InSAR (Interferometric Synthetic Aperture Radar) watershed slope displacement monitoring method and device

Abstract

The invention relates to the technical field of slope monitoring, in particular to an InSAR (Interferometric Synthetic Aperture Radar) watershed slope displacement monitoring method and device. According to the technical scheme, the method comprises the following steps: S1, acquiring multi-temporal SAR image data of a watershed area through a satellite; s2, preprocessing the SAR image data to obtain an initial interferogram; s3, de-noising processing is carried out on the interferogram; s4, eliminating terrain errors; s5, real displacement information is obtained; s6, carrying out phase calibration; s7, carrying out continuous space-time displacement monitoring on the side slope of the monitoring area; s8, constructing a displacement prediction model based on machine learning to predict the future displacement trend of the slope; s9, forming a displacement risk assessment report in combination with the slope displacement space-time distribution diagram and historical data; and S10, according to the result, sending early warning information to a related disaster prevention and reduction mechanism. According to the invention, the precision, real-time performance and effectiveness of slope displacement monitoring are comprehensively improved, and powerful technical support is provided for early warning and protection of geological disasters.

Description

Technical Field

[0001] The invention relates to the technical field of slope monitoring, and in particular to an InSAR watershed slope displacement monitoring method and device. Background Art

[0002] Basin slope displacement monitoring refers to the technical means of real-time monitoring and evaluation of the displacement changes of the slopes in the basin (usually the slopes on both sides of water bodies such as rivers and lakes). This monitoring is of great significance for ensuring the safety of water conservancy projects in the basin, preventing geological disasters, and protecting the lives and property of people along the coast.

[0003] After searching, the patent with patent announcement number CN201910686984.6 discloses a close-range photogrammetry monitoring method for river basin reservoir bank deformation based on a ground stable platform. Although the device combines the close-range photogrammetry system with GPS technology when in use, and applies ordinary digital cameras to the deformation monitoring of river basin reservoir banks through the principle of close-range photogrammetry, it expands the application scope of photogrammetry, while improving the efficiency and safety of slope monitoring, and realizes the automation and informatization of river basin reservoir bank monitoring. However, the device is limited by the equipment position and shooting angle when in use, and it is difficult to cover large areas or complex terrain river basin reservoir banks, and the accuracy may decrease at long distances or under complex conditions, especially when monitoring the early tiny displacements of geological disasters. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention provides a method and device for monitoring watershed slope displacement using InSAR, which solves the problems raised in the background technology.

[0005] The solution of the present invention to solve the above technical problems is as follows:

[0006] An InSAR watershed slope displacement monitoring method includes the following steps:

[0007] S1. Use the SAR data acquisition module to collect multi-temporal SAR images of the target watershed area to ensure continuous and regular data acquisition and provide time series information for subsequent monitoring;

[0008] S2, the data preprocessing module receives the collected SAR image data, generates the initial interferogram using InSAR technology, and identifies noise and errors during the image generation process;

[0009] S3. In the data preprocessing module, an adaptive noise suppression algorithm is added to effectively remove the noise in the interference pattern through time-frequency domain filtering, which greatly reduces the impact of atmospheric delay, orbit error and terrain error on monitoring accuracy;

[0010] S4, the data preprocessing module further uses a high-precision digital elevation model to correct the interference phase, eliminate the phase error caused by the terrain height difference, and ensure the accuracy of the phase information;

[0011] S5, the phase unwrapping module receives the corrected interferometric phase data, and uses a phase unwrapping algorithm to convert the complex phase value into continuous displacement information to restore the actual surface displacement;

[0012] S6. The phase unwrapping module calibrates the unwrapped displacement information in combination with the ground reference point to ensure the reliability and accuracy of the calculation results;

[0013] S7, displacement analysis module uses distributed point target analysis method to conduct in-depth analysis on the calibrated displacement data, generate the spatiotemporal displacement distribution of the slope in the monitoring area, and provide accurate displacement evolution information;

[0014] S8, displacement prediction module combines historical slope displacement data and uses machine learning models for prediction;

[0015] S9, the early warning system module receives data from the displacement analysis and prediction module, generates a spatiotemporal distribution map of slope displacement, and forms a displacement risk assessment report in combination with historical displacement data for safety assessment of the monitoring area;

[0016] S10. The early warning system module automatically sends out early warning signals to relevant disaster prevention and mitigation agencies based on the displacement trend prediction results and risk assessment reports, and links external systems in real time to ensure that protective measures are taken in a timely manner.

[0017] Based on the above technical solution, the present invention can also be improved as follows.

[0018] Furthermore, the adaptive noise suppression algorithm used in S3 is a time-frequency domain filtering algorithm based on wavelet transform, which is used to effectively reduce noise interference.

[0019] The beneficial effects of adopting the above further scheme are:

[0020] Wavelet transform can provide localized information in both time and frequency domains, which is not available in traditional Fourier transform. It is particularly effective for analyzing non-stationary signals (such as signals containing noise) because the statistical characteristics of non-stationary signals change over time. This feature of wavelet transform enables it to effectively separate signals and noise at different time and frequency resolutions. Wavelet transform analyzes signals at different scales (or resolutions) by gradually subdividing them. This allows it to more flexibly match the characteristics of the signal and remove noise based on the performance of the signal at different scales. For example, at high-frequency scales, noise may be more obvious, while at low-frequency scales, the characteristics of the signal are more prominent. Combining wavelet transform and adaptive filtering algorithm, adaptive suppression of noise can be achieved. This means that the parameters of the filter can be dynamically adjusted according to the characteristics of the signal and noise to achieve the best filtering effect. For example, when the noise level is high, the filter can strengthen the denoising strength; when the signal quality is good, the denoising can be appropriately weakened to retain more signal details. The time-frequency domain filtering algorithm based on wavelet transform can more effectively remove noise, especially those noises that have large differences in characteristics with the signal in the time-frequency domain. Through wavelet decomposition, signals and noise can be separated at different scales, and noise can be removed by threshold processing and other methods. This method is more sophisticated and effective than traditional global filtering methods. While removing noise, the filtering algorithm based on wavelet transform can retain as much signal detail information as possible. This is because wavelet transform can decompose signals at different scales, so that the various components of the signal are better preserved at their respective scales. Wavelet transform provides a variety of wavelet basis functions to choose from, which have different properties and characteristics. According to the specific situation of the signal and the denoising requirements, the most suitable wavelet basis function can be selected for filtering. In addition, the filtering effect can be optimized by adjusting the number of layers and thresholds of wavelet decomposition.

