Self-adaptive phase unwrapping ground-based SAR deformation monitoring method and system

Abstract

The invention relates to the technical field of geological monitoring, and discloses a self-adaptive phase unwrapping foundation SAR deformation monitoring method and system. The method comprises the steps of collecting ground-based SAR echo data, extracting phase gradient features and noise distribution, constructing a model based on a compressed sensing theory to solve a phase unwrapping result, establishing a deformation field dynamic evolution model fusing multiple factors, and generating a deformation monitoring value through an adaptive weighted fusion algorithm. And an abnormal deformation detection model is also constructed to realize adaptive re-calibration of model parameters, and a multi-sensor collaborative calibration network is constructed to improve the monitoring precision. The problem of phase unwrapping is effectively solved, multiple influence factors are comprehensively considered, surface deformation can be accurately monitored in real time, abnormity can be timely found, the model can be calibrated, multi-sensor data can be fused, and the method has important application value in the fields of geological disaster early warning, engineering structure safety assessment and the like.

Description

Technical Field

[0001] The present invention relates to the field of geological monitoring technology, and in particular to an adaptive phase unwrapping ground-based SAR deformation monitoring method and system. Background Art

[0002] In the field of geological monitoring, accurate and timely monitoring of surface deformation is of vital importance. Whether in preventing geological disasters, protecting people’s lives and property, or ensuring the stable operation of large-scale infrastructure projects, accurate deformation monitoring is an indispensable key link.

[0003] Traditional means of surface deformation monitoring, such as level measurement and total station measurement, can obtain deformation data to a certain extent, but they have many limitations. Level measurement relies on fixed measurement points and has a limited measurement range, making it difficult to achieve large-area synchronous monitoring. For areas with complex terrain and inconvenient transportation, implementation is difficult and inefficient. Total station measurement is also limited by observation distance and line of sight conditions. In mountainous areas, urban high-rise areas and other environments, measurement work often faces many obstacles and cannot fully and accurately obtain deformation information.

[0004] Synthetic Aperture Radar (SAR) is an active microwave remote sensing technology that uses radar sensors mounted on satellites, aircraft or ground platforms to transmit electromagnetic waves to the surface and receive reflected signals, and reconstruct high-resolution deformation information of the target area using the principle of synthetic aperture and interferometry (InSAR). Unlike traditional optical remote sensing, SAR has the ability to penetrate clouds, rain and fog and work at night, and can monitor tiny surface deformations (millimeter level) all day and all weather. According to the platform type, SAR is mainly divided into two categories: spaceborne SAR and ground-based SAR: spaceborne SAR has a wide coverage range but limited temporal and spatial resolution, while ground-based SAR achieves high-precision local monitoring through near-ground deployment.

[0005] Although satellite SAR technology can achieve large-area observation, its spatial resolution is relatively low, so the monitoring accuracy of some local micro-deformations is insufficient. In addition, the satellite observation period is long, which cannot meet the needs of real-time dynamic monitoring of deformation.

[0006] The emergence of ground-based SAR technology has brought new solutions to surface deformation monitoring. It has high spatial and temporal resolution and can continuously observe the monitoring area. However, in practical applications, ground-based SAR also faces a series of challenges. The information contained in the echo data is complex, and the phase unwrapping problem has become a key factor restricting the monitoring accuracy. The phase unwrapping process is easily interfered by noise, resulting in errors in the unwrapping results and affecting the judgment of the real deformation.

[0007] In addition, existing deformation monitoring methods often fail to fully consider the combined effects of multiple factors on deformation, such as surface physical parameters, ambient temperature field, and external loads. In actual situations, these factors interact with each other and cause surface deformation. Ignoring any one of these factors may lead to deviations in monitoring results and fail to accurately reflect the true situation of deformation.

