Deep foundation pit micro-deformation InSAR remote sensing method and system and storage medium

Abstract

The invention relates to the technical field of deep foundation pit monitoring, and discloses a deep foundation pit micro-deformation InSAR remote sensing method and system and a storage medium. The method comprises the following steps: synchronously acquiring data through a ground radar and an InSAR satellite, and obtaining a point target set and an original scattering characteristic sequence; obtaining a net deformation phase sequence without influence through registration residual correction, phase deduction and weighted fusion processing, and carrying out phase consistency quality control and unwrapping processing to obtain an accumulated deformation quantity sequence; constructing a deformation mechanism correlation characteristic matrix based on correlation analysis of the cumulative deformation quantity sequence and radar reflection intensity amplitude attenuation characteristics; and fusing the deformation mechanism correlation feature matrix and the real-time differential interference image group, extracting deformation rate and phase gradient change, and further generating a deep foundation pit local micro-deformation risk early warning index set. According to the invention, the tiny deformation of the surrounding area of the foundation pit can be effectively detected, potential risks are warned in advance, and the safety and monitoring efficiency in the construction process are improved.

Description

Technical Field

[0001] This application relates to the field of deep foundation pit monitoring technology, and in particular to an InSAR remote sensing method, system and storage medium for micro-deformation of deep foundation pits. Background Technology

[0002] Deep foundation pit construction is an indispensable part of modern urban construction. With the continuous increase in building depth and scale, the demand for foundation pit deformation monitoring is growing. Traditional foundation pit deformation monitoring methods, such as ground settlement observation and laser scanning, can provide some deformation information, but they have limitations such as uneven distribution of monitoring points, long data acquisition cycles, and great susceptibility to environmental interference, making it difficult to meet the precise monitoring requirements of large-scale deep foundation pit projects.

[0003] With the development of remote sensing technology, especially the continuous maturation of InSAR (Synthetic Aperture Radar Interferometry) technology, the methods for monitoring foundation pit deformation have been significantly improved. InSAR technology, by analyzing the phase changes of radar wave reflection signals, can achieve high-precision, high-spatial-resolution ground deformation monitoring. Particularly in deep foundation pit engineering, InSAR can achieve simultaneous monitoring over a wide area, overcoming the point-based limitations of traditional methods. However, existing InSAR monitoring methods still face technical challenges in monitoring micro-deformation in deep foundation pits, such as atmospheric interference, radar signal noise, and data processing complexity, and cannot achieve real-time, high-precision, multi-dimensional monitoring and early warning.

[0004] Therefore, how to combine ground radar and InSAR remote sensing technology to accurately monitor the micro-deformation of deep foundation pits and to conduct effective risk assessment and early warning based on real-time monitoring has become a key technical problem that urgently needs to be solved in this field. Summary of the Invention

[0005] This application provides an InSAR remote sensing method, system, and storage medium for micro-deformation of deep foundation pits, aiming to solve the problems of poor real-time performance, accuracy, and operability in the existing technology for monitoring the deformation of deep foundation pits. This invention achieves accurate monitoring and risk warning of deep foundation pit areas by integrating ground radar and InSAR remote sensing technology and combining multi-temporal monitoring data.

[0006] In a first aspect, this application provides an InSAR remote sensing method for micro-deformation of deep foundation pits, the method comprising: S1. By simultaneously acquiring continuous reflection signals and differential interferometric image sets of the deep foundation pit area through ground radar and InSAR satellite, a highly stable set of point targets and original scattering characteristic sequence are obtained. S2. Based on the point target set and the original scattering characteristic sequence, registration residual correction and phase subtraction processing are performed to obtain the joint time series characteristic sequence of fused radar scattering and InSAR phase. S3. Perform phase noise assessment and weighted fusion on the joint temporal feature sequence to separate the atmospheric delayed phase and the true deformation phase, and obtain the net deformation phase sequence after removing the atmospheric influence. S4. Perform phase consistency quality control and phase unwrapping on the net deformation phase sequence to obtain a continuous and reliable cumulative deformation sequence. S5. Conduct correlation analysis between the cumulative deformation sequence and the amplitude attenuation characteristics of radar reflection signals, and construct a deformation mechanism correlation feature matrix; S6. Perform deviation calculation and secondary quality control on the deformation mechanism correlation feature matrix to obtain the updated deformation mechanism correlation feature matrix; S7. The updated deformation mechanism associated feature matrix is ​​spatiotemporally fused with the real-time differential interferometric image group to extract the deformation rate and phase gradient changes at the current moment, and generate a set of early warning indicators for local micro-deformation risks in deep foundation pits.

[0007] Secondly, this application provides a deep foundation pit micro-deformation InSAR remote sensing system, the system comprising: The data acquisition module is used to simultaneously collect continuous reflection signals and differential interferometric image sets of the deep foundation pit area through ground radar and InSAR satellite, and obtain a highly stable set of point targets and the original scattering characteristic sequence. The correction and fusion module is used to perform registration residual correction and phase subtraction processing based on the point target set and the original scattering characteristic sequence to obtain the joint time series feature sequence of fused radar scattering and InSAR phase. The noise reduction and separation module is used to perform phase noise assessment and weighted fusion on the joint temporal feature sequence, separate the atmospheric delayed phase from the true deformation phase, and obtain the net deformation phase sequence after removing the atmospheric influence. The unwrapping integration module is used to perform phase consistency quality control and phase unwrapping processing on the net deformation phase sequence to obtain a continuous and reliable cumulative deformation sequence. The matrix construction module is used to perform correlation analysis between the cumulative deformation sequence and the amplitude attenuation characteristics of radar reflection signals, and to construct a deformation mechanism correlation feature matrix. The deviation update module is used to perform deviation calculation and secondary quality control on the deformation mechanism correlation feature matrix to obtain the updated deformation mechanism correlation feature matrix. The early warning output module is used to perform spatiotemporal fusion of the updated deformation mechanism associated feature matrix with the real-time differential interferometric image group, extract the deformation rate and phase gradient changes at the current moment, and generate a set of early warning indicators for local micro-deformation risks in deep foundation pits.

[0008] Thirdly, this application provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the aforementioned InSAR remote sensing method for micro-deformation of deep foundation pits.

[0009] Compared with the prior art, the beneficial effects of the technical solution of this application are at least as follows: 1. By integrating ground-based radar and InSAR remote sensing technologies, high-precision micro-deformation monitoring of deep foundation pit areas can be achieved over a large area, with high real-time performance. Compared with traditional monitoring methods, it can continuously track the deformation process in a dynamic environment, promptly detect minute deformation changes, and improve monitoring accuracy and real-time performance.

[0010] 2. By performing phase subtraction, atmospheric delay separation, and weighted fusion processing, the influence of atmospheric delay is removed, significantly reducing interference caused by atmospheric changes and signal noise, thereby improving the quality and reliability of deformation data and the accuracy of monitoring results.

[0011] 3. This invention can not only detect the micro-deformation of deep foundation pits, but also classify the deformation risk of different areas by analyzing the deformation rate along the line of sight and the local phase gradient change rate. This technology enables foundation pit monitoring to go beyond point data and provide a more comprehensive spatial and temporal deformation risk assessment, achieving multi-dimensional risk warning and ensuring timely detection and early warning of potential risks.

[0012] 4. By introducing automated analysis based on data fusion, monitoring parameters can be automatically adjusted under different deep foundation pit construction environments, demonstrating strong adaptability and intelligence. Even in complex geological and construction environments, it can still provide accurate deformation monitoring and early warning services, improving the system's stability and reliability.

[0013] 5. The deep foundation pit micro-deformation monitoring and early warning system of the present invention can operate automatically, reducing manual intervention and lowering reliance on manual operation and maintenance costs. Through automated data acquisition, processing, and early warning, monitoring efficiency is significantly improved, and the impact of human factors on monitoring results is reduced. Attached Figure Description

[0014] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 This is a flowchart of an InSAR remote sensing method for micro-deformation of deep foundation pits according to this application; Figure 2 This is a schematic diagram of a scatter plot illustrating the correlation analysis between deformation and amplitude attenuation in an embodiment of this application. Figure 3 This is a schematic diagram of the spatial distribution of deformation and amplitude attenuation in an embodiment of this application; Figure 4 This is a schematic diagram of the structure of a deep foundation pit micro-deformation InSAR remote sensing system according to this application. Detailed Implementation

[0016] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a particular order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms “comprising” or “having,” and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0017] For ease of understanding, the specific process of the embodiments of this application is described below. Figure 1 The diagram shows a flowchart of an InSAR remote sensing method for micro-deformation of deep foundation pits provided by this invention. The flowchart specifically includes the following steps: S1. By simultaneously acquiring continuous reflection signals and differential interferometric image sets of the deep foundation pit area through ground radar and InSAR satellite, a highly stable set of point targets and original scattering characteristic sequences are obtained.