[0021] Furthermore, the ground reference points used in S6 are GPS monitoring points and known stable buildings deployed in the watershed to improve the phase calibration accuracy.

[0022] The beneficial effects of adopting the above further scheme are:

[0023] GPS monitoring points can provide high-precision location information. As a known stable reference, this location information can greatly improve the accuracy of phase unwrapping and displacement prediction. By calibrating the phase information regularly or irregularly, the error accumulation caused by long-term observation or environmental changes can be reduced, ensuring the accuracy and reliability of the monitoring data. Known stable buildings are used as reference points, and their positions remain relatively stable over a long period of time, providing a reliable reference for phase calibration. Compared with relying solely on satellite signals for phase unwrapping, combining ground reference points can reduce the impact of external factors such as atmospheric delay and multipath effects on phase information.

[0024] Furthermore, the machine learning model in S8 is a nonlinear displacement prediction model based on long short-term memory network.

[0025] The beneficial effects of adopting the above further scheme are:

[0026] LSTM can effectively preserve long-term memory through its unique gating structure (forget gate, input gate, output gate). This ability enables the LSTM model to perform well when processing sequence data with long-term dependencies, and can capture relevant information with a large span in the data, thereby more accurately predicting future trends in nonlinear displacement prediction. Traditional recurrent neural networks (RNNs) are prone to gradient vanishing or gradient explosion problems when processing long sequence data, making the model difficult to train. LSTM, through its gating mechanism, effectively alleviates this problem, making the model more stable during training and able to learn more complex data features. The network structure of LSTM allows information to flow flexibly in the network, which provides a high degree of flexibility. At the same time, the LSTM model has strong generalization ability and can be applied to sequence prediction tasks in different fields, including but not limited to nonlinear displacement prediction. Compared with traditional time series prediction methods (such as ARIMA or exponential smoothing), LSTM has stronger modeling capabilities. It can capture complex nonlinear relationships and adapt to nonlinear changes in data, thereby providing more accurate prediction results in nonlinear displacement prediction. The gating structure of LSTM enables it to adapt to sequence data of different lengths. Since LSTM can capture long-term dependencies and prevent gradient problems, its prediction accuracy is usually high when it is well trained.

[0027] Furthermore, the SAR image data is acquired once a month or more frequently to achieve high-efficiency displacement monitoring.

[0028] The beneficial effects of adopting the above further scheme are:

[0029] High-frequency SAR image data acquisition ensures real-time monitoring. In scenarios such as geological disasters and slope stability assessment, it can quickly capture changes in surface displacement and provide timely information support for decision makers. Once abnormal displacement or potential risks are found, high-frequency data can immediately trigger an early warning mechanism, enabling relevant departments to take prompt action to reduce disaster losses. High-frequency data acquisition provides a continuous record of surface displacement observations, which helps to more accurately analyze displacement trends and rates. Through multiple observations and data analysis, random errors can be averaged to improve the accuracy and reliability of monitoring results. High-frequency data constitutes a complete time series, which is convenient for time series analysis, such as trend analysis, periodicity analysis, mutation point detection, etc., so as to gain a deeper understanding of the laws and characteristics of surface displacement. High-frequency data supports monitoring and analysis at different time scales, such as day, week, month, year, etc., to meet different research and application needs.

[0030] An InSAR basin slope displacement monitoring device comprises a vehicle body, on which a SAR data acquisition module is rotatably mounted via a rotating base;

[0031] A data preprocessing module and a phase unwrapping module are installed on one side of the SAR data acquisition module on the vehicle body, and a displacement analysis module and a displacement prediction module are installed on one side of the data preprocessing module and the phase unwrapping module on the vehicle body. The SAR data acquisition module is located on the center line of the vehicle body, and the data preprocessing module and the phase unwrapping module are a group, and the displacement prediction module and the rotating base are a group and are symmetrically distributed on both sides of the center line of the vehicle body;

[0032] SAR data acquisition module: contains high-resolution radar imaging devices, supports the acquisition of SAR image data in different polarization modes, ensures the diversity and richness of data sources, and the collected SAR image data will be passed to the data preprocessing module for further processing;

[0033] Data preprocessing module: including noise suppression unit and terrain correction unit, which are used to perform preliminary denoising and correction on SAR data, reduce noise and terrain errors in the interference pattern, and improve data quality. The preprocessed data is output to the phase unwrapping module to generate displacement information;

[0034] Phase unwrapping module: unwraps the preprocessed interference phase to extract the real slope displacement information to restore the original surface displacement information. The unwrapped displacement information will be passed to the displacement analysis module for subsequent spatiotemporal displacement analysis;

[0035] Displacement analysis module: Based on the distributed point target analysis method, the disentangled displacement information is analyzed in time and space, and the time and space displacement distribution map of the slope in the monitoring area is generated to realize the continuous monitoring of the slope displacement. The analyzed time and space displacement data will be transmitted to the displacement prediction module to predict the future displacement trend;

[0036] Displacement prediction module: Based on the long short-term memory network model, it performs nonlinear prediction of slope displacement data, evaluates future displacement trends, and improves prediction accuracy through continuous iterative training. The prediction results will be transmitted to the external early warning system module to generate an early warning report;

[0037] Early warning system module: The early warning system module integrates the data from the displacement prediction module to generate a detailed displacement risk assessment report and issues a warning signal based on the predicted risk level. This module is linked with external disaster prevention and mitigation agencies to ensure real-time transmission of early warning information. The early warning system module not only generates reports, but also transmits real-time displacement monitoring data back to the data storage system of the external system for further analysis or model updates.

[0038] Furthermore, the SAR data acquisition module includes a high-resolution SAR data acquisition module imaging device, which supports the acquisition of SAR image data in different polarization modes.

