A landslide monitoring method and monitoring system based on multi-source InSAR data fusion and deep learning in a densely vegetated area

Abstract

The application provides a kind of based on multi-source InSAR data fusion and deep learning's vegetation lush area landslide monitoring method and monitoring system, method includes: data acquisition step, obtains the multi-source InSAR data of target area;Deformation interpretation step, using preset MT-InSAR method is carried out deformation interpretation processing to the target area of multi-source InSAR data, obtains deformation rate data;Regional deformation feature identification step, using preset multi-source InSAR fusion optimization monitoring strategy is carried out fusion analysis processing to deformation rate data, obtains the regional deformation detection data of first period, and regional deformation detection data is carried out regional boundary sketching processing, obtains regional deformation feature data;Regional morphological feature identification step, to be identified optical remote sensing image is input to the landslide classifier model based on deep learning construction, to output landslide morphological identification data;Landslide monitoring grading step, according to regional deformation feature data and landslide morphological identification data, obtains landslide monitoring data.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of landslide monitoring, and in particular to a landslide monitoring method and system based on multi-source InSAR data fusion and deep learning in a densely vegetated area. BACKGROUND

[0002] Landslides are prone to occur in high mountain and canyon areas. In particular, in densely vegetated areas, landslides are difficult to monitor in real time due to natural characteristics such as cloudy and foggy weather, deep valleys, and dense vegetation. Traditional landslide monitoring methods are limited by cloud cover in optical remote sensing and decorrelation in Interferometric Synthetic Aperture Radar (InSAR) technology, resulting in poor landslide monitoring and identification results. In traditional monitoring methods, a single InSAR data identification method is usually used. On the one hand, due to the short wavelength of single InSAR data (such as C-band), it is difficult to obtain continuous deformation information in high-vegetation-covered areas. On the other hand, for the obtained identification information, traditional manual interpretation methods are low in efficiency and strong in subjectivity, and cannot meet the needs of rapid identification and risk classification of large-scale landslide hazards.

[0003] With the development of monitoring technology, there are monitoring methods combining multi-source InSAR data in the prior art, but there is a lack of zoning monitoring strategies for vegetation coverage differences, and the monitoring accuracy is insufficient. In order to effectively monitor landslides in densely vegetated areas, a landslide monitoring method that can target vegetation coverage differences is needed to improve landslide monitoring accuracy. SUMMARY

[0004] The purpose of the present application is to provide a landslide monitoring method that can target vegetation coverage differences to improve landslide monitoring accuracy, and to provide intelligent landslide risk classification.

[0005] To achieve the above-mentioned purpose, the first aspect of the present application provides a landslide monitoring method based on multi-source InSAR data fusion and deep learning in a densely vegetated area, which comprises:

[0006] a data acquisition step for acquiring multi-source InSAR data of a target area; wherein the multi-source InSAR data includes Sentinel-1A data of a first time period, ALOS-2 PALSAR-2 ScanSAR data of the first time period, and LiCSAR data of the first time period;

[0007] The deformation interpretation step uses a preset MT-InSAR method to perform deformation interpretation processing on a target region of the multi-source InSAR data to obtain deformation rate data; wherein the deformation rate data includes first deformation rate data, second deformation rate data and third deformation rate data corresponding to Sentinel-1A data, ALOS-2 PALSAR-2 ScanSAR data and LiCSAR data respectively;

[0008] The regional deformation feature recognition step uses a preset multi-source InSAR fusion optimization monitoring strategy to perform fusion analysis processing on the first deformation rate data, the second deformation rate data and the third deformation rate data, to obtain regional deformation detection data of the first time period, and performs regional boundary sketching processing on the regional deformation detection data to obtain regional deformation feature data.

[0009] The regional morphological feature recognition step inputs the optical remote sensing image to be recognized into a landslide classifier model constructed based on deep learning to output landslide morphological recognition data.

[0010] The landslide monitoring grading step obtains landslide monitoring data according to the regional deformation feature data and the landslide morphological recognition data; wherein the landslide monitoring data includes high-risk landslide point data and medium-risk landslide point data.

[0011] Preferably, the deformation interpretation step specifically includes:

[0012] The GCP point data in the Sentinel-1A data and the ALOS-2 PALSAR-2 ScanSAR data are extracted using a preset PS-InSAR method;

[0013] The GCP point data is used as input of a preset SBAS-InSAR method to perform phase correction processing to obtain the first deformation rate data and the second deformation rate data;

[0014] The LiCSAR data is processed using a preset LICSBAS method to obtain third deformation rate data.

[0015] Preferably, the preset multi-source InSAR fusion optimization monitoring strategy specifically includes:

[0016] Low vegetation coverage areas and high vegetation coverage areas are divided based on visual interpretation of optical remote sensing images;

[0017] The low vegetation coverage areas are monitored based on C-band data;

[0018] The high vegetation coverage areas are monitored based on L-band data;

[0019] The LiCSAR data is taken as supplementary monitoring result data.

[0020] Preferably, the method further comprises a landslide classifier model construction step, comprising:

[0021] A dataset construction step acquires optical remote sensing images within a first time period, constructs a landslide identification sample dataset based on the optical remote sensing images, and performs data enhancement processing on the landslide identification sample dataset to obtain a landslide identification dataset.

[0022] A model construction step uses a preset deep learning model to perform model training on the landslide identification dataset to obtain the landslide classifier model.

[0023] Further preferably, the data enhancement processing on the landslide identification sample dataset to obtain a landslide identification dataset specifically comprises:

[0024] annotating the landslide regions in the optical remote sensing images based on a labelme annotation tool to obtain annotated optical remote sensing images;

[0025] performing transformation processing on the annotated optical remote sensing images to obtain transformed annotated optical remote sensing images; wherein the transformation processing comprises at least four of rotating 90°, rotating 180°, rotating 270°, horizontal mirroring, and vertical mirroring.

[0026] constructing binary mask data corresponding to the annotated optical remote sensing images and the transformed annotated optical remote sensing images; setting landslide sample values in the binary mask data to 0 and non-landslide sample values to 1.

[0027] Preferably, the landslide monitoring grading step specifically comprises:

[0028] superimposing and intersecting all regional deformation feature data and landslide morphology identification data, and determining the overlapping region as high-risk landslide point data;

[0029] determining data outside the overlapping region as medium-risk landslide point data.

[0030] Preferably, the landslide monitoring grading step further comprises obtaining landslide monitoring data according to the regional deformation feature data, landslide morphology identification data, and historical landslide point data.

