Landslide disaster monitoring and early warning system based on time sequence InSAR technology

Abstract

The landslide disaster monitoring and early warning system based on the time sequence InSAR technology comprises a satellite data receiving module used for obtaining multi-temporal SAR image data; the time sequence InSAR processing module is used for carrying out phase unwrapping and deformation calculation by adopting an improved SBAS algorithm; the multi-source data fusion module is used for integrating InSAR deformation data, rainfall data, geological structure data and underground water level data; the intelligent early warning analysis module is used for establishing a deformation prediction model based on an LSTM neural network; and the early warning information issuing module is used for realizing visual output of the graded early warning signals. The system has the advantages that phase unwrapping and deformation calculation are carried out through an improved SBAS algorithm, and precise monitoring and timely early warning of landslide disasters are realized in combination with multi-source data fusion and intelligent early warning analysis.

Description

Technical Field

[0001] The present invention relates to a landslide disaster monitoring and early warning system based on time-series InSAR technology, especially for the monitoring and early warning of landslide disasters. By using time-series InSAR (Interferometric Synthetic Aperture Radar) technology and combining multi-source data fusion and intelligent analysis, accurate monitoring and timely early warning of landslide disasters can be achieved. Background Art

[0002] Landslide disasters are a common type of geological disaster, and their occurrence is often accompanied by serious casualties and property losses. Traditional landslide monitoring methods such as geological exploration and surface displacement measurement have problems such as limited monitoring range and poor real-time performance. With the continuous development of remote sensing technology, InSAR technology has shown great potential in landslide disaster monitoring due to its characteristics of large-scale, high-precision, and continuous monitoring. However, traditional InSAR technology is easily affected by factors such as atmospheric interference and surface coverage, resulting in inaccurate extraction of deformation information. Therefore, it is of great significance to develop a landslide disaster monitoring and early warning system based on time-series InSAR technology, combined with multi-source data fusion and intelligent early warning analysis. Summary of the Invention

[0003] The purpose of the present invention is to provide a landslide disaster monitoring and early warning system based on time-series InSAR technology. The system performs phase unwrapping and deformation calculation through an improved SBAS algorithm, and combines multi-source data fusion and intelligent early warning analysis to achieve accurate monitoring and timely early warning of landslide disasters.

[0004] A landslide disaster monitoring and early warning system based on time-series InSAR technology includes a satellite data receiving module for acquiring multi-temporal SAR image data; a time-series InSAR processing module that performs phase unwrapping and deformation calculation using an improved SBAS algorithm; a multi-source data fusion module that integrates InSAR deformation data, rainfall data, geological structure data, and groundwater level data; an intelligent early warning analysis module that establishes a deformation prediction model based on an LSTM neural network; and an early warning information publishing module that realizes the visual output of graded early warning signals.

[0005] Furthermore, the improved SBAS algorithm includes using an adaptive filtering algorithm to eliminate the atmospheric delay phase, specifically by constructing an atmospheric phase screen (APS) model: where P is atmospheric pressure, T is temperature, h is elevation, H is a scale parameter, and α, β, γ are regression coefficients; introducing a coherence-weighted least squares phase unwrapping method with a coherence threshold set to 0.3 - 0.5; and selecting interferometric image pairs through a spatio-temporal baseline combination optimization strategy with spatio-temporal baseline thresholds set to 150m and 60 days respectively.

[0006] Furthermore, the intelligent early warning analysis module includes: a spatial registration unit that registers the InSAR deformation field with the geological map using affine transformation, with a positioning error of less than 1 pixel; a feature extraction unit that calculates the deformation acceleration index: where is the deformation amount at time t, and Δt is the time interval; a dynamic weight allocation unit that calculates the weights of each factor based on the improved entropy method: where is the standardized value of the j-th index for the i-th sample.

[0007] Furthermore, the intelligent early warning analysis module includes: A system docking unit that interacts with the emergency management platform in real time through the REST API, and the transmission protocol complies with the ISO 22326 standard; a field verification interface that supports automatic access and comparison of drone inspection data; a feedback learning mechanism that dynamically updates the weights of the LSTM model according to false alarm / missed alarm records.