[0039] The beneficial effects of adopting the above further scheme are:

[0040] The high-resolution SAR data acquisition module can generate more detailed SAR images, and finely depict various features of the surface, including terrain undulations, details of objects, etc., providing a more accurate data basis for subsequent image interpretation and object classification. It supports SAR data acquisition in multiple polarization modes such as HH, VV, HV, and VH, and can obtain object scattering information under different polarization modes. These polarization information reflects the different physical properties and scattering characteristics of objects, and provides a richer data source for object classification, target recognition, and scene interpretation. Through the joint analysis of multi-polarization data, different types of objects can be more effectively distinguished, and the accuracy of object classification and target recognition can be improved. For example, HH polarization is more sensitive to dihedral scattering, which is suitable for detecting tall vegetation and man-made objects; VV polarization is more sensitive to rough surface scattering, which is suitable for detecting bare land, etc.; HV and VH polarization are more sensitive to parameters such as water content and humidity of vegetation and soil.

[0041] Furthermore, the data preprocessing module includes a noise suppression unit and a terrain correction unit, which are respectively used to reduce noise and terrain errors in the interference pattern.

[0042] The beneficial effects of adopting the above further scheme are:

[0043] The noise suppression unit can significantly improve the quality of data by removing or reducing the noise in the interference pattern. Noise is a random fluctuation or error in the data, which may mask the real signal and features. Through noise suppression, the real information in the data can be restored. In the subsequent data analysis, processing or modeling process, high-quality data can reduce the risk of overfitting of the model and improve the accuracy and generalization ability of the model. Noise suppression helps the model focus more on the real data features, so as to make more accurate predictions or classifications. Noise suppression can also improve the clarity of the interference pattern, making the ground features in the image more distinct, which is conducive to subsequent image interpretation and target recognition.

[0044] Terrain undulations can cause geometric distortion and phase errors in SAR images, affecting the accuracy and reliability of the data. The terrain correction unit can restore the real information in the data and improve the accuracy of the data by calculating and removing these terrain effects. In large-scale SAR image monitoring, the complexity of terrain undulations may lead to inconsistencies in data from different regions. Terrain correction can uniformly correct these inconsistencies, making the data of the entire monitoring area more consistent and comparable. The corrected data is closer to the actual surface conditions and can support more accurate geospatial analysis and applications. For example, in the fields of disaster monitoring, topographic mapping, urban planning, etc., terrain-corrected data can provide more accurate surface information support.

[0045] Furthermore, the machine learning algorithm built into the displacement prediction module has an automatic update function, which improves the prediction accuracy through continuous iterative training.

[0046] The beneficial effects of adopting the above further scheme are:

[0047] Machine learning algorithms can receive new data inputs in real time and automatically adjust and optimize model parameters based on this data. This dynamic adjustment capability allows the algorithm to adapt to environmental changes and new data features in real time, thereby improving the accuracy and reliability of predictions. As the number of iterations increases, the algorithm can learn more data features and patterns, thereby continuously improving the accuracy of predictions. This continuous optimization process ensures that the prediction results are closer to the actual situation. The automatic update function automates the iterative training process of machine learning algorithms and reduces the need for human intervention. This not only improves work efficiency, but also reduces the risk of human error. Through automated updates, the algorithm can continue to maintain its predictive performance, reducing the cost of frequent retraining and maintenance due to outdated models or performance degradation.

[0048] Furthermore, the early warning system module is connected to the river basin disaster prevention and mitigation platform and can transmit monitoring data and early warning information in real time.

[0049] The beneficial effects of adopting the above further scheme are:

[0050] After networking, the early warning system can instantly transmit the monitored data to the basin disaster prevention and mitigation platform to achieve real-time data sharing. This helps relevant departments quickly grasp the real-time situation in the basin and provide a basis for timely decision-making. The transmission of real-time data enables early warning information to be generated and disseminated quickly, which helps to take immediate measures before or when a disaster occurs to reduce disaster losses. Through networking, the basin disaster prevention and mitigation platform can integrate and analyze data from multiple early warning systems to improve the accuracy and efficiency of early warning. This helps to identify potential disaster risks and take preventive measures in advance. The networked early warning system can use the extensive communication channels of the basin disaster prevention and mitigation platform to quickly and accurately release early warning information to relevant departments and the public, thereby improving the coverage and reception rate of early warning information.

[0051] The present invention provides a method and device for monitoring watershed slope displacement using InSAR, which has the following beneficial effects:

[0052] 1. By introducing a time-frequency domain filtering algorithm based on wavelet transform, the common noise problems in InSAR technology are innovatively solved, especially the impact of atmospheric delay, orbit error and terrain effect on monitoring accuracy. This method enhances the noise suppression effect and improves the accuracy of displacement monitoring by adaptively adjusting the filtering parameters.

[0053] 2. The invention creatively combines the long short-term memory network (LSTM) to predict the future displacement trend of the slope. This prediction model based on historical displacement data can effectively handle the nonlinear changes of slope displacement and provide early warning support for disaster prevention and mitigation. It is an innovative application that has never been combined with traditional InSAR technology.

[0054] 3. By innovatively applying the distributed point target analysis method (DPS) to the spatiotemporal displacement monitoring of slopes, the limitations of traditional InSAR technology in point target monitoring are effectively overcome, making it possible to conduct large-scale, detailed and continuous monitoring of slopes in the entire area.

[0055] 4. A real-time early warning system is introduced into the device, which can send early warning information to relevant disaster prevention and mitigation agencies based on real-time monitoring data and machine learning prediction results. This innovative design enables the system to not only have monitoring functions, but also proactively warn, which enhances the application value of the monitoring system.

[0056] 5. The machine learning model in the displacement prediction module has an automatic update function, which can continuously iterate the model training according to the new monitoring data during operation. This design makes the prediction accuracy gradually improve over time, enhancing the intelligence and adaptability of the system.

[0057] 6. Innovative use of GPS monitoring points and known stable buildings as reference points, combined with InSAR technology for phase calibration, ensures phase accuracy in large-scale slope monitoring and reduces the risk of error accumulation. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] The drawings described herein are used to provide further understanding of the present invention and constitute a part of this application. The illustrative embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0059] In the attached picture:

[0060] Figure 1 It is a schematic diagram of the module processing flow of the present invention;

[0061] Figure 2 It is a schematic diagram of the workflow of the present invention;

[0062] Figure 3 It is a schematic diagram of the appearance of the present invention.