[0031] Further preferably, obtaining landslide monitoring data according to the regional deformation feature data, landslide morphology identification data, and historical landslide point data specifically comprises:

[0032] superimposing and intersecting all regional deformation feature data, landslide morphology identification data, and historical landslide point data, and determining the overlapping region as high-risk landslide point data;

[0033] data outside the coincidence area is determined as medium-risk landslide data.

[0034] In a second aspect, the present application provides a landslide monitoring system for a vegetation dense area based on multi-source InSAR data fusion and deep learning, comprising a data acquisition module, a deformation interpretation module, a regional deformation feature recognition module, a regional morphological feature recognition module, and a landslide monitoring grading module, wherein

[0035] The data acquisition module is configured to acquire multi-source InSAR data of a target area; wherein the multi-source InSAR data comprises Sentinel-1A data of a first time period, ALOS-2 PALSAR-2 ScanSAR data of the first time period, and LiCSAR data of the first time period.

[0036] The deformation interpretation module is configured to perform deformation interpretation processing on the target area of the multi-source InSAR data using a preset MT-InSAR method to obtain deformation rate data; wherein the deformation rate data comprises first deformation rate data, second deformation rate data, and third deformation rate data corresponding to the Sentinel-1A data, the ALOS-2 PALSAR-2 ScanSAR data, and the LiCSAR data, respectively.

[0037] The regional deformation feature recognition module is configured to perform fusion analysis processing on the first deformation rate data, the second deformation rate data, and the third deformation rate data using a preset multi-source InSAR fusion optimization monitoring strategy to obtain regional deformation detection data of the first time period, and perform regional boundary sketching processing on the regional deformation detection data to obtain regional deformation feature data.

[0038] The regional morphological feature recognition module is configured to input a to-be-recognized optical remote sensing image into a landslide classifier model constructed based on deep learning to output landslide morphological recognition data.

[0039] The landslide monitoring grading module is configured to obtain landslide monitoring data according to the regional deformation feature data and the landslide morphological recognition data; wherein the landslide monitoring data comprises high-risk landslide point data and medium-risk landslide point data.

[0040] Preferably, the system further comprises a landslide classifier model construction module, which comprises a data set construction module and a model construction module, wherein

[0041] The data set construction module is configured to acquire optical remote sensing images within a first time period, construct a landslide recognition sample data set based on the optical remote sensing images, and perform data enhancement processing on the landslide recognition sample data set to obtain a landslide recognition data set.

[0042] The model construction module is configured to use a preset deep learning model to perform model training on the landslide identification data set to obtain the landslide classifier model.

[0043] The landslide monitoring method and system based on multi-source InSAR data fusion and deep learning provided by the embodiment of the present application can effectively improve the detection rate in the high-vegetation coverage area.

[0044] 1. The present application can effectively improve the detection rate in the high-vegetation coverage area.

[0045] 2. The present application combines multi-source InSAR and deep learning technology to effectively improve the overall accuracy of landslide hazard point identification.

[0046] 3. The present application proposes a "deformation-morphology" element fusion grading standard to effectively improve the reliability of landslide hazard grading.

[0047] 4. The present application also proposes a "deformation-morphology-history" three-element fusion grading standard to further improve the reliability of landslide hazard grading.

[0048] 5. The present application is based on high-quality regional deformation rate results and high-resolution optical images, and combines deep learning model application in landslide detection to replace manual detection and achieve automatic detection, which significantly improves the identification accuracy and work efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0049] Figure 1 The flowchart of the landslide monitoring method based on multi-source InSAR data fusion and deep learning provided by the first embodiment of the present application is shown in the figure.

[0050] Figure 2 The deformation monitoring schematic diagram of the Sentinel-1A data after deformation interpretation processing provided by the embodiment of the present application is shown in the figure.

[0051] Figure 3 The deformation monitoring schematic diagram of the ALOS-2 PALSAR-2 ScanSAR data after deformation interpretation processing provided by the embodiment of the present application is shown in the figure.

[0052] Figure 4 The deformation monitoring schematic diagram of the LiCSAR data after deformation interpretation processing provided by the embodiment of the present application is shown in the figure.

[0053] Figure 5A fusion analysis processed multi-source InSAR fusion optimization deformation monitoring schematic diagram provided for an embodiment of the present application;

[0054] Figure 6 A high deformation area schematic diagram provided for an embodiment of the present application after regional deformation feature recognition;

[0055] Figure 7 A recognition result schematic diagram provided for an embodiment of the present application after regional morphological feature recognition;

[0056] Figure 8 A landslide hidden danger point grade division schematic diagram provided for a first embodiment of the present application after landslide monitoring hierarchical processing;

[0057] Figure 9 A system block diagram of a vegetation dense area landslide monitoring system based on multi-source InSAR data fusion and deep learning provided for the first embodiment of the present application;

[0058] Figure 10 A flowchart of a vegetation dense area landslide monitoring method based on multi-source InSAR data fusion and deep learning provided for a second embodiment of the present application;

[0059] Figure 11 An enhanced processing and corresponding mask schematic diagram of a sample in a landslide recognition sample data set provided for an embodiment of the present application;

[0060] Figure 12 A system block diagram of a vegetation dense area landslide monitoring system based on multi-source InSAR data fusion and deep learning provided for the second embodiment of the present application;

[0061] Figure 13 Another system block diagram of a vegetation dense area landslide monitoring system based on multi-source InSAR data fusion and deep learning provided for the second embodiment of the present application;

[0062] Figure 14 A flowchart of a vegetation dense area landslide monitoring method based on multi-source InSAR data fusion and deep learning provided for a third embodiment of the present application;

[0063] Figure 15 A flowchart of a vegetation dense area landslide monitoring method based on multi-source InSAR data fusion and deep learning provided for a fourth embodiment of the present application. DETAILED DESCRIPTION

[0064] In order to make the objects, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work belong to the scope of protection of the present application.

[0065] The technical solutions of the present application will be further described in detail below with reference to the drawings and embodiments.

[0066] The method provided by the embodiment of the present application is a landslide monitoring method based on multi-source InSAR data fusion and deep learning in a vegetation dense area. The multi-source InSAR data and deep learning technology are applied to the landslide monitoring technology, the way of identification based only on deformation or shape in the traditional monitoring method is changed, the landslide point identification way of deformation, shape and / or historical point data element fusion is realized, and the element fusion grading standard is realized.

[0067] [First embodiment]

[0068] [The method is a landslide monitoring method based on multi-source InSAR data fusion and deep learning in a vegetation dense area]

[0069] Figure 1 The flowchart of the landslide monitoring method based on multi-source InSAR data fusion and deep learning in a vegetation dense area provided by the first embodiment of the present application is as follows, which will be described below in combination with Figure 1 The technical solutions of the present application will be further described in detail below with reference to the drawings and embodiments.