[0018] In one specific embodiment, the process of performing step S1 may specifically include the following steps: For the retaining structure and soil surface in the deep foundation pit area, the coherence coefficient of each pixel in multiple time phases is calculated based on the differential interferometric image group. The coherence coefficient is compared with a preset coherence threshold, and pixels with a coherence coefficient higher than the preset coherence threshold are selected as candidate permanent scatterers. Candidate permanent scatterers are validated over time, and surface permanent scatterers are identified based on the validation results to form a set of point targets; For each target point in the set of point targets, extract the radar reflection intensity value and radar phase value of the continuous time series from the continuous reflection signal; The radar reflection intensity value and radar phase value at each time point are aligned and integrated in chronological order to form the original scattering characteristic sequence.

[0019] Specifically, for the retaining structures (such as concrete pile walls and retaining walls) and soil surfaces in the deep foundation pit area, ground-based radar and InSAR satellites are simultaneously activated to acquire data, obtaining continuous reflection signals from the ground-based radar and differential interferometric image sets from the InSAR satellite. For example, the ground-based radar continuously transmits microwave signals at a sampling frequency of 10Hz and receives reflected signals, forming a continuous reflection signal stream covering the entire deep foundation pit area and a 50-meter radius around it; the InSAR satellite, following a preset orbit, images the area every three days, acquiring multiple sets of differential interferometric image sets with an image resolution of 3 meters to ensure the capture of subtle features of the retaining structures and soil surfaces.

[0020] Based on the acquired differential interferometric image set, for each pixel in the image, its coherence coefficient is calculated across multiple temporal phases (e.g., 12 consecutive temporal phases). The coherence coefficient is used to quantify the similarity and stability of the signal between two SAR images acquired at different times for the same pixel. The coherence coefficient is calculated using a complex signal correlation algorithm. For example, the coherence coefficient of each pixel across multiple consecutive temporal phases is calculated and then summed using the complex signal correlation algorithm. The specific formula is as follows: , in, S is the coherence coefficient of the i-th pixel. i (t) represents the complex image signal of the i-th pixel at time phase t. Indicates the following The complex conjugate of the complex image signal of the i-th pixel in time phase. Indicates signal S i The square of the amplitude of (t), Indicates the following The square of the amplitude of the complex image signal at the i-th pixel in time phase. Preferably, The value is 1. The coherence coefficient ranges from 0 to 1. A higher value indicates that the signal of the pixel is more consistent at different times and has stronger stability.

[0021] A permanent scatterer is a point target whose reflection characteristics are very stable across multiple time phases. These are typically structural targets or objects with high reflectivity on the ground. Candidate permanent scatterers are pixels identified in the initial screening stage based on a relatively lenient criterion that may possess stable scattering characteristics. In deep foundation pit monitoring scenarios, since the surfaces of retaining structures (such as metal supports and concrete surfaces) typically provide stable reflection, while loose soil or temporary stockpiles exhibit unstable reflection, setting a coherence threshold for screening is practically meaningful. The preset coherence threshold is adjusted according to the environment of the deep foundation pit. For example, if the foundation pit is located in a dry urban area with relatively stable operating conditions, the preset coherence threshold is set to 0.75; if the foundation pit is located in a rainy area or an area with significant variations in soil moisture content, the preset coherence threshold is adjusted to 0.65. The calculated coherence coefficients of each pixel are compared with the corresponding preset coherence thresholds. Pixels with coherence coefficients higher than the preset thresholds are selected as candidate permanent scatterers because their scattering characteristics exhibit preliminary stability over time.

[0022] The candidate permanent scatterers selected above may include some temporary or incidentally stable points, such as vehicles that have been parked briefly, reflections from accumulated water surfaces, or temporary debris. Therefore, time-series verification is required to confirm their permanence. Time-series verification checks the stability and consistency across multiple time phases to ensure that the selected scatterers do not exhibit significant fluctuations over time. The verification process involves checking whether the scatterer maintains high coherence consistently across multiple consecutive time phases (e.g., at least 24 time points, covering a complete construction phase or a one-month period) in an interferometric image set. If it does not, or if the number of times its coherence coefficient exceeds a preset coherence threshold reaches a preset threshold, and the fluctuation amplitude of its coherence coefficient is less than a preset fluctuation threshold, then it is determined to be a permanent surface scatterer. For example, the verification period is set to 6 months. The coherence coefficient changes of each candidate permanent scatterer in 24 temporal images over 6 months are continuously tracked. If the coherence coefficient of a candidate permanent scatterer remains above a preset threshold in at least 20 of the 24 temporal phases, and the fluctuation range of the coherence coefficient does not exceed 0.1, then the candidate permanent scatterer is confirmed as a surface permanent scatterer. If the coherence coefficient of a candidate permanent scatterer is below the preset threshold in more than 5 of the 24 temporal phases, or the fluctuation range exceeds 0.1, then it is determined to be a temporary scatterer and is removed. All surface permanent scatterers that have passed time-series verification are integrated to form a highly stable set of point targets. This set includes the three-dimensional coordinates of each surface permanent scatterer and its corresponding temporal coherence records.

[0023] Ground-based radar continuously acquires data at a high sampling rate, thus obtaining a high-time-density observation sequence for each point target. For each target point in the target set, the radar reflection intensity and radar phase values ​​at each time point in the continuous reflection signal acquired by the ground radar are extracted. The radar reflection intensity reflects the energy intensity of the echo signal from the point target, typically the square of the reflected signal amplitude. The radar phase value is the instantaneous phase observation obtained from the continuous reflection signal, ranging from [-π, π] in radians. It reflects the total phase change along the signal propagation path from the ground radar to the target point, including components such as target deformation, atmospheric disturbance, and system noise. Based on the intensity and phase values ​​extracted from the continuous reflection signal, the original scattering characteristic time-series data for each target point are formed.

[0024] The radar reflection intensity value and radar phase value of each target point are aligned and integrated in chronological order, with the timestamp accurate to the millisecond level, to ensure that the radar reflection intensity value and radar phase value at the same time point correspond one-to-one, forming the original scattering characteristic sequence. The original scattering characteristic sequence is a multivariate time series, and each element of it contains three data items: timestamp, radar reflection intensity value, and radar phase value.

[0025] By simultaneously acquiring continuous reflection signals from ground radar and differential interferometric image groups from InSAR satellites, the technical problems of incomplete data coverage and insufficient spatiotemporal resolution of single monitoring equipment have been solved. The high-frequency sampling of ground radar ensures the temporal continuity of the data, while the wide-area imaging of InSAR satellites ensures the spatial coverage of the data. The combination of the two enables comprehensive and high-density monitoring of deep foundation pit areas, effectively improving the stability of point targets and the reliability of monitoring results.

[0026] S2. Based on the point target set and the original scattering characteristic sequence, registration residual correction and phase subtraction processing are performed to obtain the joint time series characteristic sequence of fused radar scattering and InSAR phase.

[0027] In one specific embodiment, the process of performing step S2 may specifically include the following steps: The differential interferometric image group was spatially registered and the residual was corrected using the master-slave image registration method to obtain the registered differential interferometric image group. Based on the reference ellipsoid model and digital elevation model, orbital error and terrain phase subtraction processing is performed on the registered differential interferometric image group to obtain the corrected InSAR phase sequence; Based on the corrected InSAR phase sequence, the temporal phase stability index is calculated. At the same time, the original radar amplitude sequence corresponding to each point target is extracted from the original scattering characteristic sequence. Based on the temporal phase stability index, the original radar amplitude sequence and the corrected InSAR phase sequence are fused to generate a joint temporal feature sequence.

[0028] Specifically, in the monitoring of micro-deformation in deep foundation pits, there are spatial and temporal inconsistencies between InSAR satellite imagery and ground radar data. Direct fusion will introduce registration errors and phase interference, affecting the accuracy of deformation extraction. Therefore, spatial alignment and phase purification processing of multi-source data are necessary.

[0029] In the InSAR processing flow, the differential interferometric image set is generated from multiple synthetic aperture radar images through differential interferometry, containing phase information of the same deep foundation pit area at different time phases. Due to factors such as satellite orbital offset and attitude jitter, there may be sub-pixel-level spatial offsets between images at different time phases, i.e., registration residuals. To eliminate this residual, the image with the complete coverage of the deep foundation pit area and the best signal quality in the differential interferometric image set is selected as the master image, and the remaining images are used as auxiliary images. During the registration process, using the pixel coordinates of the master image as a reference, for each pixel of the auxiliary image, at least a preset number (e.g., 200) of corresponding feature points (such as corner points of the deep foundation pit retaining structure, fixed markers on the soil surface) are extracted using the SIFT feature matching algorithm. The coordinate deviation of the feature points in the master and auxiliary images is calculated, and the auxiliary images are geometrically transformed to eliminate the spatial registration residuals caused by imaging geometric differences. Preferably, a quadratic polynomial transformation model is used to perform geometric transformation on the auxiliary images to complete the spatial registration residual correction and output the registered differential interferometric image set. The specific formula of the quadratic polynomial transformation model is: , , Where (x3, y3) are the pixel coordinates of the auxiliary image, (x3', y3') are the registered coordinates, and p0~p5 and q0~q5 are polynomial coefficients. The coefficients p0~p5 and q0~q5 are solved using the least squares method to complete image resampling and alignment, outputting the registered differential interferometric image group. This step solves the phase error problem caused by the spatial positional offset between the primary and auxiliary images, ensuring that pixels in images from different time phases correspond to the same geographic coordinates.