[0063] In the accompanying drawings, the components represented by the reference numerals are listed as follows:

[0064] 1. SAR data acquisition module; 2. Data preprocessing module; 3. Phase unwrapping module; 4. Displacement analysis module; 5. Displacement prediction module; 6. Rotating base; 7. Vehicle body. DETAILED DESCRIPTION

[0065] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0066] See also Figures 1 to 3 As shown,

[0067] Embodiment 1

[0068] The present invention provides a technical solution for a method for monitoring watershed slope displacement using InSAR:

[0069] An InSAR watershed slope displacement monitoring method includes the following steps:

[0070] S1. Use the SAR data acquisition module to collect multi-temporal SAR images of the target watershed area to ensure continuous and regular data acquisition. These data should have a certain overlapping area to facilitate subsequent interference processing and provide time series information for subsequent monitoring;

[0071] S2, the data preprocessing module receives the collected SAR image data, generates the initial interference pattern using InSAR technology, and identifies noise and errors during the image generation process. If the SAR image data is stored in a non-standard format or a format not supported by the processing software, it needs to be converted to a supported format, such as GeoTIFF, HDF, etc. The acquisition frequency of SAR image data is once a month or more to achieve high-time displacement monitoring. The continuous acquisition of high-frequency SAR image data ensures the real-time response capability of surface displacement monitoring. In emergency application scenarios such as geological disaster warning and slope stability assessment, high-frequency data can quickly capture small surface displacement changes and provide decision makers with immediate and critical information support. Once abnormal displacement signs or potential risk signals are detected, high-frequency data can immediately trigger the early warning system, prompting relevant departments to respond quickly and effectively reduce the impact and losses of potential disasters. High-frequency data acquisition not only provides a continuous and detailed time series record of surface displacement, but also significantly improves the accuracy and reliability of monitoring results through data fusion and error averaging of multiple observations. This complete time series data set further supports in-depth time series analysis, including trend analysis, periodicity identification, mutation point detection, etc., providing strong data support for in-depth understanding of the dynamic laws and characteristics of surface displacement. At the same time, high-frequency data supports monitoring and analysis needs across multiple time scales, such as days, weeks, months, and years, meeting the diverse needs of different scientific research and application scenarios.

[0072] S3. In the data preprocessing module, an adaptive noise suppression algorithm is added to effectively remove the noise in the interference pattern through time-frequency domain filtering, which greatly reduces the impact of atmospheric delay, orbit error and terrain error on monitoring accuracy. The adaptive noise suppression algorithm is a time-frequency domain filtering algorithm based on wavelet transform, which is used to effectively reduce noise interference.

[0073] Wavelet transform, with its unique ability, can realize localized analysis of information in both time domain and frequency domain, which is significantly different from the traditional Fourier transform, which has limitations in this respect. Wavelet transform shows extraordinary performance for non-stationary signals, especially those whose statistical characteristics change significantly over time (including noisy signals). It can accurately separate signal and noise by flexibly adjusting the resolution of time and frequency. This process benefits from its layer-by-layer analysis strategy, which can accurately capture the dynamic characteristics of signals at different scales (resolutions).

[0074] In the high frequency band, noise is often more significant, and the wavelet transform can identify and effectively suppress these high-frequency noises; on the contrary, in the low frequency band, the main features of the signal are clearer and preserved. Combined with the adaptive filtering algorithm, the wavelet transform achieves dynamic suppression of noise, and the filtering parameters can be intelligently adjusted according to the actual characteristics of the signal and noise, thereby enhancing the denoising effect when the noise is severe, and reducing interference when the signal quality is good, retaining more details.

[0075] The time-frequency domain filtering algorithm based on wavelet transform can effectively remove noise that is significantly different from the signal characteristics by virtue of its precise time-frequency analysis capability. Through wavelet decomposition, the signal and noise are cleverly separated at different scales, and then the noise is effectively removed by means of threshold processing, etc. This process is more sophisticated and efficient than the traditional global filtering method. At the same time, while ensuring the removal of noise, the algorithm retains the original details of the signal to the greatest extent possible, because the multi-scale analysis strategy of wavelet transform allows the different components of the signal to be properly preserved at their respective appropriate scales.

[0076] In addition, the flexibility of wavelet transform is also reflected in its rich selection of wavelet basis functions. These basis functions have their own characteristics and can be optimally matched according to the specific characteristics of the signal and the denoising requirements. Furthermore, by fine-tuning the level and threshold of wavelet decomposition parameters, the filtering effect can be further optimized to achieve more accurate signal denoising.

[0077] S4, the data preprocessing module further uses a high-precision digital elevation model to correct the interference phase, eliminate the phase error caused by the terrain height difference, and ensure the accuracy of the phase information;

[0078] S5, the phase unwrapping module receives the corrected interferometric phase data, and uses a phase unwrapping algorithm to convert the complex phase value into continuous displacement information to restore the actual surface displacement;

[0079] The interferometric phase image is unwrapped by a phase unwrapping algorithm. The unwrapping process usually involves starting from a certain point (or region) in the phase image and gradually calculating and restoring the true phase value of each pixel along a certain path. In the unwrapping process, it is necessary to deal with problems such as phase discontinuity (such as caused by noise, occlusion, etc.) and phase ambiguity (such as integer ambiguity). Phase unwrapping algorithms include minimum cost flow algorithm, branch cutting method, quality graph guided method (such as Goldstein algorithm, Snaphu algorithm, etc.) and network optimization-based methods.

[0080] The unwrapped phase image is converted into surface displacement information. This usually involves converting the phase difference into distance change (using parameters such as radar wavelength and angle of incidence), and taking into account the effects of other factors such as the curvature of the earth and atmospheric delay.