[0070] As shown in Figure 1 The method provided by the present application is a landslide monitoring method based on multi-source InSAR data fusion and deep learning in a vegetation dense area, which includes the following steps:

[0071] S1100, data acquisition step, acquiring multi-source InSAR data of a target area.

[0072] Specifically, in the present application, the target area refers to the landslide risk monitoring area to which the method of the embodiments of the present application is applied, that is, a specific area for which landslide risk needs to be confirmed. Multi-source InSAR data is obtained from the outside, such as reading from a special database or searching for corresponding data from publicly available data on the Internet. The multi-source InSAR data includes Sentinel-1A data of a first period, Advanced Land Observing Satellite-2 L-band Phased Array Synthetic Aperture Radar 2 ScanSAR (ALOS-2 PALSAR-2 ScanSAR) data of the first period, and The Looking into Continents from Space with Synthetic Aperture Radar (LiCSAR) data of the first period. The first period refers to a specific time period, for example: January 1, 2020 to December 30, 2024, the last 5 years from the date of using the method, the last 10 years, etc. The specific time period can be set according to the specific circumstances of using the method and the available data. The data types include Sentinel-1A data, ALOS-2 PALSAR-2 ScanSAR data and LiCSAR data. Each data type can also include ascending track images and / or descending track images. In the embodiments of the present application, the number of each data type is not restricted when obtaining InSAR data, and can be adjusted and set according to the actual situation of using the present application.

[0073] In a specific example of the embodiments of the present application, the obtained multi-source InSAR data specifically includes the following data:

[0074] Sentinel-1A ascending track images from January 12, 2020 to December 16, 2024, a total of 60 images;

[0075] Sentinel-1A descending track images from January 19, 2020 to December 23, 2024, a total of 60 images;

[0076] ALOS-2 PALSAR-2 ScanSAR descending track images from January 3, 2020 to March 24, 2023, a total of 33 images;

[0077] LiCSAR from January 2020 to December 2024, a total of 1082 ascending track and 500 descending track interference pairs and unwrapping results.

[0078] S1200, a deformation interpretation step, using a preset MT-InSAR method to perform deformation interpretation processing on a target region of multi-source InSAR data, to obtain deformation rate data.

[0079] Specifically, the predetermined Multi-Temporal Interferometric Synthetic Aperture Radar (MT-InSAR) is an extension of the InSAR technology, which generates time series information of surface deformation by using multi-temporal SAR image data. This technology is often used to monitor small changes in surface deformation, such as earthquakes, volcanic activity, landslides, or ground subsidence, etc. The MT-InSAR method in the present application is a preset method, and the corresponding method suitable for the multi-source InSAR data of the present application is selected from the prior art.

[0080] For the three different types of InSAR data, different methods are used for deformation interpretation processing to obtain deformation rate data corresponding to each type of data. For example, the Small Baseline Subset Interferometric Synthetic Aperture Radar (SBAS-InSAR) method is used for processing Sentinel-1A data and ALOS-2 PALSAR-2 ScanSAR data, and the reference point in the Persistent Scatterer Interferometric Synthetic Aperture Radar (PS-InSAR) method is applied to SBAS-InSAR as a phase-optimized Ground Control Point (GCP), and the The Looking into Continents with Small Baseline Subset (LICSBAS) method is used for processing regional LiCSAR data. Among them, the SBAS-InSAR method, the PS-InSAR method and the LICSBAS method are general methods available in the prior art and improved methods thereof.

[0081] After processing, the obtained deformation rate data includes first deformation rate data corresponding to Sentinel-1A data, second deformation rate data corresponding to ALOS-2 PALSAR-2 ScanSAR data, and third deformation rate data corresponding to LiCSAR data, respectively.

[0082] In a specific example of the embodiment of the present application, the step further comprises the following steps:

[0083] S1201, using a preset PS-InSAR method to extract GCP point data in Sentinel-1A data and ALOS-2 PALSAR-2 ScanSAR data.

[0084] Specifically, PS-InSAR is a Persistent Scatterer (PS) based InSAR technology, which extracts high-precision ground deformation time series by identifying and analyzing the interference phase of PS points. It is suitable for millimeter-level deformation monitoring in urban areas, densely populated areas or stable terrain areas. The present application uses a preset PS-InSAR method to extract GCP point data in Sentinel-1A data and ALOS-2 PALSAR-2 ScanSAR data. In the embodiment of the present application, the preset PS-InSAR method is selected from the general PS-InSAR method available in the prior art and the method improved therefrom.

[0085] S1202, taking the GCP point data as input of a preset SBAS-InSAR method to perform phase correction processing to obtain first deformation rate data and second deformation rate data.

[0086] Specifically, SBAS-InSAR is an InSAR technology combined with small baseline set strategy, which generates high-density and high-reliability deformation time series by screening high-quality image pairs. It is suitable for large-scale ground deformation monitoring, such as earthquakes, volcanoes, landslides, urban subsidence, etc., and performs excellently in vegetation-covered or complex terrain areas. In the embodiment of the present application, the GCP point data is taken as input of a preset SBAS-InSAR method to perform phase correction processing to obtain first deformation rate data and second deformation rate data. In the embodiment of the present application, the preset SBAS-InSAR method is selected from the general SBAS-InSAR method available in the prior art and the method improved therefrom.

[0087] S1203, using a preset LICSBAS method to process LiCSAR data to obtain third deformation rate data.

[0088] Specifically, LICSBAS is an open source InSAR time series analysis package integrated with LiCSAR, which can be used for landslide monitoring and analysis using SAR data. The present application uses a preset LICSBAS to process LiCSAR data to obtain corresponding third deformation rate data.

[0089] The implementation process of the embodiment of the present application in one specific example includes the following steps S1201-S203:

[0090] Firstly, the reference points in PS-InSAR are applied to SBAS-InSAR as GCP control points for phase optimization, one SAR image in the time series is selected as the master image, all auxiliary images are registered with the master image, a total of 59 relative (32 relative for ALOS-2 PALSAR-2 ScanSAR data) are obtained, all auxiliary images are interferometrically processed with the master image, the GACOS data is used to correct the interference results, the multi-view is set to 4:1 (2:1 for ALOS-2 PALSAR-2 ScanSAR data), the Shuttle Radar Topography Mission Version 1 Digital Elevation Model 30-meter Resolution (SRTMV1 30m DEM) data is used for registration. All interference results are first inverted, the maximum area of the reference point is set to 25 sqkm, the sub-area overlap ratio is 30%, and the number of candidates is 300; the reference GCP point file in the first inversion result is converted into a GCP file in the SAR coordinate system; the GCP file is applied to the SBAS-InSAR first inversion for phase optimization and re-flattening.