[0030] In differential interferometric phase analysis, in addition to deformation phase, there are also orbital error phases caused by satellite orbit inaccuracies and terrain phases caused by topographic relief. To extract the true deformation signal, these interference terms need to be subtracted. For the spatially registered differential interferometric image group, an ideal Earth surface is simulated using a reference ellipsoid model, and combined with a high-precision digital elevation model, the phase contributions caused by satellite orbital errors and topographic relief are subtracted respectively, resulting in a corrected InSAR phase sequence containing only deformation information, atmospheric delay, and noise. The reference ellipsoid model uses the WGS-84 ellipsoid (semi-major axis a = 6378137 meters, flattening f = 1 / 298.257223563), and the digital elevation model uses regional topographic data with a resolution of 1 meter, including information such as the top elevation of the deep foundation pit retaining structure and the soil surface elevation. First, the orbital error phase is calculated: the satellite orbital parameters corresponding to the main and auxiliary images are obtained through satellite ephemeris data, and a linear orbital error model (formula: ...) is used. ,in, Let k0~k2 be the orbital error phase, k1~k2 be the orbital error coefficients, and (x1, y1) be the image pixel coordinates. By fitting the orbital error, the orbital error phase sequence is obtained. Then, the terrain phase is calculated: based on the reference ellipsoid model and digital elevation model, the terrain phase of each pixel is calculated using the InSAR geometric relationship formula, which is: ,in, Where Bl is the terrain phase, Bl is the vertical baseline of the satellite orbit, and h is the elevation value of the target point in the digital elevation model. Where is the radar wavelength, and R is the slant range from the satellite to the target point. The radar incident angle is the angle between the radar line-of-sight direction and the normal direction of the target point. Finally, the orbital error phase and terrain phase are sequentially subtracted from the phase of the registered differential interferometric image group to obtain the corrected InSAR phase sequence. This step solves the problem of interference from orbital offset and terrain undulation on the InSAR phase, and retains effective phase information including the micro-deformation of the deep foundation pit.

[0031] The temporal phase stability index is used to quantify the phase fluctuation of each point target in the time dimension, reflecting the reliability of that point as a deformation monitoring point. It is calculated as follows: for each point target in the corrected InSAR phase sequence, the phase standard deviation is calculated over multiple preset time phases (e.g., 12 consecutive time phases). This phase standard deviation is used as the temporal phase stability index; the smaller the index value, the stronger the temporal phase stability and the lower the susceptibility to noise. Simultaneously, from the original scattering characteristic sequence, the radar reflection intensity value (i.e., amplitude value) of each point target in the corresponding time phase is selected and arranged in chronological order to form a sequence, generating the original radar amplitude sequence. For example, if the radar reflection intensity values ​​of a point target in 12 time phases are 12.3dB, 12.5dB, ..., 12.6dB, the resulting sequence is the original radar amplitude sequence for that point target.

[0032] The corrected InSAR phase is weighted based on the temporal phase stability index and then combined with the original radar amplitude sequence to generate a joint temporal feature sequence. The purpose of fusing the original radar amplitude sequence and the corrected InSAR phase sequence is to combine the high sensitivity of radar amplitude to surface physical states with the high-precision detection capability of InSAR phase to deformation, constructing a multi-dimensional temporal feature vector for subsequent analysis. For each point target, its temporal phase stability index is used in the weighted fusion process; points with higher stability (i.e., smaller index values) have greater weight in the fusion. For example, the weighting coefficient w is defined as: Wherein, τ is a scale parameter, set according to the accuracy requirements of deep foundation pit monitoring, with an exemplary value of 0.1 to 0.3 rad. The smaller the value of τ, the higher the sensitivity of the weight to the phase stability index. This is a temporal phase stability index. Then, the feature vector Fj of each phase in the joint temporal feature sequence consists of the weighted phase and amplitude: Fj = [ [Aj], where, Let Aj be the corrected InSAR phase for the j-th time phase, and Aj be the original radar amplitude for the j-th time phase. The above operation is performed on all time phases to generate a joint temporal feature sequence {F1, F2, ..., FN} for each point target. This sequence simultaneously contains deformation phase information and scattering intensity information, and enhances the contribution of high-reliability point targets through stability weighting while suppressing the influence of noise points. In deep foundation pit monitoring scenarios, this fusion method helps identify points with stable phases but significantly attenuated amplitudes. These points may correspond to early risks such as cracking of the retaining structure or local soil instability, thus solving the technical problem that relying solely on phase or amplitude data may miss weak precursors of deformation.

[0033] S3. Perform phase noise assessment and weighted fusion on the joint time series feature sequence to separate the atmospheric delay phase and the true deformation phase, and obtain the net deformation phase sequence after removing the atmospheric influence.

[0034] In one specific embodiment, the process of performing step S3 may specifically include the following steps: For each point target in the joint temporal feature sequence, the temporal coherence index of each point target is calculated based on the InSAR phase sequence; A coherence weighting method is used to fuse multi-temporal InSAR phase sequences to obtain noise-suppressed fused phase sequences; For the fused phase sequence, the atmospheric delayed phase component is separated by spatial low-pass filtering and time series analysis; The net deformation phase sequence is obtained by subtracting the atmospheric delay phase component from the InSAR phase in the joint time series feature sequence.

[0035] Specifically, phase noise assessment and weighted fusion of the joint temporal feature sequences aim to address the core issue of severe interference between InSAR phase signals and atmospheric delay, system noise, and decoherence effects in deep foundation pit micro-deformation monitoring. Atmospheric delay phase exhibits spatial correlation and slow temporal variation; its magnitude often masks the true micro-deformation signals generated by foundation pit excavation, support structure deformation, and soil displacement, leading to reduced reliability of monitoring results and delayed risk warnings. Therefore, it is essential to effectively separate and subtract the atmospheric component from the observed phase.

[0036] For each point target in the joint temporal feature sequence, the temporal coherence index of each point target is calculated based on the corrected InSAR phase sequence. The temporal coherence index is used to quantify the correlation and stability of the phase signals of the point target across multiple time phases. The calculation uses a complex signal correlation algorithm, and the formula is as follows: ,in, The temporal coherence index ranges from [0, 1]. A higher temporal coherence index indicates that the phase of the point target is more stable in time and less affected by noise, changes in vegetation cover, or drastic changes in soil moisture. M represents the number of time phases involved in the calculation, and its value is determined based on the monitoring cycle requirements of deep foundation pits. For example, M is set to 15, meaning there are 15 consecutive time phases. P k Let be the InSAR complex phase signal of the k-th time phase (in the form c_k + jd_k, where c_k is the real part, d_k is the imaginary part, and j is the imaginary unit). The complex conjugate of the complex phase signal at time k+1 is... This is the squared magnitude of the complex phase signal at time k. This step assigns a quantitative reliability label to each monitoring point, solving the problem of not being able to distinguish between high-quality and low-quality phase observation data in subsequent processing, and providing a basis for reliability-based data fusion.

[0037] A coherence-weighted method is employed to fuse multi-temporal InSAR phase sequences, obtaining a noise-suppressed fused phase field sequence. This processing is performed independently for each temporal phase, aiming to integrate phase observations of all point targets in the spatial domain and utilize data from high-reliability points to suppress noise contributions from low-reliability points. The core logic of weighted fusion is to assign greater weights to point targets with higher temporal coherence indices in the spatial fusion, thereby enhancing the contribution of reliable signals and suppressing noise interference.

[0038] Based on the temporal coherence index, a coherence-weighted vector summation method is used to achieve fusion. For example, for the m1th time phase, there are N1 point targets within the deep foundation pit retaining structure area, and the corrected InSAR phase value of the n1th point target is... Its temporal coherence index is η n1The fusion phase field of this phase Calculated in the following way: Where β is the weight amplification factor (i.e., the coherence enhancement index), and for example, β is set to 3, which is used to amplify the weight ratio of high coherence point targets. The arg{·} operation represents taking the principal argument of the complex number. Let x and y be the complex exponential phase signal of the n1-th target at time m1, where x and y are the planar geographic coordinates of the deep foundation pit monitoring area. This calculation essentially treats the phase of each point as a unit complex vector, using the β power of its coherence index as weights for spatial domain vector superposition. The observation direction of high-coherence points dominates the synthesis result, while the noise directions of low-coherence points cancel each other out or are weakened. This operation is performed on all M1 time phases to generate a fused phase field sequence. , , ..., This step significantly suppressed random phase noise through spatial domain statistical fusion, generating a sequence with a higher signal-to-noise ratio that better reflects the common phase changes in the region (mainly large-scale atmospheric delay). This created conditions for subsequent separation of atmospheric components and solved the technical problems of large phase noise interference at single measurement points and difficulty in extracting regional atmospheric features.