[0008] At the same time, the monitoring data of a single sensor has limitations, and it is difficult to fully and accurately describe the deformation characteristics. Different sensors have different measurement principles and precisions. How to effectively fuse the data of multiple sensors to improve the reliability and accuracy of monitoring is also a problem that needs to be solved urgently. In terms of abnormal deformation detection, the existing methods lack efficient and intelligent detection mechanisms, and are unable to timely detect potential safety hazards and issue early warnings. In summary, it is of great practical significance to develop an adaptive phase unwrapping ground-based SAR deformation monitoring method and system that can overcome the above problems. Summary of the invention

[0009] The object of the present invention is to provide an adaptive phase unwrapping ground-based SAR deformation monitoring method and system to solve the problems raised in the above background technology.

[0010] To achieve the above object, the present invention provides the following technical solution: an adaptive phase unwrapping ground-based SAR deformation monitoring method, the method comprising: Collect echo data of ground-based SAR, including interference phase, polarization scattering matrix and time series observation information; Extracting phase gradient characteristics and noise distribution from the echo data; A phase unwrapping optimization model is constructed based on compressed sensing theory, and the global phase unwrapping result is solved by a sparse constrained iterative algorithm. Establishing a dynamic evolution model of deformation field, wherein the dynamic evolution model of deformation field integrates surface physical parameters, ambient temperature field and external load action data; According to the phase unwrapping result and the deformation field dynamic evolution model, a deformation monitoring value is generated through an adaptive weighted fusion algorithm.

[0011] Preferably, the phase gradient characteristics include a local phase difference matrix, a frequency domain coherence spectrum and a polarization phase consistency characteristic.

[0012] Preferably, the sparse constrained iterative algorithm comprises: Constructing a phase residual objective function, wherein the function includes an L1 norm sparsity term and a total variation regularization term; The alternating direction multiplier method is used to decompose the objective function and generate iterative update rules; The convergence is accelerated by an adaptive step-size adjustment strategy, and a residual feedback mechanism is introduced to dynamically correct the iteration path.

[0013] Preferably, the deformation field dynamic evolution model includes: The surface medium is discretized based on the finite element method, and the partial differential equations of node displacement and strain tensor are established; Integrate temperature gradient field data and correct deformation field boundary conditions through thermal-mechanical coupling analysis; Combined with the Kalman filter algorithm, the deformation field is predicted and updated in real time.

[0014] Preferably, the adaptive weighted fusion algorithm includes: Calculate the frequency band weight coefficient according to the signal-to-noise ratio and spatial resolution of the phase unwrapping result; Construct a multi-scale fusion pyramid and perform weighted superposition of deformation fields at various scales; The rasterization artifacts of the fused result are eliminated by edge-preserving filtering.

[0015] Preferably, the method further comprises: An abnormal deformation detection model is constructed based on a generative adversarial network, wherein the model uses historical deformation data as real samples and real-time monitoring data as generated samples; When the probability of abnormality output by the discriminator exceeds a threshold, adaptive recalibration of the deformation field model parameters is triggered.

[0016] Preferably, the adaptive recalibration comprises: Define the credibility intervals and physical constraints of deformation field parameters; The Bayesian optimization algorithm is used to search for the optimal parameter combination, and the parameter posterior distribution is verified by Markov chain Monte Carlo sampling.

[0017] Preferably, the method further comprises: Constructing a multi-sensor collaborative calibration network that integrates GNSS displacement data, inclinometer measurement data, and SAR deformation monitoring values; The spatiotemporal correlation between sensors is modeled through graph convolutional neural networks to generate joint calibration corrections.

[0018] Preferably, the present invention further includes an adaptive phase unwrapping ground-based SAR deformation monitoring system, the system comprising: Data acquisition module: used to collect echo data of ground-based SAR; Phase filtering module: extracting phase gradient characteristics and noise distribution from the echo data; Disentanglement optimization module: builds a phase disentanglement optimization model based on compressed sensing theory, and solves the global phase disentanglement result through sparse constrained iterative algorithm; Deformation modeling module: establishes a dynamic evolution model of the deformation field, integrating surface physical parameters and ambient temperature field data; Fusion output module: Generates deformation monitoring values ​​through adaptive weighted fusion algorithm based on the phase unwrapping results and the deformation field model.