[0091] Secondly, the regional Sentinel-1A data and ALOS-2 PALSAR-2 ScanSAR data are processed by using the SBAS-InSAR method, including: selecting a SAR image in the time series as the main image, checking the time baseline and spatial baseline distribution, setting the maximum spatial baseline as 4.6% (300m), the maximum time baseline as 120d, and the minimum connection number as 10; registering all auxiliary images with the main image, obtaining 215 relative ascending tracks and 204 relative descending tracks (85 relative for ALOS-2 PALSAR-2 ScanSAR data); performing interference processing on all auxiliary images and the main image, applying the GCP point file generated in step S211 to phase optimization and re-levelling, using GACOS data to correct the interference results, setting the multi-view as 4:1 (2:12 for ALOS-2 PALSAR-2 ScanSAR data), using SRTMV130m DEM data for registration, selecting the Minimum Cost Flow (MCF) as the unwrapping method, setting the unwrapping decomposition level as level 1, the correlation threshold as 0.3, and the Goldstein as the filtering method; performing the first inversion on all interference processing results, applying the GCP point file generated in step S211 to phase optimization and re-levelling, setting the product correlation threshold as 0.3, selecting the Liner as the deformation model, setting the spatial wavelet size as 120m, selecting the Minimum Cost Flow as the unwrapping method, setting the unwrapping decomposition level as level 1, the correlation threshold as 0.3, and the Goldstein as the filtering method; performing the second inversion on the basis of the first inversion results, applying the GCP point file generated in step S211 to phase optimization and re-levelling, setting the product correlation threshold as 0.3, the atmospheric low-pass filter as 1600m, and the atmospheric high-pass filter as 365d; performing geographic coding on the second inversion results, outputting the results in the WGS-84 geographic coordinate system, setting the resolution as 30m, and obtaining the preliminary deformation rate, including the first deformation rate data corresponding to the Sentinel-1A data and the second deformation rate data corresponding to the ALOS-2 PALSAR-2 ScanSAR data.

[0092] Then, the regional LiCSAR data was processed using the LICSBAS-based method, including: preprocessing all downloaded LiCSAR data products, applying GACOS data for atmospheric correction with a mask threshold of 0.2; performing quality checks on all preprocessed results to identify and remove defective interferograms with an unwrapping threshold of 0.3 and a correlation threshold of 0.2; performing loop closure checks on all quality-checked results to identify and remove defective interferograms with a loop error parameter set to 3 rad; performing small baseline inversion on all quality-checked results; calculating the standard deviation of deformation rate on the inversion results; and performing masking and filtering on the time series to output the filtered deformation rate, which is the third deformation rate data corresponding to the LiCSAR data.

[0093] In the above specific examples of the present invention, after processing through the above steps, the first deformation rate data corresponding to Sentinel-1A data, the second deformation rate data corresponding to ALOS-2 PALSAR-2 ScanSAR data, and the third deformation rate data corresponding to LiCSAR data are obtained respectively. Figure 2 This is a schematic diagram of Sentinel-1A data deformation monitoring after deformation interpretation processing provided in an embodiment of the present invention. It shows a schematic diagram of Sentinel-1A data deformation monitoring after deformation interpretation processing of Sentinel-1A down-orbit images, as follows. Figure 2 As shown, it contains the first deformation rate data. Figure 3 This invention provides a schematic diagram of ALOS-2 PALSAR-2 ScanSAR data deformation monitoring, which includes first deformation rate data. The diagram illustrates ALOS-2 PALSAR-2 ScanSAR data deformation monitoring after deformation interpretation processing of ALOS-2 PALSAR-2 ScanSAR de-orbiting images. Figure 3 As shown, it contains second deformation rate data. Figure 4 This is a schematic diagram of LiCSAR data deformation monitoring after deformation interpretation processing provided in an embodiment of the present invention. It illustrates the deformation monitoring of LiCSAR data after deformation interpretation processing of LiCSAR de-orbiting images. Figure 4 As shown, it contains third deformation rate data.

[0094] S1300, Regional deformation feature identification step: The first deformation rate data, the second deformation rate data and the third deformation rate data are fused and analyzed using a preset multi-source InSAR fusion optimization monitoring strategy to obtain regional deformation detection data for the first time period. The regional deformation detection data is then processed to delineate the regional boundaries to obtain regional deformation feature data.

[0095] Specifically, in the embodiments of the present application, the preset multi-source InSAR fusion optimization monitoring strategy is a strategy preset before the implementation of the present application, which specifically includes:

[0096] The low-vegetation coverage area and the high-vegetation coverage area are divided based on visual interpretation of optical remote sensing images. Specifically, the differences in the influence of vegetation on SAR signals include: low-vegetation coverage area: low vegetation height and small density, the backscattering of C-band SAR mainly comes from the ground, which is suitable for monitoring terrain deformation or ground changes. High-vegetation coverage area: dense vegetation, C-band signals are easily reflected or attenuated by vegetation, while L-band has strong penetration ability, which can reduce the interference of vegetation and obtain information of the ground under the forest. Through the GIS platform, the low-vegetation coverage area and the high-vegetation coverage area are divided based on visual interpretation of optical remote sensing images, which can realize selection of appropriate bands for different vegetation coverage types, avoid data redundancy, and improve monitoring accuracy and efficiency.

[0097] The low-vegetation coverage area is monitored based on C-band data. Specifically, C-band has high spatial resolution (such as 10-20 m of Sentinel-1), which is suitable for monitoring subtle ground changes (such as soil cracks and small-scale subsidence) in low-vegetation areas. It is sensitive to short vegetation (such as grassland and crops), and can analyze vegetation growth or land use changes through time series analysis. In the low-vegetation coverage area, C-band (Sentinel-1A) data is used for monitoring, and C-band data can accurately capture ground deformation with an error of millimeters to centimeters.

[0098] The high-vegetation coverage area is monitored based on L-band data. Specifically, L-band has strong penetration ability and can penetrate dense vegetation (such as forest canopy) to obtain information of the ground under the forest. It is sensitive to soil moisture and ground roughness, and is suitable for monitoring hydrological or geological changes (such as landslides and permafrost thaw) in high-vegetation areas. In the high-vegetation coverage area, L-band (ALOS-2 PALSAR-2 ScanSAR) data is used for monitoring, and L-band data can effectively penetrate vegetation, reduce canopy interference, and extract information of the ground under the forest with a vertical accuracy of several decimeters to meters.