[0039] For the fused phase field sequence, the atmospheric delay phase component is separated through spatial low-pass filtering and time series analysis. Spatial low-pass filtering is used to extract the spatially gently distributed atmospheric delay signal from the fused phase sequence. Atmospheric delay exhibits low-frequency variations in space, while the micro-deformation of deep foundation pits mostly shows localized high-frequency variations. This spatial frequency domain difference provides the basis for separation. For example, a Gaussian filtering algorithm is used, with the following filtering kernel function: Where G(x2,y2) is the weight of the filter kernel at coordinate (x2,y2). As the standard deviation for filtering, and considering the area of ​​the deep foundation pit (assuming a side length of 100 meters), we take, for example, the standard deviation. =5 pixels (pixel resolution 3 meters, corresponding to an actual spatial distance of 15 meters). During the filtering process, this kernel function is used to perform convolution operations on the fused phase sequence to obtain the spatial low-frequency phase component. Although the low-frequency spatial component has been filtered to remove local high-frequency deformation noise, it may still contain the following components: 1) spatially smooth but temporally coupled residual signals; 2) the true atmospheric delay phase dominated by long-period atmospheric variations (such as seasonal temperature and humidity fluctuations); and 3) possible low-frequency orbital residual errors or systematic drift. Further time-series analysis is needed to separate the gradual component characterizing the atmospheric delay from this component.

[0040] The core objective of time series analysis is to establish a model of the temporal variation of the spatial low-frequency phase component and extract a slowly varying trend term consistent with atmospheric delayed physical processes. One specific implementation involves using a model combining polynomial and periodic functions to analyze the time series of the spatial low-frequency phase component at each spatial point. , ,..., We then perform a least-squares fit. The model can be expressed as: , Where R1 is the polynomial order (e.g., 2), used to capture the linear / nonlinear trends of atmospheric delay; S1 is the number of periodic terms considered (e.g., 1), corresponding to the annual cycle; Ts is the cycle length, in units of the number of time phases; and a r1 b s1 and c s1 The coefficients to be fitted are denoted as . For example, if the revisit period of an InSAR satellite is 12 days, then the annual period corresponds to Ts = 365 / 12 ≈ 30 time phases. This is achieved by minimizing the objective function. Solve for the coefficients to be fitted. The fitted coefficients are obtained. This is the atmospheric delay phase estimate of the spatial point at time m1.

[0041] This model simultaneously captures both the trend and periodic fluctuations of atmospheric delay, such as the regular changes in atmospheric water vapor content caused by seasonal changes. In scenarios involving deep foundation pit monitoring spanning months or even years, such periodic signals are particularly significant. If not separated, their cumulative effect may be misjudged as slowly developing soil creep or support structure creep. Preferably, to improve the robustness of the estimation, outlier detection and removal can be performed on the spatial low-frequency phase component sequence before fitting. For example, the moving median and standard deviation of the sequence are calculated, and phases deviating from the median by more than three times the standard deviation are marked as outliers and ignored during fitting. This can reduce the local instantaneous interference caused by sudden weather events (such as short-duration heavy rainfall) on atmospheric phase estimation.

[0042] The atmospheric delay phase component is subtracted from the InSAR phase in the joint temporal feature sequence to obtain the net deformation phase sequence. Specifically, for each point target at each time phase, the atmospheric delay phase component is subtracted from the corresponding corrected InSAR phase in the joint temporal feature sequence. The net deformation phase values ​​of all time phases are arranged chronologically to form the net deformation phase sequence for that point target. The net deformation phase sequences of all point targets are then integrated to obtain the net deformation phase sequence covering the deep foundation pit area. This step directly removes the interference of atmospheric delay on the phase signal, solving the technical problem of deformation monitoring data distortion caused by atmospheric factors. The final output net deformation phase sequence contains only the true micro-deformation information of the deep foundation pit retaining structure and soil, providing a clean data source for subsequent cumulative deformation extraction.

[0043] S4. Perform phase consistency quality control and phase unwrapping on the net deformation phase sequence to obtain a continuous and reliable cumulative deformation sequence.

[0044] In one specific embodiment, the process of performing step S4 may specifically include the following steps: For the group of point targets within the enclosure structure area, the standard deviation of the phase sequence corresponding to each point target is calculated based on the net deformation phase sequence, which serves as the phase consistency quality control index for the point targets. When the phase consistency quality control index of any target point is higher than the preset index threshold, jump point detection is performed on the phase sequence corresponding to any target point to identify possible abrupt changes or abnormal data during the deformation process. By employing the minimum cost flow method or statistical testing method, differential interference phase unwrapping is performed on the detected phase jump points to eliminate the influence of phase jumps and obtain the cumulative deformation sequence.

[0045] Specifically, after completing atmospheric delay phase subtraction to obtain the net deformation phase sequence, although the phase values ​​in this sequence have removed large-scale atmospheric interference, they still exist in the main value range of [-π,π] in a tangled form, and contain phase jumps caused by residual noise, local decoherence, or abrupt changes in true deformation. Therefore, phase consistency quality control needs to be implemented to screen reliable data, and phase untangling needs to be performed to recover the true continuous deformation.

[0046] The retaining structure of deep foundation pits is a rigid or semi-rigid structure, and its micro-deformation exhibits significant spatial continuity and temporal stability. The corresponding net deformation phase sequence should show low fluctuation characteristics, and the phase consistency index directly quantifies this degree of fluctuation. The smaller the index value, the more stable the phase sequence and the higher the data reliability. The calculation process is carried out independently for each point target. Based on the net deformation phase value of any point target in all monitoring time phases, the standard deviation of its corresponding phase sequence is calculated. The physical meaning of this standard deviation is to quantify the dispersion of the phase observation values ​​of any point target throughout the entire monitoring period. The smaller the standard deviation value, the smaller the phase fluctuation and the higher the consistency in the time dimension, and the lower the influence of residual noise or random decoherence. This step assigns a quantitative index characterizing the temporal observation stability of each point target, solving the technical problem of not being able to automatically identify and screen high-reliability deformation observation data sources from a large number of monitoring points, and avoiding the waste of effective data or the omission of abnormal data caused by directly processing all point targets uniformly.

[0047] The preset threshold values ​​need to be calibrated in conjunction with the accuracy requirements of deep foundation pit monitoring and the deformation characteristics of the retaining structure. An example value is 0.05 rad (corresponding to a deformation error of 0.22 mm). This threshold can be adaptively adjusted according to the type of retaining structure in the monitoring area and the resolution of satellite data. For example, the deformation fluctuation of the steel structure support system is slightly higher than that of the concrete structure, and the threshold can be relaxed to 0.06 rad. If the phase consistency quality control index is higher than the preset threshold, it is considered that there may be a jump point in the phase sequence at that point caused by phase entanglement, gross errors, or abrupt changes in actual deformation, and the jump point detection process needs to be initiated. The jump point detection is implemented using a sliding window method. For example, the window size is set to three time phases. Using the median of the phase data within the window as a benchmark, the deviation between the current time phase value and the median is calculated. When the deviation exceeds twice the standard deviation of the phase within the window, the time phase is marked as a candidate jump point. Trend fitting is performed on the phase data of the two time phases before and after the candidate jump point. If the slope of the fitted curve at the candidate jump point changes abruptly beyond a preset threshold (e.g., 0.1 rad / time phase), then the point is confirmed as a jump point. This step solves the technical problem of difficulty in identifying hidden abrupt changes in the net deformation phase sequence, avoids the distortion of subsequent phase unwrapping results caused by jump points, and ensures the temporal continuity of the data before unwrapping.

[0048] The InSAR phase corresponding to the micro-deformation of deep foundation pits is a tangled phase, with its value range limited to [-π, π]. However, the accumulation of actual deformation will cause the phase to exceed this range, resulting in phase tangling. The superimposed influence of transition points will further exacerbate the difficulty of untangling. The core of phase untangling is to solve for integer ambiguities and restore the tangled phase to a continuous absolute phase.