[0019] Compared with the prior art, the present invention has the following beneficial effects: In the data processing and phase unwrapping process, the echo data of ground-based SAR is collected, and the phase gradient characteristics and noise distribution are extracted. This process can fully mine the effective information in the data and provide a basis for subsequent accurate unwrapping and analysis. Based on the compressed sensing theory, a phase unwrapping optimization model is constructed, and the global phase unwrapping result is solved by combining the sparse constrained iterative algorithm. The algorithm constructs a phase residual objective function containing L1 norm sparse terms and total variation regularization terms, decomposes and generates iterative update rules using the alternating direction multiplier method, and adopts an adaptive step size adjustment strategy to accelerate convergence and introduces a residual feedback mechanism to dynamically correct the iterative path. These operations can not only effectively suppress the interference of noise on phase unwrapping and improve the unwrapping accuracy, but also are more stable and accurate when processing complex data compared to traditional unwrapping methods, and can more accurately restore the real phase information, providing a reliable basis for subsequent deformation analysis.

[0020] The established dynamic evolution model of deformation field integrates surface physical parameters, ambient temperature field and external load data. By discretizing the surface medium based on the finite element method, partial differential equations of node displacement and strain tensor are established, and the deformation field boundary conditions are corrected by thermomechanical coupling analysis combined with temperature gradient field data, and the deformation field is predicted and updated in real time using the Kalman filter algorithm. This enables the model to comprehensively and accurately reflect the dynamic change process of surface deformation, fully considering the influence of various practical factors on deformation. Compared with traditional models, the prediction results are closer to the actual situation, and potential deformation trends can be discovered in advance, providing a more reliable basis for disaster warning and engineering safety assessment.

[0021] In terms of deformation monitoring value generation, an adaptive weighted fusion algorithm is used to calculate the frequency band weight coefficient based on the signal-to-noise ratio and spatial resolution of the phase unwrapping result, and a multi-scale fusion pyramid is constructed to perform weighted superposition of the deformation field at all levels of scale, and the rasterization artifacts of the fusion result are eliminated through edge protection filtering. This fusion method can give full play to the advantages of data in different frequency bands, improve the accuracy and resolution of the monitoring results, and effectively eliminate the artifacts generated during the fusion process, making the monitoring results clearer and more reliable, and facilitating intuitive analysis and judgment of deformation conditions.

[0022] An abnormal deformation detection model is constructed based on a generative adversarial network, with historical deformation data as real samples and real-time monitoring data as generated samples. When the probability of abnormality output by the discriminator exceeds the threshold, the adaptive recalibration of the deformation field model parameters is triggered. By defining the credibility interval and physical constraints of the deformation field parameters, the Bayesian optimization algorithm is used to search for the optimal parameter combination, and the posterior distribution of the parameters is verified by Markov chain Monte Carlo sampling. This mechanism can detect abnormal deformation in real time and intelligently, discover potential safety hazards in a timely manner, and ensure the accuracy and reliability of the model through adaptive recalibration, greatly improving the early warning capability of the monitoring system.

[0023] The constructed multi-sensor collaborative calibration network integrates GNSS displacement data, inclinometer measurement data and SAR deformation monitoring values, and generates joint calibration corrections by modeling the spatiotemporal correlation between sensors through graph convolutional neural networks. The network realizes the effective fusion of multi-sensor data, fully utilizes the advantages of different sensors, complements and verifies each other, improves the reliability and accuracy of monitoring data, and provides strong support for a more comprehensive and accurate understanding of surface deformation. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 This is a working principle diagram of the adaptive phase unwrapping ground-based SAR deformation monitoring method of the present invention; Figure 2 The workflow diagram of the sparse constrained iterative algorithm; Figure 3 Flowchart for abnormal deformation detection and recalibration triggering; Figure 4 The architecture diagram of the adaptive phase unwrapping ground-based SAR deformation monitoring system. DETAILED DESCRIPTION

[0025] 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.