[0099] LiCSAR data is used as supplementary monitoring result data. Specifically, LiCSAR data can provide lightweight and fast processing of monitoring results, making up for the lack of C / L-band data in time resolution or coverage range, and can enhance data diversity. In areas with missing data or complex terrain (such as mountainous areas and polar regions), LiCSAR data can be used as a supplement to improve the robustness of monitoring, which has special scene adaptability. LiCSAR data as supplementary monitoring result data can enhance the rapid response capability to extreme environments (such as heavy rain and earthquakes), and assist in quickly identifying disaster events (such as earthquakes and volcanoes) or temporary ground changes.

[0100] The multi-source InSAR fusion optimization monitoring strategy provided in this embodiment of the invention classifies vegetation cover types through visual interpretation of optical images and combines the complementary advantages of C-band and L-band SAR data to achieve accurate monitoring of low / high vegetation areas. Simultaneously, the introduction of LiCSAR data as a supplement enhances the comprehensiveness of the monitoring. This multi-band collaborative strategy has significant advantages in improving accuracy, adapting to complex environments, and increasing efficiency, and is suitable for the geological morphology and natural conditions of the target area of ​​this invention.

[0101] In this embodiment of the invention, this step specifically includes the following steps:

[0102] S1301, a preset multi-source InSAR fusion optimization monitoring strategy is used to perform fusion analysis on the first deformation rate data, the second deformation rate data, and the third deformation rate data to obtain the regional deformation detection data for the first time period. Specifically, after fusion analysis through the preset multi-source InSAR fusion optimization monitoring strategy, the corresponding regional deformation detection data is obtained. Figure 5 This is a schematic diagram of multi-source InSAR fusion-optimized deformation monitoring after fusion analysis and processing, provided in an embodiment of the present invention. Figure 5 As shown, it includes regional deformation detection data.

[0103] S1302, Perform region boundary delineation processing on the region deformation detection data to obtain region deformation feature data. Specifically, use a preset deep learning method, such as a deep learning model based on "Unet-Resnet50", to perform region boundary delineation processing on the region deformation detection data to obtain region deformation feature data.

[0104] In this embodiment of the invention, this step can also be performed by the following steps: after performing fusion analysis on the first deformation rate data, the second deformation rate data and the third deformation rate data using a preset InSAR fusion optimization monitoring strategy, the regions with an annual average deformation rate greater than 20 mm / y in the ascent and descent orbit deformation results are plotted according to the deformation monitoring results to obtain the high deformation region. Figure 6 This is a schematic diagram of a high-deformation region after regional deformation feature recognition provided in an embodiment of the present invention, as shown below. Figure 6 As shown, this includes high-deformation regions indicated by high-deformation region data.

[0105] S1400, Regional Morphological Feature Recognition Step: The optical remote sensing image to be identified is input into a landslide classifier model built based on deep learning to output landslide morphological recognition data.

[0106] Specifically, in this embodiment of the invention, the optical remote sensing image to be identified refers to an optical remote sensing image that needs to be used for landslide monitoring and identification. The landslide classifier model based on deep learning provided in this embodiment of the invention can be a pre-trained landslide classifier model, which is embedded in the method of this invention. Alternatively, it can be a landslide classifier model based on deep learning constructed using the method provided by this invention. In the second embodiment of the invention, the method for constructing the landslide classifier model provided by this invention will be described in detail.

[0107] In this step, the optical remote sensing image to be identified is input into a landslide classifier model built based on deep learning, and the model can directly output landslide morphology identification data. Figure 7 This is a schematic diagram of the recognition result after regional morphological feature recognition provided in an embodiment of the present invention, such as... Figure 7 As shown, it includes landslide morphology identification data.

[0108] S1500, the landslide monitoring classification step, obtains landslide monitoring data based on regional deformation characteristic data and landslide morphology identification data.

[0109] Specifically, in this embodiment of the invention, this step involves overlaying and intersecting all regional deformation characteristic data and landslide morphology identification data to identify the overlapping areas as high-risk landslide point data; and identifying the data outside the overlapping areas as medium-risk landslide point data. In other words, the landslide monitoring data includes both high-risk and medium-risk landslide point data. Figure 8 This is a schematic diagram illustrating the classification of landslide hazard points after landslide monitoring and grading processing, provided in the first embodiment of the present invention. Figure 8 As shown, this includes high-risk landslide locations formed from high-risk landslide data and medium-risk landslide locations formed from medium-risk landslide data, which is landslide monitoring data.

[0110] The above description, with reference to the accompanying drawings, details a landslide monitoring method for densely vegetated areas based on multi-source InSAR data fusion and deep learning, according to the first embodiment of the present invention. The present invention also provides a landslide monitoring system for densely vegetated areas based on multi-source InSAR data fusion and deep learning, used to implement the landslide monitoring method for densely vegetated areas provided by the present invention. The following description, with reference to the accompanying drawings, details a landslide monitoring system for densely vegetated areas based on multi-source InSAR data fusion and deep learning provided by the present invention.

[0111] [A Landslide Monitoring System for Densely Vegetated Areas Based on Multi-Source InSAR Data Fusion and Deep Learning]

[0112] Figure 9A system block diagram of a landslide monitoring system in densely vegetated areas based on multi-source InSAR data fusion and deep learning, as provided in the first embodiment of the present invention, is shown below. Figure 9 As shown, the monitoring system 1000 provided by the present invention includes a data acquisition module 1001, a deformation interpretation module 1002, a regional deformation feature recognition module 1003, a regional morphological feature recognition module 1004, and a landslide monitoring and grading module 1005, wherein...

[0113] The data acquisition module 1001 is used to acquire multi-source InSAR data of the target area; wherein, the multi-source InSAR data includes Sentinel-1A data of the first time period, ALOS-2 PALSAR-2 ScanSAR data of the first time period, and LiCSAR data of the first time period.

[0114] The deformation interpretation module 1002 is used to perform deformation interpretation processing on the target area of ​​multi-source InSAR data using a preset MT-InSAR method to obtain deformation rate data; wherein, the deformation rate data includes first deformation rate data, second deformation rate data and third deformation rate data corresponding to Sentinel-1A data, ALOS-2PALSAR-2 ScanSAR data and LiCSAR data respectively.

[0115] The regional deformation feature identification module 1003 is used to perform fusion analysis on the first deformation rate data, the second deformation rate data and the third deformation rate data using a preset multi-source InSAR fusion optimization monitoring strategy to obtain the regional deformation detection data for the first time period, and to perform regional boundary delineation on the regional deformation detection data to obtain regional deformation feature data.

[0116] The regional morphological feature recognition module 1004 is used to input the optical remote sensing image to be identified into the landslide classifier model built based on deep learning, so as to output landslide morphological recognition data.

[0117] The landslide monitoring and grading module 1005 is used to obtain landslide monitoring data based on regional deformation characteristic data and landslide morphology identification data; the landslide monitoring data includes data on high-risk landslide points and data on medium-risk landslide points.