[0081] S6. The phase unwrapping module calibrates the unwrapped displacement information in combination with ground reference points to ensure the reliability and accuracy of the calculation results. The ground reference points are GPS monitoring points and known stable buildings deployed in the watershed, which are used to improve the phase calibration accuracy. The GPS monitoring stations can output high-precision geographic location coordinates. These coordinates serve as fixed stable reference benchmarks, which significantly improve the accuracy of the phase unwrapping process and displacement prediction. By implementing regular or adaptive calibration strategies and fine-tuning the phase information, the error accumulation effect induced by long-term observations or dynamic environmental changes can be effectively curbed, ensuring the accuracy and high reliability of the monitoring data. In addition, buildings with known stable structures are used as ground reference points, and their position stability is excellent in long-term observations, laying a solid benchmark foundation for phase calibration. Compared with the phase unwrapping method that relies solely on satellite signals, the strategy combined with ground reference points can significantly reduce the impact of external interference factors such as atmospheric delay and multipath effects on phase information;

[0082] S7, displacement analysis module uses distributed point target analysis method to conduct in-depth analysis on the calibrated displacement data, generate the spatiotemporal displacement distribution of the slope in the monitoring area, and provide accurate displacement evolution information;

[0083] S8, displacement prediction module combines historical slope displacement data and uses machine learning model for prediction. The machine learning model is a nonlinear displacement prediction model based on long short-term memory network. The core is intelligent prediction based on time series data. Long short-term memory network (LSTM) achieves effective memory and retention of long-term dependent information with its unique gating mechanism architecture - forget gate, input gate and output gate. This feature enables LSTM to show excellent performance when processing sequence data containing long-term correlation, and can capture relevant information across longer intervals in the data, thereby achieving high-precision prediction of future change trends in nonlinear displacement prediction tasks. In contrast, traditional recurrent neural networks (RNNs) are prone to gradient vanishing or explosion problems when processing long sequence data, which limits their training efficiency. LSTM, through its sophisticated gating design, effectively alleviates the above problems, ensures the stability and efficiency of the model training process, and enables it to learn more complex and deep data features. The flexibility of the LSTM network architecture gives it high adaptability in the information processing process, while showing strong generalization ability, and can be widely used in multi-field sequence prediction tasks including but not limited to nonlinear displacement prediction. Compared with traditional time series analysis methods (such as ARIMA or exponential smoothing models), LSTM, with its powerful modeling capabilities, can more accurately capture the nonlinear relationships and dynamic changes in the data, and provide more accurate prediction results for nonlinear displacement prediction. In addition, the gating structure of LSTM also gives it the flexibility to process sequence data of different lengths. Combined with sufficient training, it can usually achieve higher prediction accuracy;

[0084] S9, the early warning system module receives data from the displacement analysis and prediction module, generates a spatiotemporal distribution map of slope displacement, and forms a displacement risk assessment report in combination with historical displacement data for safety assessment of the monitoring area;

[0085] S10. The early warning system module automatically sends out early warning signals to relevant disaster prevention and mitigation agencies based on the displacement trend prediction results and risk assessment reports, and links external systems in real time to ensure that protective measures are taken in a timely manner.

[0086] Embodiment 2

[0087] In order to generate an initial interference pattern, the present invention exemplarily further comprises:

[0088] The data preprocessing module uses InSAR technology processing method:

[0089] Radiometric calibration of SAR images is performed to convert the original SAR data into physically meaningful parameters such as backscatter coefficient or brightness temperature, and the influence of the atmosphere on the SAR signal, including atmospheric attenuation and scattering, is removed to improve the accuracy of the data, correct the orbit error of the SAR image, ensure the geographic positioning accuracy of the image, and use appropriate filtering methods to remove noise in the image, such as speckle noise. Common filtering methods include mean filtering, median filtering, Gaussian filtering, etc.

[0090] Geometric correction is performed on SAR images at different time points so that they are aligned to the same geographic coordinate system. This is usually achieved by using ground control points or feature point matching methods. The registered images can ensure the accuracy of subsequent interferometric processing.

[0091] The registered SAR images are used for interferometric processing to calculate the phase difference between two time points. This phase difference contains information such as surface deformation and elevation change; the phase difference is converted into an interferogram. An interferogram is an image that shows phase differences and contains information about surface deformation and elevation change.

[0092] Process the interferometric phase and unwrap the redundant 2π phase to obtain accurate values ​​of surface deformation and elevation change. Perform error analysis on the interferogram and phase unwrapping results to evaluate data quality and reliability. This includes analyzing problems such as noise levels and phase discontinuities, and taking appropriate measures to correct or improve them.

[0093] The output is the initial interferogram containing noise and errors.

[0094] Embodiment 3

[0095] In order to remove noise in the interference pattern, illustratively, the present invention further includes:

[0096] Steps to remove noise from the interference pattern:

[0097] Load the SAR image data, make sure the data format is correct, and have been subjected to basic radiometric calibration and atmospheric correction. If necessary, perform preliminary denoising on the data, such as using a simple spatial domain filtering method to reduce obvious noise points.

[0098] Convert the interference pattern from the spatial domain to the wavelet domain. This step is to decompose the image into sub-bands of different frequencies and scales, where the high-frequency sub-bands contain the details and noise of the image, and the low-frequency sub-bands contain the smooth part of the image. Select the appropriate wavelet basis and the number of decomposition layers. The choice of wavelet basis affects the decomposition effect. Common wavelet functions include Haar, Daubechies, Symlets, etc. The number of decomposition layers is determined according to the image characteristics and noise level. The wavelet transform not only has time domain characteristics (can reflect the changes of signals or images in time or space), but also has frequency domain characteristics (can reflect the changes of signals or images in frequency). This gives the wavelet transform a unique advantage in processing signals or images with complex time-frequency characteristics.

[0099] In the wavelet domain, each sub-band is thresholded to remove noise. The choice of threshold is critical. Too high a threshold will lead to loss of image details, while too low a threshold will not effectively remove noise. The adaptive threshold algorithm can dynamically adjust the threshold according to the noise level and image characteristics of the sub-band, such as setting the threshold based on statistical characteristics such as the standard deviation and median of the sub-band coefficients.

[0100] The wavelet coefficients after threshold processing are inversely transformed from the wavelet domain back to the spatial domain to obtain the denoised interference pattern.

[0101] The quality of the denoised interference pattern is evaluated to check whether the noise is effectively removed and the image details are retained. According to the evaluation results, the parameters of the wavelet transform (such as wavelet basis, number of decomposition layers, threshold, etc.) are adjusted to achieve the best denoising effect.