[0026] See also Figure 1-Figure 4 The present invention provides a technical solution: an adaptive phase unwrapping ground-based SAR deformation monitoring method, the method comprising: Specialized ground-based SAR equipment is used to observe the monitoring area and obtain echo data. These data include interference phase, polarization scattering matrix and time series observation information. Interference phase can reflect the phase difference at different locations in the monitoring area, which is an important basis for subsequent deformation analysis; polarization scattering matrix contains polarization characteristic information of ground objects, which helps to understand the ground features of the monitoring area more comprehensively; time series observation information records observation data at different time points, which can be used to analyze the trend of deformation over time. In the actual acquisition process, according to the scope of the monitoring area, the complexity of the terrain and the requirements of monitoring accuracy, the parameters of the ground-based SAR equipment, such as transmission frequency, bandwidth, observation angle, etc., are reasonably set to ensure the acquisition of high-quality echo data.

[0027] The collected echo data is processed to extract phase gradient characteristics and noise distribution. Phase gradient characteristics can reflect the change of phase, which is of great significance for subsequent phase unwrapping; the acquisition of noise distribution helps to suppress noise in subsequent processing and improve monitoring accuracy. In the extraction process, special data processing algorithms, such as algorithms based on signal processing theory, are used to analyze and calculate echo data to obtain accurate phase gradient characteristics and noise distribution information.

[0028] Based on the theory of compressed sensing, a phase unwrapping optimization model is constructed. The model uses the sparsity of the signal to solve the global phase unwrapping result through a sparse constrained iterative algorithm. When constructing the model, the characteristics of the monitoring data and actual needs are fully considered, and the model parameters are reasonably set. The sparse constrained iterative algorithm can improve the computational efficiency while ensuring the unwrapping accuracy. In the solution process, it is continuously iterated and optimized to gradually obtain accurate global phase unwrapping results.

[0029] A dynamic evolution model of the deformation field is established, which integrates surface physical parameters, ambient temperature field and external load data. Surface physical parameters include soil type, geological structure, etc., which will affect the deformation characteristics of the surface; changes in the ambient temperature field will cause thermal expansion and contraction of surface materials, thereby affecting deformation; external loads, such as the weight of buildings and traffic loads, are also important factors causing surface deformation. By integrating these data, the dynamic evolution process of the deformation field can be described more accurately. When establishing the model, relevant physical theories and mathematical methods are used to incorporate these factors into the model, and reasonable parameter settings and model calibration are performed.

[0030] The phase unwrapping results are combined with the dynamic evolution model of the deformation field, and the deformation monitoring value is generated using the adaptive weighted fusion algorithm. The adaptive weighted fusion algorithm automatically adjusts the weight according to the characteristics and reliability of different data, fully combining the advantages of the two to obtain more accurate deformation monitoring values. In practical applications, the algorithm parameters are reasonably adjusted according to the specific conditions and needs of the monitoring area to achieve the best monitoring effect.

[0031] The present invention will be further described below in conjunction with Examples 1 to 6: Example 1

[0032] When processing echo data to extract phase gradient features, we focus on the local phase difference matrix, frequency domain coherence spectrum, and polarization phase consistency features. The local phase difference matrix is ​​used to describe the phase difference between adjacent pixels. By calculating the phase difference between adjacent pixels, the phase change trend can be clearly displayed. Assume that the phases of adjacent pixels are and , then the elements in the local phase difference matrix can be expressed as The frequency domain coherence spectrum is obtained by performing frequency domain analysis on the echo data, which reflects the correlation between different frequency components. When calculating the frequency domain coherence spectrum, the echo data is first Fourier transformed to obtain frequency domain data, and then the coherence coefficients between different frequency components are calculated. The polarization phase consistency feature is extracted based on the polarization scattering matrix, which measures the degree of phase consistency between different polarization channels. For the polarization scattering matrix , the phases of different polarization channels are calculated through a specific algorithm, and then the polarization phase consistency features are obtained. These phase gradient features reflect the characteristics of the echo data from different angles, providing rich information for subsequent phase unwrapping and deformation analysis.