[0049] For point target sequences with phase consistency indices exceeding a threshold and detected abrupt changes, a time-series version of the minimum cost flow method is used for untangling. This method treats each monitoring phase of a single point target as a network node, and the phase difference between adjacent phases as an edge between nodes. The weight of the edge is set according to how close the adjacent phase difference is to π; the closer the phase difference is to π, the higher the weight (for example, the weight calculation formula is w=|Δφ| / π, where Δφ is the phase difference between adjacent phases). This transforms the problem of solving integer ambiguity into solving the minimum cost path problem on a time-series network. In specific implementation, the tangling phase of the first phase of the point target is used as the initial value of the starting node, and a time-series network containing L nodes (L is the number of monitoring phases) and L-1 edges is constructed. The edge weights corresponding to the detected abrupt changes are multiplied by a penalty coefficient (e.g., 1.5) to avoid the untangling process relying excessively on abnormal phase difference data. By iteratively calculating using a minimum-cost flow solver, an integer ambiguity sequence {K1, K2, ..., KL} that makes the overall phase sequence smoothest is obtained. The entanglement phase of each time phase is then superimposed with 2π×Kl to obtain a continuous unentangled absolute phase sequence. This method, through global optimization of the temporal phase relationship, solves the technical problems of traditional temporal unentanglement methods being sensitive to jump points and prone to overall unentanglement distortion due to local anomalies, ensuring that the unentanglement results conform to the temporal development characteristics of deep foundation pit deformation.

[0050] For sequences with excellent phase consistency and few abrupt changes (exemplarily, fewer than two abrupt changes), a time-series local unwrapping method based on statistical tests is adopted. For each detected abrupt change, the phase sequences of the preceding and following phases (e.g., three) are extracted as two independent samples. For example, if the abrupt change occurs at phase 10, phases 7-9 are taken as the preceding samples, and phases 11-13 as the following samples. The Mann-Whitney U test is used to test the significance difference between the two samples, with a significance level threshold of 0.05. If the p-value obtained by the test is <0.05, the abrupt change is confirmed as a true deformation abrupt change, and the phase data is retained and marked as a deformation feature point. The abrupt change amplitude is directly included in the phase difference calculation before and after the phase during unwrapping. If the p-value is >0.05, it is determined to be a noise spurious abrupt change, and the phase data is removed. The phase sequence is then completed using the phase data of the two phases before and after the abrupt change as a benchmark, and a linear interpolation method is used. For the processed complete phase sequence, the temporal differential accumulation method is used for unwrapping. The first time phase of the target at that point is used as the initial unwrapping value. The phase difference between adjacent time phases is calculated sequentially. If the absolute value of the phase difference is greater than π, compensation is added or subtracted by an integer multiple of 2π according to the phase change trend (add 2π if the phase is increasing, subtract 2π if it is decreasing). This is accumulated phase by phase to obtain a continuously unwrapped absolute phase sequence. This method does not rely on external reference points and can complete unwrapping using only single-point time-series data. It solves the technical problem of traditional unwrapping methods being unable to distinguish between real deformation abrupt changes and noise-induced spurious jumps. It suppresses noise interference, accurately preserves real deformation characteristics, and has higher computational efficiency.

[0051] After unwrapping, the continuous unwrapped phases are converted into cumulative deformations through the conversion relationship between phase and deformation. These cumulative deformation sequences for each point target are then arranged chronologically. The sequences of all point targets are integrated to form a cumulative deformation sequence covering the entire retaining structure area. For example, the core conversion relationship between unwrapped phase and cumulative deformation is: Deformation = Unwrapped Phase × Radar Wavelength / (4π). If the unwrapped phase sequence of a point target is 0.5 rad, 0.7 rad, 0.9 rad, ..., 2.3 rad, and the radar wavelength is 0.056 meters, the calculated cumulative deformations are 0.0022 meters, 0.0031 meters, 0.0040 meters, ..., 0.0102 meters, respectively. The integration of the cumulative deformation sequences of all point targets forms a cumulative deformation sequence covering the entire retaining structure area. This technique solves the core technical problems of the wrapped phase not directly reflecting the true cumulative deformation and the failure of unwrapping due to jump points. The output sequence directly provides core data support for subsequent deformation trend analysis and early warning.

[0052] S5. Correlation analysis is performed on the cumulative deformation sequence and the amplitude attenuation characteristics of radar reflection signal to construct a deformation mechanism correlation feature matrix.

[0053] In one specific embodiment, the process of performing step S5 may specifically include the following steps: Based on the cumulative deformation sequence and satellite geometric parameters, deformation decomposition projection is performed to obtain the projection results of the deformation components; From the projection results, the horizontal displacement component sequence of the retaining structure and the vertical heave component sequence of the soil are extracted respectively. Extract the radar reflection intensity value sequence corresponding to each point target from the original scattering characteristic sequence, and calculate the amplitude attenuation ratio sequence based on the radar reflection intensity value sequence; The horizontal displacement component sequence, the vertical bulge component sequence, and the amplitude attenuation ratio sequence are time-series aligned and correlated to establish the correspondence between deformation components and amplitude attenuation characteristics. Based on the correspondence and projection results, a deformation mechanism association feature matrix containing the correlation relationships of multiple deformation components is constructed.

[0054] Specifically, the correlation analysis between the cumulative deformation sequence and the amplitude attenuation characteristics of radar reflection signals aims to address the key issue in deep foundation pit micro-deformation monitoring where relying solely on single deformation measurement data is insufficient to determine the physical mechanism of deformation. The cumulative deformation sequence describes "how much" displacement occurred, but cannot directly explain "why" it occurred. In contrast, radar amplitude attenuation characteristics reflect changes in the scattering properties of the surface or structure, potentially correlated with deformation-inducing factors such as soil loosening, structural cracking, or material fatigue. By combining these two analyses, we can move from phenomenological observation to mechanism inference, providing a physical basis for risk assessment.

[0055] InSAR observations yield deformation along the satellite line of sight, a one-dimensional projection. To separate the horizontal and vertical displacements, which are more significant for engineering safety, decomposition using multi-field observations or a single field of view combined with a priori models is necessary. For example, if this application uses single satellite ascent or descent data, decomposition must be performed in conjunction with the geometric constraints of the deep foundation pit retaining structure deformation. For retaining pile walls perpendicular to the edge of the foundation pit, their deformation can be mainly decomposed into horizontal displacement perpendicular to the wall surface and vertical displacement parallel to the wall surface. Let the unit vector of the satellite line of sight be I = [le, ln, lu] (cosines of the east, north, and sky directions), and the unit vector of the retaining pile wall surface normal be Nv = [ne, nn, 0]. Then, the relationship between the line-of-sight deformation dlos and the wall surface normal horizontal displacement dh and vertical displacement dv can be approximately expressed as: dlos = dh(I⋅Nv) + dv⋅lu. This equation is a single-point underdetermined equation. To address this problem, a regional global least squares fitting method based on spatial smoothing priors is employed. Taking the smoothness of continuous deformation of the retaining structure as a prior assumption, the cumulative line-of-sight deformation dlos(t1) of all point targets within the defined retaining structure / soil region is jointly fitted. By minimizing the regional global fitting residuals, the time-series estimates of the horizontal displacement component dh(t1) and the vertical displacement component dv(t1) of each point target within the region at each time phase t1 are obtained. This step transforms one-dimensional observations into two-dimensional deformation components that better conform to engineering mechanics interpretations. It solves the technical problem that the physical meaning of line-of-sight deformation in InSAR is not intuitive and difficult to directly use for structural safety evaluation. Its accuracy depends on the degree of matching between the spatial smoothing prior constraints and the actual engineering scenario.

[0056] From the decomposition results, the horizontal displacement component sequence of the retaining structure and the vertical heave component sequence of the soil are extracted separately. The horizontal displacement component of the retaining structure is selected as the horizontal displacement component dh, perpendicular to the orientation of the retaining structure, as this displacement is directly related to the lateral deformation risk of the retaining structure and is a core indicator for deep foundation pit stability monitoring. The vertical heave component of the soil is directly selected as the vertical displacement component dv, which reflects the rebound or settlement state of the soil at the bottom of the pit during excavation. During the extraction process, based on the boundaries between the retaining structure area and the soil area defined in step S2, the decomposed two-dimensional deformation component sequence is matched regionally. The dh of all point targets within the retaining structure area is arranged in chronological order to form the horizontal displacement component sequence of each point target. After integration, the overall horizontal displacement component sequence of the retaining structure is obtained. Similarly, the dv of all point targets within the soil area is arranged in chronological order to form the vertical heave component sequence of each point target. After integration, the overall vertical heave component sequence of the soil is obtained. This step addresses the technical issue of mismatch between deformation components and the core focus of deep foundation pit monitoring, focusing on two key risk points: lateral deformation of the retaining structure and vertical heave of the soil, thereby improving the engineering application relevance of deformation data.