[0118] The method and system provided by the first embodiment of the present invention have been described in detail above with reference to the accompanying drawings and embodiments. Next, the second embodiment of the present invention will be described in detail with reference to the accompanying drawings.

[0119] [Second Embodiment]

[0120] [A Landslide Monitoring Method Based on Multi-Source InSAR Data Fusion and Deep Learning in Densely Vegetated Areas]

[0121] In addition to the relevant method steps in the first embodiment, the second embodiment of the present invention also provides a method for constructing a landslide classifier model based on deep learning. The method provided by the second embodiment of the present invention will be described in detail below with reference to the accompanying drawings. Figure 10 A flowchart illustrating a landslide monitoring method in densely vegetated areas based on multi-source InSAR data fusion and deep learning, as provided in the second embodiment of the present invention, is shown below. Figure 10 As shown, in addition to steps S2100 to S2500 corresponding to steps S1110-S1500 in the first embodiment of the present invention, the following steps are also included:

[0122] S2600, Steps for building a landslide classifier model: Build a dataset to train a preset deep learning model and build a landslide classifier model.

[0123] Specifically, in this embodiment of the invention, a method for constructing a landslide classifier model is provided, which includes the following steps:

[0124] In the dataset construction step S2601, optical remote sensing images within the first time period are acquired, a landslide identification sample dataset is constructed based on the optical remote sensing images, and data augmentation processing is performed on the landslide identification sample dataset to obtain the landslide identification dataset.

[0125] Specifically, in this embodiment of the invention, optical remote sensing images within a first time period are acquired from an external source. The specific data source for acquiring these optical remote sensing images is set according to the specific target scene and available data using the method of this invention; this invention does not impose constraints on the data source. Externally acquired optical remote sensing images are obtained that are temporally consistent with multi-source InSAR data of the target area. Based on these optical remote sensing images, a landslide identification sample dataset is created and data augmentation processing is performed. Training and testing sets are then divided, thereby constructing the landslide identification dataset.

[0126] In this embodiment of the invention, data augmentation processing is performed on the landslide identification sample dataset to obtain the landslide identification dataset, specifically including:

[0127] First, the landslide areas in the optical remote sensing image are labeled using the labelme annotation tool to obtain an annotated optical remote sensing image.

[0128] Specifically, in one example provided by this invention, the optical remote sensing image is a high-resolution optical satellite image. Typical landslide areas in the high-resolution optical satellite image are labeled using the Labelme annotation tool, thus constructing an annotated optical remote sensing image. After annotating 90 optical remote sensing image data points, a self-made landslide dataset of 90 data points is obtained. Next, the annotated optical remote sensing image is transformed to obtain a transformed annotated optical remote sensing image.

[0129] Specifically, in the embodiments of the invention, the transformation processing of the labeled optical remote sensing image includes at least four of the following: rotation by 90°, rotation by 180°, rotation by 270°, horizontal mirroring, and vertical mirroring. The choice of which transformation method to use depends on the size of the dataset and the characteristics of the samples, and is not constrained here.

[0130] In this embodiment of the invention, an existing landslide dataset is also provided. The self-made landslide dataset obtained after the annotation process in the previous step together constitutes the dataset required by this invention. Random rotation of 90°, 180°, 270° and vertical mirroring operations are performed on the annotated optical remote sensing images (samples) in the existing landslide dataset and the self-made landslide dataset, resulting in a total of 3440 image data, of which 3080 are existing data and 360 are self-made data.

[0131] As shown in the examples above, the enhancement process expands the number of samples in both the existing landslide dataset and the self-made landslide dataset. Finally, binary mask data corresponding to the labeled optical remote sensing images and the transformed labeled optical remote sensing images are constructed; the landslide sample values ​​in the binary mask data are set to 0, and the non-landslide sample values ​​are set to 1. Figure 11 The figure shows an enhanced processing of a sample in a landslide identification sample dataset provided in an embodiment of the present invention and a corresponding mask diagram. It includes a sample and the sample data obtained after random rotation by 90°, 180°, 270° and vertical mirroring operations, and the corresponding mask diagram.

[0132] In this embodiment of the invention, the samples in the existing landslide dataset and the self-made landslide dataset are processed separately to obtain the landslide identification dataset of the present invention.

[0133] In this embodiment of the invention, the constructed landslide identification dataset is divided into a training set and a validation set, which are used to train and validate a preset deep learning model, respectively.

[0134] In model building step S2602, a pre-set deep learning model is used to train the landslide identification dataset to obtain a landslide classifier model.

[0135] Specifically, the preset deep learning model provided in this embodiment of the invention is the "Unet-Resnet50" deep learning model, which is built and trained based on the deep learning framework of the open-source Python machine learning library PyTorch. The landslide classifier model provided by this invention is obtained by training it using a landslide identification dataset. In this embodiment of the invention, the preset deep learning model is not limited to the "Unet-Resnet50" deep learning model; other suitable general-purpose deep learning models can also be used. The selection of the preset deep learning model can be based on the regional and natural features of the target area for which this invention is applied.

[0136] In a specific example, the construction of the landslide classifier model provided in this embodiment of the invention includes the following steps:

[0137] First, the samples in the constructed landslide identification dataset are cropped to 256×256 pixels.

[0138] Secondly, the training set and validation set were randomly divided in a 7:3 ratio, with 2408 data entries in the training set and 1032 data entries in the validation set.

[0139] Next, the "Unet-Resnet50" deep learning model was trained using the training set, iterating 2000 times.

[0140] Next, the trained deep learning model is used to predict the data in the test set to verify the accuracy of the deep learning model. The test results of the trained model are: train_loss: 0.0106, val_loss: 0.0537, lr: 0.000000, Accuracy(PA): 98.7, Precision(CPA): ['99.1', '94.6'], Recall: ['99.5', '90.9'], IoU: ['98.6', '86.4'], MPA: 96.9, MIoU: 92.5, which meets the preset requirements. At this time, the landslide classifier model of the present invention is completed.

[0141] It should be noted that the landslide classifier model construction step S2600 provided in this embodiment of the invention can be executed before step S2000 or before S2400; the specific execution order is not constrained here. Figure 10 As shown, the landslide classifier model construction step S2600 is executed before step S2400. The specific implementation methods of steps S2100 to S2500 have been described in the above embodiment one and will not be repeated here.