[0033] In actual operation, the parameters and algorithms for calculating these features are reasonably selected according to the type of ground objects in the monitoring area and the purpose of monitoring. For example, when monitoring densely built areas, the calculation window of the local phase difference matrix can be appropriately reduced to more accurately capture the phase changes at the boundaries of buildings; while when monitoring large natural areas, the calculation of the frequency domain coherence spectrum and polarization phase consistency features can use a wider frequency range and more polarization channels to obtain more comprehensive ground object information. At the same time, combined with the noise distribution information, the extracted phase gradient features are denoised to improve the accuracy and reliability of the features. Filtering algorithms, such as median filtering and Gaussian filtering, can be used to smooth the local phase difference matrix and remove abnormal phase differences caused by noise; for the frequency domain coherence spectrum and polarization phase consistency features, the frequency components and polarization channels with large noise interference can be removed by setting thresholds. Example 2

[0034] When using the sparse constrained iterative algorithm to solve the global phase unwrapping result, the phase residual objective function is first constructed. This function contains the L1 norm sparse term and the total variation regularization term. Assume that the phase to be unwrapped is , the observed phase is , then the phase residual objective function It can be expressed as: , in, and is the weight coefficient, which is used to balance the importance of different items; Indicates phase The L1 norm sparse term can promote the sparseness of the phase gradient, which is conducive to removing noise and retaining the edge information of the phase; is the total variation regularization term, which is used to maintain the smoothness of the phase and avoid excessive fluctuations in the unwrapping results; is the fitting term of the observed data, ensuring that the unwrapping result is as close to the observed data as possible.

[0035] Next, the alternating direction multiplier method is used to decompose the objective function and generate an iterative update rule. By decomposing the objective function into multiple sub-problems, solving each sub-problem separately, and then updating the variables alternately, the optimal solution is gradually approached. During the iteration process, an adaptive step adjustment strategy is used to accelerate convergence. According to the results of each iteration, the step size is dynamically adjusted. For example, if the current iteration result changes greatly, the step size is appropriately increased to speed up the convergence; if the change is small, the step size is reduced to improve the convergence accuracy. At the same time, a residual feedback mechanism is introduced to dynamically correct the iteration path. According to the phase residual after each iteration, the direction of the next iteration is adjusted to make the iteration process more stable and accurate. In practical applications, it is necessary to reasonably select the weight coefficient according to the characteristics of the monitoring data and the limitations of computing resources. and . You can test different combinations of values ​​through experiments, observe the quality of the unwrapping results, and select the optimal parameters. In addition, you also need to pay attention to the setting of the iteration termination condition to avoid excessive iterations that lead to waste of computing resources. Generally, the termination condition can be determined based on the change in the phase residual or the number of iterations. For example, when the phase residual is less than a preset threshold, or the number of iterations reaches a certain upper limit, stop the iteration. Example 3

[0036] When establishing the dynamic evolution model of the deformation field, the surface medium is discretized based on the finite element method. The surface medium is divided into multiple finite element units, each of which is composed of nodes. The partial differential equations of node displacement and strain tensor are established to describe the deformation behavior of the surface medium under stress. Assume that the node displacement is , the strain tensor is , according to the theory of elasticity, the relationship between them can be expressed by the partial differential equation: ,in is the stress tensor, and the strain tensor Through the constitutive relationship Get in touch, It is the elastic matrix, which depends on the material properties of the surface medium. The temperature gradient field data is integrated and the deformation field boundary conditions are corrected through thermal-mechanical coupling analysis.