[0057] The original scattering characteristic sequence is the scattering signal record inherent in the SAR image data acquired in step S1. During extraction, the radar reflection intensity value of each point target in each monitoring time phase is matched with the timestamp of its spatial coordinates, and arranged chronologically to form a radar reflection intensity value sequence. The amplitude attenuation ratio sequence is calculated using the sliding window method. For example, the window size is set to 5 time phases, and the window is slid in a backward alignment manner (i.e., the characteristic value of each time phase is the calculation result with that time phase as the last moment of the window). The amplitude attenuation ratio of each time phase is the ratio of the average reflection intensity of the last 3 time phases within the current window to the average reflection intensity of the first 2 time phases. The amplitude attenuation ratio ranges from (0, +∞). A ratio less than 1 indicates a decreasing trend in radar reflection intensity, with a smaller ratio indicating more significant attenuation; a ratio greater than 1 indicates an increasing trend in reflection intensity. For example, if the reflection intensity values ​​of a point target in time phases 6 to 10 are 280, 275, 260, 250, and 245 respectively, then the amplitude attenuation ratio in time phase 10 is 0.907, indicating that the reflection intensity in this time phase is attenuated by 9.3% compared to the previous phase. The amplitude attenuation ratio in time phase 9 is calculated based on the reflection intensity values ​​in time phases 5 to 9. This technical feature solves the technical problem that the amplitude change of radar reflection signal is not correlated with deformation and cannot reflect changes in structural integrity. By quantifying the stability of the scatterer through the amplitude attenuation ratio, it provides a characteristic indicator for correlating deformation and structural state.

[0058] The horizontal displacement component sequence, vertical bulge component sequence, and amplitude attenuation ratio sequence were time-series aligned and correlated to establish the correspondence between deformation components and amplitude attenuation characteristics. Time-series alignment was based on the timestamps of the monitoring phases, ensuring a one-to-one correspondence among the three sequences in the same time dimension. If data was missing in a certain phase, linear interpolation was used to complete it. Correlation analysis used Pearson correlation coefficients to calculate the correlation coefficients between the horizontal displacement component and the amplitude attenuation ratio, and between the vertical bulge component and the amplitude attenuation ratio, respectively. The correlation coefficient ranged from -1 to 1; the closer the absolute value was to 1, the stronger the correlation, and a negative value indicated a negative correlation between the two variables. For example, the correlation coefficient between the horizontal displacement component and the amplitude attenuation ratio of a target point in the retaining structure was -0.87, indicating that the larger the horizontal displacement, the smaller the amplitude attenuation ratio, and the more significant the radar reflection intensity attenuation. The correlation coefficient between the vertical bulge component and the amplitude attenuation ratio of a target point in the soil was -0.72, indicating that the more pronounced the vertical bulge, the smaller the amplitude attenuation ratio, and the more severe the radar reflection intensity attenuation. This step solves the technical problem that isolated analysis of deformation characteristics and scattering properties cannot reveal the intrinsic relationship between the two, and clarifies the corresponding law between deformation development and changes in the stability of the scatterer.

[0059] Based on the correspondence and projection results, a deformation mechanism correlation feature matrix containing various deformation component relationships is constructed. The row dimension of the feature matrix represents the point target number within the retaining structure and soil area, while the column dimension represents the core feature indicators, including but not limited to the spatial coordinates of the point targets, the mean value of the horizontal displacement component of the retaining structure, the mean value of the vertical heave component of the soil, the mean value of the amplitude attenuation ratio, the correlation coefficient between horizontal displacement and amplitude attenuation ratio, the correlation coefficient between vertical heave and amplitude attenuation ratio, the temporal change rate of horizontal displacement, and the temporal change rate of vertical heave. Among these, the various mean values ​​and change rate indicators are calculated from the time series of the point targets over the entire monitoring period, and the correlation coefficients are the global statistical values ​​of the deformation component and amplitude attenuation ratio sequences of the point targets. This feature matrix, centered on spatial point targets, integrates the spatial location, deformation characteristics, amplitude attenuation characteristics, and their correlation. It solves the technical problem that traditional monitoring only focuses on deformation data and lacks in-depth analysis of deformation mechanisms. At the same time, it avoids information redundancy of feature indicators, providing structured and analyzable feature data support for revealing the intrinsic relationship between deep foundation pit deformation and structural stability and achieving accurate early warning.

[0060] Based on the above correspondences and projection results, a deformation mechanism correlation feature matrix containing correlations of multiple deformation components is constructed. For a visual demonstration of the technical effect of the correlation analysis in step S5 of this invention, please refer to [link to relevant documentation]. Figure 2 and Figure 3 The diagram showing the correlation between deformation and amplitude attenuation is shown. Figure 2 A scatter plot showing the correlation between deformation and amplitude attenuation. Figure 3 This is a spatial distribution diagram showing the correlation between deformation and amplitude attenuation.

[0061] exist Figure 2 In the graph, the horizontal axis represents the horizontal displacement component of the retaining structure (unit: mm), and the vertical axis represents the radar amplitude attenuation ratio (dimensionless). Each scatter point represents a point target (permanent scatterer) within the deep foundation pit area. The scatter point color gradually changes according to the magnitude of the horizontal displacement (the color spectrum ranges from blue through cyan, yellow to red, indicating that the displacement increases from negative to positive), while the scatter point size remains constant. The black dashed line in the graph is the trend line obtained through linear regression, and its negative slope clearly indicates a significant negative correlation between horizontal displacement and the amplitude attenuation ratio; that is, the larger the displacement of a point target, the more obvious the amplitude attenuation of its radar reflection signal. Three typical monitoring points are marked in the graph: point A is the location with the maximum displacement at the top of the retaining pile, point B is the location of obvious cracking in the support structure, and point C is a typical monitoring point. This scatter plot visually verifies the effectiveness of the correlation analysis in step S5 of this invention, with a correlation coefficient reaching over -0.8, indicating a clear correspondence between deformation development and changes in the stability of the scatterer. Figure 3 In the middle, it shows the relationship with Figure 2The actual spatial distribution of the corresponding point targets within the deep foundation pit. The black rectangle in the background indicates the outline of the deep foundation pit. Each circular marker in the figure represents a point target, and the color of the marker also indicates the magnitude of the horizontal displacement (color spectrum and scatter plot on the left). Figure 1 The size (area) of the marker indicates the severity of the amplitude attenuation ratio—the more significant the attenuation (the smaller the ratio), the larger the marker size. The figure clearly shows that in specific areas of the foundation pit (marked as "high-risk areas"), a large number of "large red markers" appear, indicating that this area simultaneously exhibits significant displacement (red, indicating large displacement) and severe amplitude attenuation (large size, indicating a small attenuation ratio), clearly indicating a high risk of structural instability. In contrast, the markers in "low-risk areas" are mostly small blue or green, indicating small displacement and slight attenuation.

[0062] Figure 2 and Figure 3 Together, they constitute a complete characterization of the technical effect of step S5 of the present invention: not only through statistical correlation ( Figure 2 ) quantitatively proved the intrinsic relationship between deformation and changes in scattering properties, and further demonstrated this through spatial mapping ( Figure 3 This connection is pinpointed to a specific engineering spatial location. The combination of these two aspects demonstrates that this invention can not only monitor "how much" displacement has occurred, but also, through amplitude attenuation characteristics, reveal the physical mechanisms underlying "why" displacement occurred (such as structural cracking, soil loosening, etc.), thus providing intuitive and reliable visual evidence for constructing a "deformation mechanism correlation feature matrix." This correlation feature matrix forms the data foundation for subsequent accurate risk warning.

[0063] S6. Perform deviation calculation and secondary quality control on the deformation mechanism correlation feature matrix to obtain the updated deformation mechanism correlation feature matrix.

[0064] In one specific embodiment, the process of performing step S6 may specifically include the following steps: The deformation feature values ​​of each target point in the deformation mechanism associated feature matrix are compared with the corresponding phase stability reference values ​​in the historical time series phase stability sequence to calculate the feature deviation value of each target point. When the feature deviation value exceeds the preset deviation threshold, the corresponding point target is marked as an abnormal point target; Based on the differential interferometric image group, a polynomial fitting process for orbital error removal is performed on each anomalous target to eliminate the influence of orbital error on phase data and obtain a corrected phase data sequence. Based on the corrected phase data sequence, the phase consistency index of each target point is obtained by calculating the phase change gradient between adjacent time nodes. Valid point targets with a phase consistency index less than a preset quality threshold are selected, and an updated deformation mechanism association feature matrix is ​​generated based on the valid point targets.

[0065] Specifically, the purpose of bias calculation and secondary quality control on the deformation mechanism correlation feature matrix is ​​to address the problem of deformation feature distortion in long-term deep foundation pit micro-deformation InSAR monitoring caused by accumulated residual systematic errors, abrupt changes in point target scattering characteristics, or abnormal data missed in previous processing. Although the deformation mechanism correlation feature matrix has integrated deformation and radar amplitude attenuation features, the basic data may still contain systematic biases that were not identified and eliminated in steps S1 to S5. Such biases will distort the true reflection of the deformation mechanism, thereby affecting the accuracy of subsequent risk warnings. Therefore, it is necessary to conduct bias diagnosis based on historical data and implement targeted correction and secondary screening of abnormal data to ensure the reliability of the feature matrix.