[0142] [A Landslide Monitoring System for Densely Vegetated Areas Based on Multi-Source InSAR Data Fusion and Deep Learning]

[0143] Accordingly, the second embodiment of the present invention also provides another landslide monitoring system 2000 based on multi-source InSAR data fusion and deep learning in densely vegetated areas, for implementing the landslide monitoring method based on multi-source InSAR data fusion and deep learning in densely vegetated areas provided by the present invention. Figure 12 A system block diagram of a landslide monitoring system in densely vegetated areas based on multi-source InSAR data fusion and deep learning, as provided in the second embodiment of the present invention, is shown below. Figure 12 As shown, in addition to the data acquisition module 2001, deformation interpretation module 2002, regional deformation feature recognition module 2003, regional morphological feature recognition module 1004, and landslide monitoring and grading module 1005 corresponding to the data acquisition module 1001, deformation interpretation module 1002, regional deformation feature recognition module 1003, regional morphological feature recognition module 1004, and landslide monitoring and grading module 1005 of the first embodiment, the system 2000 also includes a landslide classifier model construction module 2006, which is used to construct a landslide classifier model.

[0144] Figure 13 A system block diagram of another landslide monitoring system in densely vegetated areas based on multi-source InSAR data fusion and deep learning, as provided in the second embodiment of the present invention, is shown below. Figure 13 As shown, in another specific example of this embodiment of the invention, the landslide classifier model building module 2006 further includes a dataset building module 20061 and a model building module 20062, wherein...

[0145] The dataset construction module 20061 is used to acquire optical remote sensing images within the first time period, construct a landslide identification sample dataset based on the optical remote sensing images, and perform data augmentation processing on the landslide identification sample dataset to obtain the landslide identification dataset.

[0146] The model building module 20062 is used to train a landslide identification dataset using a preset deep learning model to obtain a landslide classifier model.

[0147] The above provides a detailed description of the landslide monitoring method and system based on multi-source InSAR data fusion and deep learning in densely vegetated areas provided by the second embodiment of the present invention. The landslide classifier model constructed by the second embodiment of the present invention is highly targeted. By utilizing the landslide monitoring method based on multi-source InSAR data fusion and deep learning in densely vegetated areas implemented by it, the accuracy of landslide monitoring in the target area of ​​the present invention is improved.

[0148] Next, the third embodiment of the present invention will be described in conjunction with the accompanying drawings.

[0149] [Third Embodiment]

[0150] [A Landslide Monitoring Method Based on Multi-Source InSAR Data Fusion and Deep Learning in Densely Vegetated Areas]

[0151] The third embodiment of the present invention provides a landslide monitoring method for densely vegetated areas based on multi-source InSAR data fusion and deep learning. It proposes a three-element fusion grading standard of "deformation-morphology-history" to further improve the reliability of landslide hazard level. The method provided by the third embodiment of the present invention will be described in detail below with reference to the accompanying drawings.

[0152] Figure 14 A flowchart illustrating a landslide monitoring method in densely vegetated areas based on multi-source InSAR data fusion and deep learning, as provided in the third embodiment of the present invention, is shown below. Figure 14 As shown, the third embodiment of the present invention provides a landslide monitoring method in densely vegetated areas based on multi-source InSAR data fusion and deep learning, including steps S3100 to S3500. Steps S3100 to S3400 correspond to steps S1100 to S1400 provided in the first embodiment, respectively. The specific implementation of each step has been described in the first embodiment above and will not be repeated here. In this embodiment, step S3500 replaces S1500 in the first embodiment to achieve the fusion and grading standard of the three elements of "deformation-morphology-history".

[0153] S3500, the landslide monitoring classification steps, obtains landslide monitoring data based on regional deformation characteristic data, landslide morphology identification data, and historical landslide location data.

[0154] Specifically, in this embodiment of the invention, the historical landslide points are pre-acquired landslide point data that indicates a past landslide in the target area. The landslide monitoring data is obtained based on regional deformation characteristic data, landslide morphology identification data, and historical landslide point data by: overlaying and intersecting all regional deformation characteristic data, landslide morphology identification data, and historical landslide point data; identifying the overlapping areas as high-risk landslide point data; and identifying the data outside the overlapping areas as medium-risk landslide data.

[0155] [Fourth Embodiment]

[0156] [A Landslide Monitoring Method Based on Multi-Source InSAR Data Fusion and Deep Learning in Densely Vegetated Areas]

[0157] The fourth embodiment provided by the present invention is a variation of the second embodiment. Figure 15 A flowchart illustrating a landslide monitoring method in densely vegetated areas based on multi-source InSAR data fusion and deep learning, as provided in the fourth embodiment of the present invention, is shown below. Figure 15As shown, in addition to steps S4100 to S4400 and step S4600, which correspond to steps S2100 to S2400 and step S2600 respectively provided in the second embodiment, it also provides step S4500 to replace S2500 in the second embodiment. The specific implementation method of step S4500 corresponds to step S3500 in the third embodiment, and will not be described in detail here.

[0158] The various embodiments of the present invention have been described in detail above with reference to the accompanying drawings.

[0159] This invention provides a landslide monitoring method and system for densely vegetated areas based on multi-source InSAR data fusion and deep learning. It utilizes a zonal optimization monitoring strategy that fuses multi-source InSAR data (C-band Sentinel-1A for low vegetation cover areas, L-band ALOS-2 for high vegetation cover areas, and LICSAR as supplementary data), combined with MT-InSAR phase optimization processing technology (using reference points from PS-InSAR as GCP control points for phase optimization in SBAS-InSAR to reduce manual point selection errors) and the LICSBAS automated process. This effectively overcomes the decoherence problem in densely vegetated mountain and canyon areas, obtaining high-precision regional deformation rate maps. The invention also provides a landslide classifier model based on a pre-set deep learning model. This model uses an enhanced landslide sample dataset to achieve automated landslide morphology identification with an accuracy of 98.7%. By constructing a three-element fusion grading standard of "deformation-morphology-history" (matching deformation features, landslide morphology, and historical disaster locations), it achieves intelligent classification of high-risk and medium-risk landslide points. This method significantly improves the accuracy and efficiency of landslide hazard identification in densely vegetated areas, overcoming the limitations of cloud cover and manual interpretation in traditional monitoring methods. It provides reliable technical support for early disaster warning and risk prevention, effectively safeguarding people's lives and property. Compared with existing technologies, it has at least the following significant advantages:

[0160] 1. This invention can effectively improve the detection rate in areas with high vegetation coverage.

[0161] 2. This invention combines multi-source InSAR with deep learning technology to effectively improve the overall accuracy of identifying potential landslide sites.

[0162] 3. This invention proposes a grading standard that integrates "deformation-morphology" elements, which effectively improves the reliability of landslide hazard levels.

[0163] 4. This invention also proposes a three-element integrated grading standard of "deformation-morphology-history" to further improve the reliability of landslide hazard levels.

