in that The step of sequentially performing orbital refining and flattening processing on the second interference graph to obtain a third interference graph includes: Determine a target interference point from the interference points according to the coherence and phase information of each control point in the second interference graph; According to the target interference point, the second interference graph is sequentially subjected to orbit refinement and flattening processing to obtain a third interference graph.

8. The method according to claim 1, characterized in that The step of generating target normalized difference vegetation index NDVI time series data according to a plurality of remote sensing data in the remote sensing time series data comprises: Extracting NDVI data in a first format from the plurality of remote sensing data, and performing format conversion on the NDVI data in the first format to obtain NDVI data in a second format; According to the conversion relationship between the sinusoidal projection coordinate system and the UTM coordinate system, the NDVI data in the second format is converted to the UTM coordinate system to obtain projection data in the UTM coordinate system; Performing format conversion on the projection data to obtain conversion data; According to the converted data, target NDVI time series data is generated.

9. The method according to claim 8, characterized in that Generating target NDVI time series data according to the converted data includes: Eliminating abnormal pixel values ​​in the converted data to obtain converted data after the abnormal values ​​are removed; Performing seamless mosaicking and farmland mask extraction on the converted data after outliers are removed to obtain initial NDVI time series data; The initial NDVI time series data is smoothed to obtain target NDVI time series data.

10. A landslide identification device, characterized in that: include: A first acquisition module is used to acquire a plurality of radar images; wherein the radar images are acquired by collecting the surface of the earth through a radar satellite in response to reaching a set period; A first extraction module is used to extract the target deformation time series data of the surface from the multiple radar images; A second acquisition module is used to acquire remote sensing time series data; wherein the remote sensing time series data includes remote sensing data acquired by remote sensing satellites at multiple time points on the surface of the earth; A generating module, used for generating target normalized difference vegetation index NDVI time series data according to a plurality of remote sensing data in the remote sensing time series data; The prediction module is used to perform landslide prediction on the surface according to the target deformation time series data and the target NDVI time series data to obtain a landslide prediction result.

11. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the landslide identification method according to any one of claims 1 to 9 is implemented.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the landslide identification method as described in any one of claims 1 to 9 is implemented.

Images