[0037] Temperature changes can cause thermal expansion or contraction of the surface medium, thus generating additional stress and strain. Assume that the temperature change is , the thermal expansion coefficient is , then the thermal strain ,in is a unit tensor. Thermal strain is incorporated into the strain tensor to modify the boundary conditions of the deformation field. The Kalman filter algorithm is combined to predict and update the deformation field in real time. The Kalman filter algorithm can use the state equation and observation equation of the system to optimally estimate and predict the state of the deformation field. Suppose the state vector of the deformation field is , the observation vector is , the state equation is , the observation equation is ,in is the state transition matrix, is the observation matrix, and They are process noise and observation noise respectively. Through the prediction and update steps of the Kalman filter algorithm, the estimated value of the deformation field is continuously adjusted to make it closer to the actual deformation. In practical applications, it is necessary to accurately obtain the material parameters of the surface medium, such as elastic modulus, Poisson's ratio, thermal expansion coefficient, etc. These parameters can be obtained through field tests, geological surveys, etc. At the same time, to ensure the accuracy of the temperature gradient field data, multiple temperature sensors can be arranged in the monitoring area to collect temperature data in real time, and data fusion and processing can be performed. In addition, it is also necessary to reasonably set the parameters of the Kalman filter algorithm, such as the process noise covariance matrix and the observation noise covariance matrix, to improve the accuracy of prediction and update. Example 4

[0038] When the adaptive weighted fusion algorithm is used to generate deformation monitoring values, the frequency band weight coefficient is calculated according to the signal-to-noise ratio and spatial resolution of the phase unwrapping result. The signal-to-noise ratio of the phase unwrapping result of each frequency band is , the spatial resolution is , then the frequency band weight coefficient It can be calculated by the following formula: , in is the total number of frequency bands. In this way, frequency bands with high signal-to-noise ratio and high spatial resolution will get greater weights and play a more important role in the fusion process.

[0039] Construct a multi-scale fusion pyramid and perform weighted superposition of deformation fields at each scale. First, perform multi-scale decomposition of deformation field data of different frequency bands to obtain image pyramids of different resolutions. At each level of scale, perform weighted superposition of deformation field data of each frequency band according to the calculated frequency band weight coefficient. For example, at a certain level of scale, the fused deformation field It can be expressed as , in It is The frequency band is Deformation field data at the level of the scale. The rasterization artifacts of the fusion results are eliminated by edge protection filtering. The edge protection filtering algorithm can smooth the fusion results and remove rasterization artifacts while retaining the edge information of the deformation field. Bilateral filtering, guided filtering and other algorithms can be used to select appropriate filtering parameters, such as filter radius, standard deviation, etc., according to the characteristics and requirements of the fusion results. In practical applications, it is necessary to accurately evaluate the signal-to-noise ratio and spatial resolution. The signal-to-noise ratio can be calculated by the ratio of signal power to noise power. The characteristics of the echo data and the distribution of noise should be considered in the calculation process. The spatial resolution is related to the parameters of the ground-based SAR equipment and the data processing method, and can be determined by observing and analyzing known standard targets. In addition, when constructing a multi-scale fusion pyramid, the decomposition scale and fusion method should be reasonably selected to ensure that the fusion result can reflect the overall trend of the deformation field and retain the detailed information. At the same time, when performing edge protection filtering, care should be taken to avoid excessive filtering that leads to loss of edge information, which affects the accuracy of deformation monitoring. Example 5

[0040] An abnormal deformation detection model is built based on a generative adversarial network, with historical deformation data as real samples and real-time monitoring data as generated samples. and the discriminator Composition. Generator The role of is to generate simulated deformation data based on random noise, making it as close as possible to the real historical deformation data; the discriminator It is responsible for judging whether the input data is real historical deformation data or simulated data generated by the generator. During the training process, the generator and the discriminator compete with each other and continuously optimize their parameters. Suppose the random noise is , the simulated deformation data generated by the generator is , the real historical deformation data is , then the discriminator The objective function can be expressed as: , in is the distribution of real historical deformation data, is the distribution of random noise. When the probability of abnormal output of the discriminator exceeds the threshold, the adaptive recalibration of the deformation field model parameters is triggered.