[0164] 5. This invention is based on high-quality regional deformation rate results and high-resolution optical images, combined with a deep learning model for landslide detection, to replace manual detection and achieve automatic detection, significantly improving recognition accuracy and work efficiency.

[0165] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0166] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented in hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0167] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A landslide monitoring method in densely vegetated areas based on multi-source InSAR data fusion and deep learning, characterized in that, The method includes: The data acquisition step involves acquiring multi-source InSAR data for the target area; wherein the multi-source InSAR data includes Sentinel-1A data from the first time period, ALOS-2 PALSAR-2 ScanSAR data from the first time period, and LiCSAR data from the first time period. The deformation interpretation step involves using a preset MT-InSAR method to perform deformation interpretation processing on the target area of ​​the multi-source InSAR data to obtain deformation rate data; wherein, the deformation rate data includes first deformation rate data, second deformation rate data, and third deformation rate data corresponding to Sentinel-1A data, ALOS-2PALSAR-2 ScanSAR data, and LiCSAR data, respectively. The regional deformation feature identification step involves using a preset multi-source InSAR fusion optimization monitoring strategy to perform fusion analysis on the first deformation rate data, the second deformation rate data, and the third deformation rate data to obtain the regional deformation detection data for the first time period, and then performing regional boundary delineation on the regional deformation detection data to obtain regional deformation feature data. The regional morphological feature identification step involves inputting the optical remote sensing image to be identified into a landslide classifier model built based on deep learning to output landslide morphological identification data. The landslide monitoring and grading process involves obtaining landslide monitoring data based on the regional deformation characteristic data and landslide morphology identification data; wherein the landslide monitoring data includes data on high-risk landslide points and data on medium-risk landslide points. The landslide monitoring and grading steps are specifically as follows: All regional deformation characteristic data and landslide morphology identification data are overlaid and intersected, and the overlapping areas are identified as high-risk landslide points. Data outside the overlapping area were identified as medium-risk landslide points.

2. The method according to claim 1, characterized in that, The deformation interpretation steps specifically include: GCP point data in the Sentinel-1A data and the ALOS-2 PALSAR-2 ScanSAR data were extracted using a preset PS-InSAR method. The GCP point data is used as input to a preset SBAS-InSAR method for phase correction processing to obtain the first deformation rate data and the second deformation rate data. The LiCSAR data is processed using the preset LICSBAS method to obtain the third deformation rate data.

3. The method according to claim 1, characterized in that, The preset multi-source InSAR fusion optimization monitoring strategy is specifically as follows: Based on visual interpretation of optical remote sensing images, low vegetation cover areas and high vegetation cover areas are divided; The low vegetation cover area was monitored using C-band data; The high vegetation cover area was monitored using L-band data; The LiCSAR data was used as supplementary monitoring results.

4. The method according to claim 1, characterized in that, The method also includes a landslide classifier model construction step, including: The dataset construction steps are as follows: acquire optical remote sensing images within the first time period, construct a landslide identification sample dataset based on the optical remote sensing images, and perform data augmentation processing on the landslide identification sample dataset to obtain the landslide identification dataset. The model building step involves training the landslide identification dataset using a preset deep learning model to obtain the landslide classifier model.

5. The method according to claim 4, characterized in that, The process of performing data augmentation on the landslide identification sample dataset to obtain the landslide identification dataset specifically includes: The landslide areas in the optical remote sensing image are labeled using the labelme annotation tool to obtain an annotated optical remote sensing image; Transformation processing is performed on labeled optical remote sensing images to obtain transformed labeled optical remote sensing images; wherein, the transformation processing includes at least four of the following: rotation by 90°, rotation by 180°, rotation by 270°, horizontal mirroring, and vertical mirroring; Construct and label binary mask data corresponding to the optical remote sensing image and the transformed labeled optical remote sensing image respectively; set the landslide sample value to 0 and the non-landslide sample value to 1 in the binary mask data.

6. The method according to claim 1, characterized in that, The landslide monitoring and grading steps also include obtaining landslide monitoring data based on the regional deformation characteristic data, landslide morphology identification data, and historical landslide location data.

7. The method according to claim 6, characterized in that, The landslide monitoring data obtained based on the aforementioned regional deformation characteristic data, landslide morphology identification data, and historical landslide location data are as follows: All regional deformation characteristic data, landslide morphology identification data and historical landslide location data are overlaid and intersected to identify the overlapping areas as high-risk landslide points. Data outside the overlapping area were identified as medium-risk landslide data.

8. A landslide monitoring system for densely vegetated areas based on multi-source InSAR data fusion and deep learning, characterized in that, The system includes a data acquisition module, a deformation interpretation module, a regional deformation feature recognition module, a regional morphological feature recognition module, and a landslide monitoring and grading module. The data acquisition module is used to acquire multi-source InSAR data of the target area; wherein, the multi-source InSAR data includes Sentinel-1A data of the first time period, ALOS-2 PALSAR-2 ScanSAR data of the first time period, and LiCSAR data of the first time period. The deformation interpretation module is used to perform deformation interpretation processing on the target area of ​​the multi-source InSAR data using a preset MT-InSAR method to obtain deformation rate data; wherein, the deformation rate data includes first deformation rate data, second deformation rate data and third deformation rate data corresponding to Sentinel-1A data, ALOS-2 PALSAR-2 ScanSAR data and LiCSAR data respectively. The regional deformation feature identification module is used to perform fusion analysis on the first deformation rate data, the second deformation rate data and the third deformation rate data using a preset multi-source InSAR fusion optimization monitoring strategy to obtain regional deformation detection data for the first time period, and to perform regional boundary delineation on the regional deformation detection data to obtain regional deformation feature data. The regional morphological feature recognition module is used to input the optical remote sensing image to be identified into a landslide classifier model built based on deep learning, so as to output landslide morphological recognition data. The landslide monitoring and grading module is used to obtain landslide monitoring data based on the regional deformation characteristic data and landslide morphology identification data; wherein, the landslide monitoring data includes high-risk landslide point data and medium-risk landslide point data; Specifically, obtaining landslide monitoring data based on the regional deformation characteristic data and landslide morphology identification data involves: All regional deformation characteristic data and landslide morphology identification data are overlaid and intersected, and the overlapping areas are identified as high-risk landslide points. Data outside the overlapping area were identified as medium-risk landslide points.

9. The system according to claim 8, characterized in that, The system also includes a landslide classifier model building module, which comprises a dataset building module and a model building module, wherein... The dataset construction module is used to acquire optical remote sensing images within a first time period, construct a landslide identification sample dataset based on the optical remote sensing images, and perform data augmentation processing on the landslide identification sample dataset to obtain the landslide identification dataset. The model building module is used to train the landslide identification dataset using a preset deep learning model to obtain the landslide classifier model.

Images