[0008] Furthermore, the satellite data receiving module includes: a GPU acceleration unit that processes the original SLC data using CUDA parallel computing; a data preprocessing unit that performs radiometric calibration and multi-look processing, with an output resolution of not less than 5m×20m; a data buffer that configures an SSD solid-state drive array with a storage capacity of ≥20TB.

[0009] Furthermore, the phase unwrapping method specifically includes: constructing a phase unwrapping mathematical model based on terrain gradient:

[0010] where: is the unwrapped phase value, is the wrapped phase observation value, is the weight factor based on the coherence coefficient, L is the terrain gradient constraint matrix, is the predicted value of the prior deformation model.

[0011] The advantages of the present invention are as follows: 1. Using the time-series InSAR technology (such as SBAS-InSAR or PS-InSAR), it can monitor the small surface deformations in the long term (with an accuracy of 1-3 mm / year), overcoming the problems of low sampling rate and high cost of traditional GPS or manual surveying and mapping, and is especially suitable for monitoring in remote mountainous areas or large-scale regions. A single satellite scan covers hundreds of square kilometers (such as Sentinel-1 data covering 250×250 km), realizing the combination of regional landslide hazard point census and key target tracking.

[0012] 2. Through the formula Quantify the atmospheric delay phase, combine meteorological data with the Digital Elevation Model (DEM) to dynamically correct errors, and reduce the impact of atmospheric noise by more than 60% (measured cases show that the deformation inversion error is reduced from ±5 mm to ±2 mm), exponential term Accurately reflect the non-linear attenuation characteristics of water vapor with height, and significantly improve the monitoring reliability in mountainous areas with large elevation differences (such as canyons and mine pits).

[0013] 3. Predict the landslide evolution trend through machine learning algorithms (such as LSTM), and the early warning time window is 7 - 15 days earlier than traditional methods. Description of the Drawings

[0014] Figure 1 It is a schematic diagram of the module collaboration relationship. Specific Embodiments

[0015] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments.

[0016] The five modules of the landslide disaster monitoring and early warning system form a complete monitoring and early warning chain through data flow and functional collaboration. The multi-temporal SAR image data obtained by the satellite data receiving module is directly transmitted to the temporal InSAR processing module, which uses an improved SBAS algorithm for phase unwrapping and deformation calculation, and outputs the deformation rate and cumulative displacement. These InSAR deformation data are then input into the multi-source data fusion module, where they are fused and analyzed with external environmental data such as rainfall, geological structure, and groundwater level. The risk assessment accuracy is improved through a weight distribution model. The fused multi-dimensional data features are sent to the intelligent early warning analysis module, which establishes a deformation prediction model based on the LSTM neural network, analyzes the spatio-temporal evolution law of landslide deformation, and predicts the future deformation trend. When the prediction result reaches the preset threshold, the early warning information release module will be triggered to generate four-level early warning signals of blue, yellow, orange, and red, and achieve visual outputs such as heat maps and displacement curves through the GIS platform. At the same time, it is connected to the government emergency system for early warning information push. The entire system forms a closed-loop process from data collection, processing, fusion, analysis to early warning release. The modules work together through data-driven and feedback optimization. Among them, the early warning effect evaluation information can be transmitted back to the temporal InSAR processing module and the intelligent early warning analysis module to optimize the SBAS algorithm parameters and the LSTM model weights, thereby continuously improving the system performance.

[0017] A landslide disaster monitoring and early warning system based on time-series InSAR technology, which integrates functional modules such as satellite data reception, time-series InSAR processing, multi-source data fusion, intelligent early warning analysis, and early warning information release, aims to achieve precise monitoring and early warning of landslide disasters and improve the efficiency of disaster emergency response.

[0018] Satellite data reception module, which is responsible for receiving multi-temporal SAR image data from satellites and providing raw data support for subsequent processing.

[0019] Adopt CUDA parallel computing technology to quickly process the received raw SLC data and improve data processing efficiency. Perform radiometric calibration and multi-look processing to ensure that the resolution of the output data is not less than 5m×20m to meet the needs of subsequent analysis. Configure an SSD solid-state drive array with a storage capacity of ≥20TB to store the preprocessed data and ensure the security and accessibility of the data.

[0020] Time-series InSAR processing module, which uses an improved SBAS algorithm to perform phase unwrapping and deformation solution on multi-temporal SAR image data to obtain surface deformation information.