[0041] In the adaptive recalibration process, the credibility interval and physical constraints of the deformation field parameters are first defined. For example, for the elastic modulus of the surface medium, its reasonable value range is determined as the credibility interval based on geological data and experience; at the same time, physical constraints are set according to physical principles, such as the elastic modulus cannot be negative. Then the Bayesian optimization algorithm is used to search for the optimal parameter combination. The Bayesian optimization algorithm estimates the location of the optimal parameters by constructing a proxy model of the objective function using probability distribution. During the search process, the proxy model is continuously updated according to the new sampling points, gradually approaching the optimal solution. Finally, the posterior distribution of the parameters is verified by Markov chain Monte Carlo sampling. Markov chain Monte Carlo sampling can sample from the posterior distribution of the parameters and obtain a series of parameter samples that conform to the posterior distribution. By analyzing these samples, it can be verified whether the optimal parameter combination obtained by the Bayesian optimization algorithm is reasonable and whether it conforms to the actual situation. In practical applications, the quality and representativeness of historical deformation data should be ensured. Historical deformation data should cover different geological conditions, environmental factors and time periods so that the generator can learn the real deformation law. At the same time, when setting the threshold of the discriminator, the sensitivity and false alarm rate of monitoring should be considered comprehensively. If the threshold is set too high, some abnormal deformations may be missed; if the threshold is set too low, more false alarms may be generated. In addition, during the adaptive recalibration process, it is necessary to make full use of the existing monitoring data and prior knowledge to improve the efficiency and accuracy of recalibration. Example 6

[0042] Construct a multi-sensor collaborative calibration network that integrates GNSS displacement data, inclinometer measurement data, and SAR deformation monitoring values. GNSS displacement data can provide high-precision absolute displacement information, inclinometer measurement data can reflect the tilt change of the surface, and SAR deformation monitoring values ​​have the characteristics of large area and high resolution. Combining these three types of data can more comprehensively monitor surface deformation. The spatiotemporal correlation between sensors is modeled through a graph convolutional neural network to generate a joint calibration correction. Graph convolutional neural networks can process data with a graph structure, treating sensors as nodes in the graph and the spatiotemporal relationships between sensors as edges. Assume that the number of sensors is , the node feature matrix is ,in is the dimension of node features, and the adjacency matrix is , represents the connection relationship between sensors. The graph convolutional neural network learns the spatiotemporal correlation between sensors by performing convolution operations on the node feature matrix and the adjacency matrix. Its convolution operation can be expressed as: , in It is The node feature matrix of the layer, It is The weight matrix of the layer, is the activation function, , is the identity matrix, yes Through the learning of graph convolutional neural network, the spatiotemporal correlation model between sensors is obtained, and then the joint calibration correction is generated according to the GNSS displacement data, inclinometer measurement data and SAR deformation monitoring value.

[0043] In practical applications, the accuracy and synchronization of the data from each sensor must be ensured. GNSS equipment and inclinometers should be calibrated and maintained regularly to ensure the accuracy of the measurement data; at the same time, the observation time of SAR monitoring and other sensors should be synchronized as much as possible to avoid data fusion errors caused by time differences. In addition, when constructing a graph convolutional neural network, the structure and parameters of the network should be reasonably set, such as the number of layers, node feature dimensions, and initialization of the weight matrix. Different network structures and parameter combinations can be tested experimentally to select the optimal model to improve the accuracy of the joint calibration correction, thereby improving the overall deformation monitoring accuracy.