[0102] Embodiment 4

[0103] In order to eliminate the phase error caused by the terrain height difference and ensure the accuracy of the phase information, the present invention also includes:

[0104] Processing method using high-precision digital elevation model:

[0105] Select DEM data that is close in time to the SAR image data, has a matching spatial resolution, and is sufficiently accurate. The accuracy of the DEM directly affects the effect of terrain phase correction. Obtain the DEM data and ensure that its coordinate system is consistent with the SAR image data for subsequent processing.

[0106] The phase delay caused by terrain is calculated using DEM data and radar parameters (such as wavelength, incident angle, etc.). This step is usually achieved by simulating the propagation path of the radar beam on the terrain and calculating the corresponding phase delay.

[0107] The calculated terrain phase is subtracted from the original interferometric phase to eliminate the phase error caused by terrain height difference. This step needs to be performed point by point throughout the interferogram to ensure the accuracy and consistency of the correction.

[0108] Perform a quality assessment on the corrected interferometric phase to check whether the terrain phase error has been effectively removed. Adjust the DEM accuracy, radar parameters or correction algorithm based on the assessment results to obtain better correction results.

[0109] Embodiment 5

[0110] In order to calibrate the unwrapped displacement information, illustratively, the present invention further includes:

[0111] Calibration steps:

[0112] Select points that remain relatively stable during surface deformation monitoring as reference points, such as rock outcrops, stable building foundations, etc.; the reference points should be evenly distributed within the surface deformation area to fully reflect the surface deformation; determine the appropriate number of reference points based on the size and deformation characteristics of the monitoring area.

[0113] Use high-precision surveying equipment (such as GPS, level, etc.) to conduct field measurements of ground reference points to obtain their precise coordinates and elevation information, and record the measurement data of each reference point in detail, including measurement time, coordinates, elevation, etc.

[0114] Extract the displacement information of the corresponding ground reference point from the unwrapped phase map, compare and analyze the extracted displacement information with the actual measurement data of the ground reference point, calculate the difference between the two, and perform error correction on the unwrapped displacement information based on the results of the comparative analysis. If the displacement error of a reference point is large, consider reselecting the reference point or adjusting the unwrapping algorithm parameters. On the basis of local calibration, perform global calibration on the displacement information of the entire monitoring area to ensure the accuracy and consistency of the displacement information.

[0115] Embodiment 6

[0116] In order to combine historical slope displacement data and use a machine learning model for prediction, the present invention also includes, illustratively:

[0117] Machine learning model prediction steps:

[0118] 1. Data preparation:

[0119] Collect data: First, it is necessary to collect historical displacement data of the slope, which usually exists in the form of time series, such as daily, hourly or minute-by-minute displacement observations.

[0120] Data preprocessing: including data cleaning (removing outliers, processing missing values, etc.), feature selection (selecting features that are useful for prediction, such as timestamps, temperature, rainfall, and other factors that may affect slope displacement), data normalization or standardization, etc.

[0121] Divide the data set: Divide the preprocessed data set into training set, validation set and test set. The training set is used to train the model, the validation set is used to adjust the model parameters (such as hyperparameter tuning), and the test set is used to evaluate the performance of the model.

[0122] 2. Build LSTM model:

[0123] Design the network structure: determine the number of LSTM layers, the number of neurons in each layer, whether to use regularization techniques such as dropout to prevent overfitting, and whether to add a fully connected layer (Dense layer) to combine features and output prediction values.

[0124] Define loss function and optimizer: Choose an appropriate loss function (such as mean squared error MSE) and optimizer (such as Adam optimizer) to train the model.

[0125] Compile model: In frameworks such as Keras or TensorFlow, use the compile method to compile the model and specify the loss function, optimizer, and evaluation metrics.

[0126] 3. Training model:

[0127] Training process: Use the training set data to train the LSTM model. During the training process, the model will learn its inherent patterns and rules based on the input displacement time series data.

[0128] Validation and adjustment: By monitoring the loss value on the validation set, the model parameters such as learning rate, batch size, number of iterations, etc. are adjusted to achieve better prediction results.

[0129] 4. Model evaluation:

[0130] Test set evaluation: Use the test set to evaluate the performance of the trained LSTM model. Evaluation indicators can include the mean square error (MSE), root mean square error (RMSE), mean absolute error (MAE), etc. of the prediction.

[0131] Visualization: Compare the predicted results with the actual observed values, and visualize the predicted effects through charts, etc., so as to intuitively understand the predictive ability of the model

[0132] 5. Prediction and application:

[0133] New data prediction: The new slope displacement time series data is input into the trained LSTM model for displacement prediction.

[0134] Decision support: Based on the prediction results, provide decision support for slope stability assessment, disaster warning, etc.

[0135] Embodiment 7

[0136] The present invention provides a technical solution of an InSAR watershed slope displacement monitoring device:

[0137] An InSAR watershed slope displacement monitoring device comprises a vehicle body, a SAR data acquisition module is rotatably mounted on the vehicle body via a rotating base, a data preprocessing module and a phase unwrapping module are mounted on one side of the SAR data acquisition module on the vehicle body, a displacement analysis module and a displacement prediction module are mounted on one side of the data preprocessing module and the phase unwrapping module on the vehicle body, the SAR data acquisition module is located on the center line of the vehicle body, the data preprocessing module and the phase unwrapping module form a group, and the displacement prediction module and the rotating base form a group and are symmetrically distributed on both sides of the center line of the vehicle body;

[0138] SAR data acquisition module: It includes high-resolution radar imaging devices, supports the acquisition of SAR image data in different polarization modes, ensures the diversity and richness of data sources, and the collected SAR image data will be passed to the data preprocessing module for further processing. The SAR data acquisition module includes a high-resolution SAR data acquisition module imaging device, which supports the acquisition of SAR image data in different polarization modes. The SAR data acquisition module can describe the surface morphology in detail, including complex terrain undulations and fine details of objects, laying a solid data foundation for subsequent image analysis and accurate classification of objects. It supports the acquisition of multi-polarization modes such as HH, VV, HV, and VH, and effectively captures the scattering characteristics of objects under different polarization states. These polarization data deeply reveal the differences in the physical properties and scattering mechanisms of objects, greatly enriching the data dimensions of object classification, target recognition, and scene analysis. The comprehensive analysis of multi-polarization data can significantly improve the accuracy of ground object recognition and target identification. For example, the sensitivity of HH polarization to dihedral scattering is suitable for vegetation and building detection, while VV polarization is good at capturing the scattering characteristics of rough surfaces, and HV and VH polarizations are highly sensitive to vegetation-soil moisture status.