[0021] Construct an atmospheric phase screen (APS) model, eliminate the influence of atmospheric delay phase through regression coefficients α, β, γ, set the coherence threshold to 0.3 - 0.5 to improve the accuracy of phase unwrapping, and set the spatio-temporal baseline thresholds to 150m and 60 days respectively to optimize the selection of interferometric image pairs and ensure the reliability of deformation solution.

[0022] Quantify the linear influence of atmospheric pressure on phase delay. The signal propagation path is longer in high-pressure areas. Correct the refractive index change caused by temperature. High temperature leads to phase advance. Simulate the non-linear attenuation of tropospheric delay with height, which is strongly correlated with the atmospheric water vapor distribution.

[0023] Multi-source data fusion module, which integrates InSAR deformation data, rainfall data, geological structure data, and groundwater level data to provide comprehensive data support for intelligent early warning analysis.

[0024] Through data interfaces and algorithm integration, fuse various types of data into a unified analysis framework to achieve complementary and collaborative analysis of multi-source data.

[0025] Intelligent early warning analysis module, which builds a deformation prediction model based on the LSTM neural network to achieve intelligent early warning analysis of landslide disasters.

[0026] The InSAR deformation field is registered with the geological map using affine transformation, with the positioning error less than 1 pixel, ensuring the accurate correspondence between deformation information and geological information. Calculate the deformation acceleration index to reflect the dynamic change characteristics of surface deformation. Calculate the weights of each factor based on the improved entropy method to ensure the accuracy and reliability of early warning analysis. Interact with the emergency management platform in real time through the REST API, and the transmission protocol complies with the ISO22326 standard to achieve the rapid transmission of early warning information. And it supports the automatic access and comparison of UAV inspection data, providing a convenient means for verifying the early warning results, and dynamically updating the LSTM model weights according to false alarm / missed alarm records to continuously optimize the early warning analysis model.

[0027] Early warning information release module This module realizes the visual output of hierarchical early warning signals, providing intuitive information support for disaster emergency response. According to the results of the intelligent early warning analysis module, the early warning information is visually displayed in the form of graphs, charts, etc., facilitating emergency management personnel to quickly grasp the disaster situation and make response decisions.

[0028] The landslide disaster monitoring and early warning system based on time-series InSAR technology of the present invention realizes the precise monitoring and early warning of landslide disasters by integrating functional modules such as satellite data reception, time-series InSAR processing, multi-source data fusion, intelligent early warning analysis, and early warning information release. This system has the advantages of high data processing efficiency, accurate and reliable early warning analysis, and intuitive and convenient early warning information release, providing strong technical support for the prevention and emergency response of landslide disasters.

[0029] Application case: In this embodiment, a mountainous area in the southwest of China is selected as the demonstration area. The geological conditions in this area are complex, and multiple landslide disasters have occurred historically. The monitored area is about 50 square kilometers, including 3 known landslide hazard points and 5 potential dangerous slopes.

[0030] Data collection stage: (1) Satellite data: Receive Sentinel-1A / B satellite data (time span: January 2024 - March 2025), and obtain a total of 56 scenes of image data (VV polarization, descending orbit data), spatial resolution: 5m×20m (range direction×azimuth direction) (2) Auxiliary data: Collect the regional geological map (scale 1:50000), access the real-time rainfall data of the meteorological bureau (time resolution: 1 hour), and obtain the displacement data of 12 GNSS monitoring stations.

[0031] Data processing stage: (1) Time-series InSAR processing: The improved SBAS algorithm is used to process the data, generating 168 groups of interferometric image pairs (spatiotemporal baseline threshold: 150 m / 60 days), and the annual average deformation rate map is calculated (accuracy: ±3 mm / year).

[0032] (2)Data fusion: The InSAR deformation field is registered with the geological map (error < 5 m), and a multi-source data correlation matrix (including 12 characteristic parameters) is established.

[0033] Early warning analysis stage: (1)Model training: The LSTM network is trained using historical landslide data (training set: data from 2018 to 2023). The number of nodes in the input layer is set to 15 (corresponding to 15 characteristic parameters), and the number of iterations is 500 times (final accuracy: 92.3%).