[0066] The historical time-series phase stability sequence is constructed based on the time-series coherence sequence (single-point full-time stability comprehensive scalar value) from step S3 and the phase consistency quality control index sequence (single-point time-series standard deviation sequence) from step S4. Specifically, the construction method involves extracting the statistical baselines of two types of indicators for each target point within the historical monitoring period, including the statistical mean and standard deviation of time-series coherence and the statistical mean and standard deviation of the phase consistency index sequence. These are integrated to form a unified historical stability reference benchmark, characterizing the historical stability baseline and natural fluctuation range of each target point. Deformation characteristic values ​​are selected from core indicators in the deformation mechanism correlation characteristic matrix, including the mean of the horizontal displacement component of the retaining structure, the mean of the vertical heave component of the soil, and / or the mean of the amplitude attenuation ratio. Characteristic deviation values ​​are calculated using a normalization method, with the following formula: , among which, Fd r Let Fv be the characteristic deviation value of the r-th target point. r Let be the deformation characteristic value of the r-th point target. and These represent the mean and standard deviation of the historical reference sequence for the r-th target point, respectively. This formula enables comparable quantification of the deviation values ​​of target points with different fluctuation characteristics, intuitively reflecting the degree of deviation of the current feature value relative to the historical stable baseline. In scenarios with multiple deformation feature values, it is only necessary to independently calculate the normalized deviation value for each deformation feature and then integrate it into a comprehensive feature deviation value through the arithmetic mean (i.e., first calculate the feature deviation value of a single feature, and then perform an arithmetic mean or weighted average integration). This retains the quantification results of single feature deviations and enables anomaly judgment of multiple feature dimensions through the comprehensive value, perfectly meeting the analysis needs of multiple core features such as horizontal displacement of the retaining structure, vertical heave of the soil, and amplitude attenuation ratio in deep foundation pit monitoring. This step solves the technical problem in deep foundation pit monitoring where the lack of long-term stability references makes it impossible to effectively identify abnormal abrupt changes in deformation features, providing a clear quantitative basis for subsequent targeted treatment of anomalies.

[0067] Preferably, in the multi-dimensional feature collaborative analysis scenario, the historical time-series phase stability sequence is a statistical set of multiple features of the point target within the historical monitoring period, including a multi-feature mean vector (including the mean of time-series coherence, the mean of phase consistency index, the mean of horizontal displacement, etc.) and a multi-feature covariance matrix. This not only characterizes the historical stable baseline and natural fluctuation range of each point target's individual features but also reflects the historical correlation patterns between features. Deformation feature values ​​are multi-dimensional feature vectors, including the mean of the horizontal displacement component of the retaining structure, the mean of the vertical heave component of the soil, and / or the mean of the amplitude attenuation ratio, etc. These various dimensional features have inherent correlations in historical monitoring and need to be considered in the deviation calculation to avoid misjudgment of anomalies. The feature deviation values ​​are calculated using Mahalanobis distance for multi-feature normalization, with the formula: , among which, Fd r Let F be the characteristic deviation value of the r-th target point. r Let be the multidimensional deformation feature vector of the target at point r. Let r be the mean vector of the multidimensional historical features of the target at point r. Let r be the multi-order historical feature covariance matrix of the target point r. The formula is the inverse of the covariance matrix, with the superscript T representing matrix transpose. This formula automatically eliminates differences in feature dimensions and quantifies the correlation between features through the covariance matrix, achieving scientifically comparable quantification of deviation values ​​across multiple feature dimensions. It accurately reflects the overall deviation of the current multi-dimensional feature combination from historically stable correlation patterns, fundamentally avoiding misjudgments caused by the single-feature arithmetic mean method neglecting feature correlation.

[0068] A preset deviation threshold is determined based on the fault tolerance requirements of deep foundation pit monitoring, statistical significance, and engineering geological conditions. An exemplary value of 3 is used; however, in soft soil areas where foundation pit deformation fluctuates significantly, the threshold can be relaxed to 3.5. When the characteristic deviation value of a point target exceeds this threshold, its deformation characteristics are deemed to have become distorted due to uncorrected errors or drastic changes in the scatterer state. This point is then marked as an anomalous point target, and its spatial coordinates, anomalous feature type, and specific deviation value are recorded, forming an anomalous point target set. By setting objective statistical criteria, automated identification of anomalous point targets is achieved, solving the technical problems of low efficiency, inconsistent standards, and lack of quantitative discrimination standards for anomalous point targets in manual screening.

[0069] The differential interferometric image group consists of multi-phase phase images obtained by differential interferometry processing of the original SAR images from step S1. Orbital error is a systematic deviation introduced by satellite orbit determination error, which spatially manifests as a large-scale low-order polynomial trend and is the core factor causing distortion of the phase data of anomaly targets. For each anomaly target, the original phase values ​​of all monitoring phases are first extracted from the differential interferometric image group to form an original phase sequence. Then, a local spatial neighborhood (e.g., a local spatial neighborhood with a radius of 50 meters) is defined with the anomaly point as the center. All non-anomaly stable point targets within the neighborhood are selected as references, and a quadratic polynomial surface is fitted using the least squares method to characterize the spatial distribution of orbital error in this local area. Finally, the fitted orbital error values ​​are subtracted from the original phase values ​​of the anomaly point phase phase by phase to obtain the corrected phase data. Arranged in chronological order of the monitoring phases, the corrected phase data sequence is formed. This step specifically eliminates the residual orbital error of outlier targets, solving the technical problems of residual global orbital error in local areas and insufficient correction for special points in the early stage, improving the authenticity of outlier phase data, and providing the possibility for the "rescue" and reuse of outlier data.

[0070] The phase change gradient is used to quantify the smoothness and continuity of the corrected phase sequence in the time dimension, directly reflecting the effectiveness of the orbital error correction process. Adjacent time nodes are two consecutive monitoring phases. The change in the corrected phase value between adjacent time phases is calculated. The absolute values ​​of the phase changes of all adjacent time phases within the entire monitoring period of the point target are averaged to obtain the phase consistency index for that point target. The smaller the index value, the smaller the temporal fluctuation and the higher the consistency of the corrected phase sequence, and the stronger the data reliability. This step solves the technical problem of lacking a quantitative evaluation index for the effectiveness of the corrected phase data sequence, providing a core judgment basis for subsequent secondary screening of point targets.

[0071] A preset quality threshold is used, combined with the InSAR phase observation noise level and the time-series stationary characteristics of deep foundation pit micro-deformation. An example value of 0.1 rad is used. If the phase consistency index of anomaly point targets is less than this threshold, it is determined that the time-series consistency of its corrected phase data meets the requirements for deep foundation pit monitoring and is a valid point target. If the index is greater than or equal to this threshold, it is determined that its phase data still has significant time-series fluctuations and is an invalid point target, which is then removed. The scope of valid point target screening covers all anomaly point targets that have been corrected for track errors. At the same time, the original valid point targets that have not been marked as anomalies in the deformation mechanism association feature matrix are directly retained. After integrating all valid point targets, according to the dimensional structure and index system of the original feature matrix, the core indicators such as the spatial coordinates of the point targets, the mean of the horizontal displacement component of the retaining structure, the mean of the vertical heave component of the soil, the mean of the amplitude attenuation ratio, and the correlation coefficient between deformation and amplitude attenuation ratio are retained. The data of the corresponding rows of invalid point targets are removed, and an updated deformation mechanism association feature matrix is ​​generated.

[0072] This step achieves dynamic purification of feature matrix data through a closed-loop process of "identification-correction-verification-screening". It solves the core technical problems of the original feature matrix containing outlier targets, insufficient feature data accuracy, and a few low-quality data causing a decline in overall reliability, and significantly improves the data quality of the feature matrix associated with the deformation mechanism.

[0073] S7. The updated deformation mechanism associated feature matrix is ​​spatiotemporally fused with the real-time differential interferometric image group to extract the deformation rate and phase gradient changes at the current moment, and generate a set of early warning indicators for local micro-deformation risks in deep foundation pits.

[0074] In one specific embodiment, the process of performing step S7 may specifically include the following steps: The updated deformation mechanism associated feature matrix is ​​spatiotemporally aligned and fused with the latest acquired real-time differential interferometric image set to obtain the feature data field; The line-of-sight deformation rate of each target point at the current moment is calculated based on the feature data field. In the feature data field, the local region consisting of each target point and its spatial neighborhood is used as the analysis unit to calculate the phase gradient change rate within the corresponding local region. The deformation rate and phase gradient change rate along the line of sight are compared with preset rate thresholds and preset change rate thresholds, respectively. Based on the comparison results, the risk levels of multiple local areas in the deep foundation pit are classified. Based on risk level classification and spatial location information of local areas, a set of early warning indicators for local micro-deformation risks in deep foundation pits is generated, which includes risk level information of multiple local areas.