[0139] Data preprocessing module: including noise suppression unit and terrain correction unit, which are used to perform preliminary denoising and correction on SAR data, reduce noise and terrain errors in the interference pattern, and improve data quality. The preprocessed data is output to the phase unwrapping module to generate displacement information. The data preprocessing module includes noise suppression unit and terrain correction unit, which are used to reduce noise and terrain errors in the interference pattern. The noise suppression unit focuses on the effective suppression and reduction of noise in the interference pattern, significantly improving the purity of the data. Noise, as a random disturbance factor in the data, often obscures the real signal and features. Noise suppression technology aims to restore the essential information of the data and provide high-quality data support for subsequent analysis, processing and modeling. High-quality data can significantly reduce the risk of model overfitting, enhance the accuracy and generalization ability of model prediction and classification, make the model more focused on the real data features, and improve decision-making accuracy. In addition, noise suppression also optimizes the visual clarity of the interference pattern, highlights the characteristics of the ground objects, and helps image depth analysis and target accurate identification. The terrain correction unit accurately calculates and eliminates the geometric distortion and phase error of SAR images caused by terrain undulations, restores data authenticity, improves data accuracy and consistency, and provides more accurate surface information for disaster monitoring, terrain mapping, urban planning and other fields.

[0140] Phase unwrapping module: unwraps the preprocessed interference phase to extract the real slope displacement information to restore the original surface displacement information. The unwrapped displacement information will be passed to the displacement analysis module for subsequent spatiotemporal displacement analysis;

[0141] Displacement analysis module: Based on the distributed point target analysis method, the disentangled displacement information is analyzed in time and space, and the time and space displacement distribution map of the slope in the monitoring area is generated to realize the continuous monitoring of the slope displacement. The analyzed time and space displacement data will be transmitted to the displacement prediction module to predict the future displacement trend;

[0142] Displacement prediction module: Based on the long short-term memory network model, it performs nonlinear prediction of slope displacement data, evaluates future displacement trends, and improves prediction accuracy through continuous iterative training. The prediction results will be transmitted to the external early warning system module to generate an early warning report. The built-in machine learning algorithm of the displacement prediction module has an automatic update function, which improves prediction accuracy through continuous iterative training. The algorithm has the ability to receive real-time data and adaptively adjust model parameters, and can quickly respond to environmental changes and data feature updates, continuously optimize the prediction model, and ensure the high accuracy and reliability of the prediction results. With the deepening of iterative training, the algorithm continues to accumulate knowledge of data features and rules, and the prediction accuracy is steadily improved, gradually approaching the actual situation. The automatic update function realizes the automation of iterative training, reduces the need for manual intervention and the risk of human error, while reducing model maintenance costs and ensuring the continued stability of the algorithm's prediction performance;

[0143] Early warning system module: The early warning system module integrates the data from the displacement prediction module to generate a detailed displacement risk assessment report and issues an early warning signal according to the predicted risk level. The module is linked with external disaster prevention and mitigation agencies to ensure the real-time transmission of early warning information. The early warning system module not only generates reports, but also transmits real-time displacement monitoring data back to the data storage system of the external system for subsequent further analysis or model update. The early warning system module is connected to the basin disaster prevention and mitigation platform and can transmit monitoring data and early warning information in real time. The networked early warning system realizes the instant sharing of monitoring data to the basin disaster prevention and mitigation platform, builds a real-time data circulation network, and provides a solid information foundation for rapid response and decision-making. Real-time data transmission ensures the instant generation and efficient dissemination of early warning information, effectively shortens the disaster response time, and reduces losses. The platform integrates multi-source early warning system data, improves the comprehensive analysis capability of early warning, accurately identifies potential disaster risks, and deploys preventive measures in advance. At the same time, using the platform's extensive communication channels, early warning information can be quickly covered to relevant departments and the public, significantly improving the effective communication rate and reception of early warning information.

[0144] Working principle:

[0145] Data acquisition and preprocessing: SAR images of the target basin are collected in multiple phases through synthetic aperture radar to ensure the continuity and timeliness of the data. These images provide the necessary time series information for subsequent interferometric analysis. The collected SAR images are preprocessed, including radiation calibration, noise filtering, orbit error correction and geometric registration, to generate interferograms and remove errors such as atmosphere and terrain, ensuring image accuracy and reliability of interferometric phase.

[0146] Interference pattern processing and phase unwrapping: The initial interference pattern is generated by InSAR technology, and an adaptive noise suppression algorithm based on wavelet transform is applied to effectively remove noise and interference in the image to ensure the accuracy of the monitoring results. The phase error caused by the terrain is corrected by a high-precision digital elevation model (DEM) to make the interference phase consistent with the actual terrain morphology. The phase unwrapping algorithm is used to convert complex phase information into continuous surface displacement data. This is the core step of InSAR displacement monitoring and directly determines the accuracy of the displacement results.

[0147] Calibration and analysis: Combine ground reference points (such as GPS sites or known stable buildings) to calibrate the accuracy of the unwrapped displacement data. The stability and accuracy of the reference points ensure the reliability of the unwrapped results. Using the distributed point target analysis method, the calibrated displacement data is deeply analyzed to generate a spatiotemporal displacement distribution map of the slope area, reflecting the dynamic changes of the slope.

[0148] Displacement prediction and early warning: The long short-term memory network (LSTM) model is used to predict future displacement trends based on historical displacement data. The nonlinear modeling capability of LSTM enables this method to make high-precision predictions in complex and diverse environments. Based on the analysis and prediction results, a displacement risk assessment report is generated, and early warning signals are issued to disaster prevention and mitigation agencies in real time. The automated response mechanism of this module ensures that protective measures are taken in a timely manner to reduce disaster risks.

[0149] This method combines InSAR technology, digital elevation model and machine learning to monitor and predict slope displacement. It uses the spatiotemporal changes in radar image phase to extract surface deformation information, and uses efficient algorithms to remove errors and improve accuracy. Combined with the intelligent prediction and early warning system of historical data, it realizes the full process of automated slope monitoring from data collection, processing, analysis to early warning.