[0034] (2)Real-time monitoring: The system automatically generates a deformation time series curve, calculates the deformation acceleration index (threshold: 5 mm / day²), and dynamically updates the early warning level (four levels: blue / yellow / orange / red).

[0035] Typical early warning cases On January 15, 2025, the system issued an orange early warning: it was monitored that the deformation of hidden danger point No. 2 accelerated (daily displacement reached 8.7 mm). Combining with rainfall data (continuous rainfall > 50 mm for 3 days), the system issued an early warning 72 hours in advance. After on-site UAV verification, it was confirmed that cracks appeared on the slope (maximum width 12 cm). The local area promptly organized the evacuation of the masses, avoiding casualties.

[0036] Compared with GNSS data, the displacement monitoring error < ±2 mm, the early warning accuracy rate reaches 89.5% (verification period 6 months), the preprocessing time of a single-scene data < 15 minutes, the deformation solution time for the whole area < 4 hours, the cost is reduced by 60% compared with the traditional monitoring method, the emergency response time is shortened by 75%, and 3 landslide risks have been successfully warned.

[0037] The above-disclosed are only the preferred embodiments of the present invention. Of course, the scope of the rights of the present invention cannot be limited thereby. Therefore, equivalent changes made according to the claims of the present invention still fall within the scope covered by the present invention.

Claims

1. A landslide disaster monitoring and early warning system based on time-series InSAR technology, characterized in that: Including: A satellite data receiving module for acquiring multi-temporal SAR image data; A time-series InSAR processing module that uses an improved SBAS algorithm for phase unwrapping and deformation calculation; A multi-source data fusion module that integrates InSAR deformation data, rainfall data, geological structure data, and groundwater level data; An intelligent early warning analysis module that establishes a deformation prediction model based on an LSTM neural network; An early warning information publishing module that realizes the visual output of graded early warning signals.

2. The landslide disaster monitoring and early warning system based on the time series InSAR technology according to claim 1, characterized in that: The improved SBAS algorithm includes using an adaptive filtering algorithm to eliminate the atmospheric delay phase, specifically by constructing an atmospheric phase screen (APS) model: , where P is atmospheric pressure, T is temperature, h is elevation, H is a scale parameter, and α, β, γ are regression coefficients; introducing a coherence-weighted least squares phase unwrapping method with a coherence threshold set at 0.3 - 0.5; selecting interferometric image pairs through a spatio-temporal baseline combination optimization strategy with spatio-temporal baseline thresholds set at 150 m and 60 days respectively.

3. The landslide disaster monitoring and early warning system based on the time-series InSAR technology according to claim 2, wherein: The intelligent early warning analysis module includes: A spatial registration unit that registers the InSAR deformation field with a geological map using affine transformation, with a positioning error < 1 pixel; A feature extraction unit that calculates deformation acceleration indicators: , Among them is the deformation amount at time t, and Δt is the time interval; A dynamic weight allocation unit that calculates the weights of each factor based on an improved entropy method: , where , is the standardized value of the i-th sample for the j-th indicator.

4. The landslide disaster monitoring and early warning system based on the time-series InSAR technology according to claim 3, characterized in that: The intelligent early warning analysis module includes: A system docking unit that interacts with the emergency management platform in real time through REST API, and the transmission protocol complies with the ISO 22326 standard; A field verification interface that supports the automatic access and comparison of UAV inspection data; A feedback learning mechanism that dynamically updates the weights of the LSTM model according to false alarm / missed alarm records.

5. The landslide disaster monitoring and early warning system based on the time-series InSAR technology according to claim 4, characterized in that: The satellite data receiving module includes: A GPU acceleration unit that uses CUDA parallel computing to process raw SLC data; A data preprocessing unit that performs radiometric calibration and multi-look processing, with an output resolution of not less than 5 m × 20 m; A data buffer configured with an SSD solid-state drive array and a storage capacity of ≥ 20 TB.

6. The landslide disaster monitoring and early warning system based on the time-series InSAR technology according to claim 5, characterized in that: The phase unwrapping method specifically includes: Constructing a phase unwrapping mathematical model based on terrain gradient: , Wherein: is the unwrapped phase value, is the wrapped phase observation value, is the weight factor based on the coherence coefficient, and L is the topographic gradient constraint matrix, is the predicted value of the prior deformation model.

Images