[0044] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0045] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An adaptive phase unwrapping ground-based SAR deformation monitoring method, characterized in that: include: Collect echo data of ground-based SAR, including interference phase, polarization scattering matrix and time series observation information; Extracting phase gradient characteristics and noise distribution from the echo data; A phase unwrapping optimization model is constructed based on compressed sensing theory, and the global phase unwrapping result is solved by a sparse constrained iterative algorithm. Establishing a dynamic evolution model of deformation field, wherein the dynamic evolution model of deformation field integrates surface physical parameters, ambient temperature field and external load action data; According to the phase unwrapping result and the deformation field dynamic evolution model, a deformation monitoring value is generated through an adaptive weighted fusion algorithm.

2. The adaptive phase unwrapping ground-based SAR deformation monitoring method according to claim 1 is characterized in that: The phase gradient characteristics include a local phase difference matrix, a frequency domain coherence spectrum and a polarization phase consistency characteristic.

3. The adaptive phase unwrapping ground-based SAR deformation monitoring method according to claim 1, characterized in that: The sparse constrained iterative algorithm includes: Constructing a phase residual objective function, wherein the function includes an L1 norm sparsity term and a total variation regularization term; The alternating direction multiplier method is used to decompose the objective function and generate iterative update rules; The convergence is accelerated by an adaptive step-size adjustment strategy, and a residual feedback mechanism is introduced to dynamically correct the iteration path.

4. The adaptive phase unwrapping ground-based SAR deformation monitoring method according to claim 3 is characterized in that: The deformation field dynamic evolution model includes: The surface medium is discretized based on the finite element method, and the partial differential equations of node displacement and strain tensor are established; Integrate temperature gradient field data and correct deformation field boundary conditions through thermal-mechanical coupling analysis; Combined with the Kalman filter algorithm, the deformation field is predicted and updated in real time.

5. The adaptive phase unwrapping ground-based SAR deformation monitoring method according to claim 4 is characterized in that: The adaptive weighted fusion algorithm includes: Calculate the frequency band weight coefficient according to the signal-to-noise ratio and spatial resolution of the phase unwrapping result; Construct a multi-scale fusion pyramid and perform weighted superposition of deformation fields at various scales; The rasterization artifacts of the fused result are eliminated by edge-preserving filtering.

6. The adaptive phase unwrapping ground-based SAR deformation monitoring method according to claim 5, characterized in that: The method further comprises: An abnormal deformation detection model is constructed based on a generative adversarial network, wherein the model uses historical deformation data as real samples and real-time monitoring data as generated samples; When the probability of abnormality output by the discriminator exceeds a threshold, adaptive recalibration of the deformation field model parameters is triggered.

7. The adaptive phase unwrapping ground-based SAR deformation monitoring method according to claim 6, characterized in that: The adaptive recalibration comprises: Define the credibility intervals and physical constraints of deformation field parameters; The Bayesian optimization algorithm is used to search for the optimal parameter combination, and the parameter posterior distribution is verified by Markov chain Monte Carlo sampling.

8. The adaptive phase unwrapping ground-based SAR deformation monitoring method according to claim 1, characterized in that: The method further comprises: Constructing a multi-sensor collaborative calibration network that integrates GNSS displacement data, inclinometer measurement data, and SAR deformation monitoring values; The spatiotemporal correlation between sensors is modeled through graph convolutional neural networks to generate joint calibration corrections.

9. An adaptive phase unwrapping ground-based SAR deformation monitoring system, characterized in that: include: Data acquisition module: used to collect echo data of ground-based SAR; Phase filtering module: extracting phase gradient characteristics and noise distribution from the echo data; Disentanglement optimization module: builds a phase disentanglement optimization model based on compressed sensing theory, and solves the global phase disentanglement result through sparse constrained iterative algorithm; Deformation modeling module: establishes a dynamic evolution model of the deformation field, integrating surface physical parameters and ambient temperature field data; Fusion output module: Generates deformation monitoring values ​​through adaptive weighted fusion algorithm based on the phase unwrapping results and the deformation field model.

Images