[0150] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and the present invention can be implemented in other specific forms without departing from the spirit or basic features of the present invention. Therefore, no matter from which point of view, the embodiments should be regarded as exemplary and non-restrictive. The scope of the present invention is defined by the attached claims rather than the above description, and it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention. Any figure mark in the claims should not be regarded as limiting the claims involved.

[0151] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

Claims

1. A method for monitoring watershed slope displacement using InSAR, characterized in that: The following steps are involved: S1. Use the SAR data acquisition module to collect multi-temporal SAR images of the target watershed area to ensure continuous and regular data acquisition and provide time series information for subsequent monitoring; S2, the data preprocessing module receives the collected SAR image data, generates the initial interferogram using InSAR technology, and identifies noise and errors during the image generation process; S3. In the data preprocessing module, an adaptive noise suppression algorithm is added to effectively remove the noise in the interference pattern through time-frequency domain filtering, which greatly reduces the impact of atmospheric delay, orbit error and terrain error on monitoring accuracy; S4, the data preprocessing module further uses a high-precision digital elevation model to correct the interference phase, eliminate the phase error caused by the terrain height difference, and ensure the accuracy of the phase information; S5, the phase unwrapping module receives the corrected interferometric phase data, and uses a phase unwrapping algorithm to convert the complex phase value into continuous displacement information to restore the actual surface displacement; S6. The phase unwrapping module calibrates the unwrapped displacement information in combination with the ground reference point to ensure the reliability and accuracy of the calculation results; S7, displacement analysis module uses distributed point target analysis method to conduct in-depth analysis on the calibrated displacement data, generate the spatiotemporal displacement distribution of the slope in the monitoring area, and provide accurate displacement evolution information; S8, displacement prediction module combines historical slope displacement data and uses machine learning models for prediction; S9, the early warning system module receives data from the displacement analysis and prediction module, generates a spatiotemporal distribution map of slope displacement, and forms a displacement risk assessment report in combination with historical displacement data for safety assessment of the monitoring area; S10. The early warning system module automatically sends out early warning signals to relevant disaster prevention and mitigation agencies based on the displacement trend prediction results and risk assessment reports, and links external systems in real time to ensure that protective measures are taken in a timely manner.

2. The InSAR watershed slope displacement monitoring method according to claim 1, characterized in that: The adaptive noise suppression algorithm used in S3 is a time-frequency domain filtering algorithm based on wavelet transform, which is used to effectively reduce noise interference.

3. The InSAR watershed slope displacement monitoring method according to claim 1, characterized in that: The ground reference points used in S6 are GPS monitoring points and known stable buildings deployed in the watershed to improve the phase calibration accuracy.

4. The InSAR watershed slope displacement monitoring method according to claim 1, characterized in that: The machine learning model in S8 is a nonlinear displacement prediction model based on long short-term memory network, and its core is intelligent prediction based on time series data.

5. The InSAR watershed slope displacement monitoring method according to claim 1, characterized in that: The SAR image data is acquired once a month or more frequently to achieve high-efficiency displacement monitoring.

6. An InSAR basin slope displacement monitoring device, comprising a vehicle body, on which a SAR data acquisition module is rotatably mounted via a rotating base, characterized in that: A data preprocessing module and a phase unwrapping module are installed on one side of the SAR data acquisition module on the vehicle body, and a displacement analysis module and a displacement prediction module are installed on one side of the data preprocessing module and the phase unwrapping module on the vehicle body. The SAR data acquisition module is located on the center line of the vehicle body, and the data preprocessing module and the phase unwrapping module are a group, and the displacement prediction module and the rotating base are a group and are symmetrically distributed on both sides of the center line of the vehicle body; SAR data acquisition module: contains high-resolution radar imaging devices, supports the acquisition of SAR image data in different polarization modes, ensures the diversity and richness of data sources, and the collected SAR image data will be passed to the data preprocessing module for further processing; Data preprocessing module: including noise suppression unit and terrain correction unit, which are used to perform preliminary denoising and correction on SAR data, reduce noise and terrain errors in the interference pattern, and improve data quality. The preprocessed data is output to the phase unwrapping module to generate displacement information; Phase unwrapping module: unwraps the preprocessed interference phase to extract the real slope displacement information to restore the original surface displacement information. The unwrapped displacement information will be passed to the displacement analysis module for subsequent spatiotemporal displacement analysis; Displacement analysis module: Based on the distributed point target analysis method, the disentangled displacement information is analyzed in time and space, and the time and space displacement distribution map of the slope in the monitoring area is generated to realize the continuous monitoring of the slope displacement. The analyzed time and space displacement data will be transmitted to the displacement prediction module to predict the future displacement trend; Displacement prediction module: Based on the long short-term memory network model, it performs nonlinear prediction of slope displacement data, evaluates future displacement trends, and improves prediction accuracy through continuous iterative training. The prediction results will be transmitted to the external early warning system module to generate an early warning report; Early warning system module: The early warning system module integrates the data from the displacement prediction module to generate a detailed displacement risk assessment report and issues a warning signal based on the predicted risk level. The module works in conjunction with external disaster prevention and mitigation agencies to ensure real-time transmission of warning information. The early warning system module not only generates reports, but also transmits real-time displacement monitoring data back to the data storage system of the external system for subsequent further analysis or model updates.

7. The InSAR watershed slope displacement monitoring device according to claim 6, characterized in that: The SAR data acquisition module includes a high-resolution SAR data acquisition module imaging device, which supports the acquisition of SAR image data in different polarization modes.

8. The InSAR watershed slope displacement monitoring device according to claim 6, characterized in that: The data preprocessing module includes a noise suppression unit and a terrain correction unit, which are used to reduce noise and terrain errors in the interference pattern respectively.

9. The InSAR watershed slope displacement monitoring device according to claim 6, characterized in that: The machine learning algorithm built into the displacement prediction module has an automatic update function, which improves the prediction accuracy through continuous iterative training.

10. The InSAR watershed slope displacement monitoring device according to claim 6, characterized in that: The early warning system module is connected to the river basin disaster prevention and mitigation platform and can transmit monitoring data and early warning information in real time.

Images