[0075] Specifically, the updated deformation mechanism correlation feature matrix is ​​spatiotemporally fused with a real-time differential interferometric image set. The aim is to address the core issues in deep foundation pit safety monitoring: delayed early warning information, the isolation between historical deformation mechanism knowledge and the latest observational data, and the inability to reflect instantaneous deformation dynamics. The updated deformation mechanism correlation feature matrix contains the historical deformation trends and structural state correlation patterns of effective point targets, but it lacks the ability to capture the current deformation situation. The real-time differential interferometric image set contains the latest phase change signals, but it is difficult to interpret the meaning of deformation in isolation. By fusing the two, the synergy between historical knowledge and real-time observation can be achieved, completing the leap from "historical deformation analysis" to "real-time risk perception."

[0076] The updated deformation mechanism association feature matrix includes core information such as the spatial coordinates of effective point targets, horizontal displacement trends, vertical uplift trends, amplitude attenuation ratios, and feature correlation coefficients. The spatial dimension adopts the geographic coordinate system of the deep foundation pit monitoring area. The real-time differential interferometric image set is a single-phase image obtained by differential interferometry processing of the latest raw SAR image acquired in step S1, containing the real-time phase observation values ​​of each pixel. The time dimension corresponds to the current monitoring time, and the spatial dimension adopts the geographic coordinate projection consistent with the feature matrix. Spatiotemporal alignment includes time alignment and spatial registration: time alignment ensures that the most recent historical phases covered by the feature matrix are continuous with the real-time image acquisition phases. For example, if the latest phase of the feature matrix is ​​day T1-1, the real-time image was acquired on day T1. Spatial registration uses the geographic coordinates of the point targets in the feature matrix as a reference, and resamples the real-time differential interferometric image. The registration error is controlled within 0.5 pixels to ensure that the coordinates of each point target completely overlap in the two sets of data. The fusion processing employs a combination of data layer overlay and feature layer correlation: for each point target, the multidimensional historical correlation feature vector in the feature matrix is ​​used as prior knowledge and combined with local spatial statistics such as the latest differential interferometric phase value and the phase standard deviation within a 3×3 window in the real-time image to form an extended feature set; for the blank area outside the effective point targets, Kriging interpolation is used, with the effective point targets as sample points, and a semi-variogram model is constructed based on the spatial distance between points and feature correlation to estimate the feature values ​​of the blank area, ultimately forming a feature data field that covers the entire deep foundation pit monitoring area, is spatially continuous, and has consistent spatiotemporal attributes. This technology solves the technical problems of historical deformation mechanism knowledge and the latest deformation observation data being isolated from each other, the feature matrix containing only discrete point information, and the inability to achieve continuous collaborative analysis across the entire domain, providing a unified data foundation for subsequent index calculations.

[0077] The line-of-sight deformation rate of each target point at the current moment is calculated based on the feature data field. Within the feature data field, the local region formed by each target point and its spatial neighborhood is used as the analysis unit to calculate the phase gradient change rate within the corresponding local region. The line-of-sight deformation rate reflects the instantaneous velocity of deformation development, and the formula is: , where v n2 Let n2 be the line-of-sight deformation rate of the target point n2. The radar wavelength (0.055 meters for example). The current temporal phase value of the n2th point target (obtained from real-time imagery). The phase value of the n2th point target at the previous moment (obtained from the phase history sequence associated with the feature matrix). This represents the time interval between consecutive phases (i.e., the satellite revisit period, such as 12 days); if only phase m2 is available, the short-term predicted value of the historical deformation trend in the characteristic matrix can be used instead. Perform the calculation.

[0078] The local region is divided using each target point as the center, defining a square region (or a 5×5 pixel window) with a preset side length (e.g., 30 meters) as the analysis unit. The phase gradient change rate is used to quantify the spatial inhomogeneity of the deformation field in the local region. The calculation process is as follows: First, the phase gradient components of all pixels within the window in the east-west and north-south directions are calculated using the Sobel operator. The formula is: , ,in, Let x and y be the differential interferometric phase values ​​of pixels in the real-time differential interferometric image, and x and y be the planar geographic coordinate components of the deep foundation pit monitoring area, thus obtaining the phase gradient magnitude; then, calculate the relative change between the current gradient magnitude and the historical average gradient magnitude, using the following formula: Where n3 is the local analysis unit number within the deep foundation pit monitoring area. This represents the mean of the historical gradient magnitudes of the local analysis cell numbered n3 (calculated from the historical phase field associated with the feature matrix). This refers to the phase gradient magnitude of the local analysis unit numbered n3 at the current monitoring time. This technical feature solves the technical problems of traditional time series analysis, which cannot provide the deformation rate at the current moment and lacks quantitative indicators of deformation spatial inhomogeneity. It achieves near real-time perception through real-time phase difference calculation and captures the precursors of local stress concentration through the relative change of phase gradient, providing a two-dimensional core indicator for risk classification.

[0079] The preset rate thresholds are set based on the allowable deformation rate of deep foundation pits, specification requirements, and engineering experience, and are exemplaryly divided into three levels: V_th1 = 0.002 m / day (2 mm / day, low-risk threshold), V_th2 = 0.005 m / day (5 mm / day, medium-risk threshold), and V_th3 = 0.01 m / day (10 mm / day, high-risk threshold). The preset change rate thresholds are set based on the historical gradient fluctuation level of the region, and are exemplaryly divided into three levels: R_th1 = 1.0 (gradient growth 100%, low-risk threshold), R_th2 = 1.5 (gradient growth 150%, medium-risk threshold), and R_th3 = 2.0 (gradient growth 200%, high-risk threshold). The arithmetic mean of the line-of-sight deformation rate of all point targets within the local area is calculated to obtain the comprehensive line-of-sight deformation rate of the local area, representing the overall deformation rate level of the local area, and is used for regional risk assessment. The risk level classification adopts a dual-index matrix discrimination method, combined with prior knowledge in the feature matrix for weighted adjustment: when the comprehensive line-of-sight deformation rate ≤ V_th1 and the phase gradient change rate ≤ R_th1, it is judged as low risk; when V_th1 < comprehensive line-of-sight deformation rate ≤ V_th2 or R_th1 < phase gradient change rate ≤ R_th2, it is judged as medium risk; when the comprehensive line-of-sight deformation rate > V_th2 and the phase gradient change rate > R_th2, or when any one of the following conditions is met, the comprehensive line-of-sight deformation rate ≥ V_th3 and the phase gradient change rate ≥ R_th3, it is judged as high risk; if the comprehensive line-of-sight deformation rate < 0 (phase reverse change, possibly deformation rebound) and the phase gradient change rate ≤ R_th1, it is judged as no risk; if the historical amplitude attenuation in a local area is significant or the deformation-attenuation correlation coefficient is high, the risk level can be weighted up by one level. This technology solves the problem that traditional early warning systems rely on a single threshold and cannot comprehensively reflect the combined risks of deformation rate and spatial variation. It achieves precise and reliable risk classification through a combination of multiple indicators.

[0080] Spatial location information of local areas is obtained through geographic coordinates in the feature data field, with the center coordinates of each local area serving as the spatial identifier. The risk warning indicator set is constructed in the form of a structured data list or map product, with each entry corresponding to a local area and including core indicators: center coordinates, specific values ​​of line-of-sight deformation rate, specific values ​​of phase gradient change rate, risk level (low risk / medium risk / high risk / no risk), risk judgment basis, and corresponding deep foundation pit functional zoning information (such as support structure area, pit bottom soil area).

[0081] This technology addresses the technical problems of scattered, unintuitive, and weak decision support in risk level classification, transforming complex multi-source data processing results into spatialized risk instructions that can be directly understood and operated on-site.

[0082] The above describes an InSAR remote sensing method for micro-deformation of deep foundation pits in an embodiment of this application. The following describes an InSAR remote sensing system for micro-deformation of deep foundation pits in an embodiment of this application. Please refer to [link to relevant documentation]. Figure 4 This application provides a schematic diagram of the structure of a deep foundation pit micro-deformation InSAR remote sensing system, which includes: The data acquisition module 10 is used to simultaneously collect continuous reflection signals and differential interferometric image sets of the deep foundation pit area through ground radar and InSAR satellite, and obtain a highly stable set of point targets and original scattering characteristic sequence.

[0083] The correction and fusion module 20 is used to perform registration residual correction and phase subtraction processing based on the point target set and the original scattering characteristic sequence to obtain the joint time series characteristic sequence of fused radar scattering and InSAR phase.

[0084] The denoising and separation module 30 is used to perform phase noise evaluation and weighted fusion on the joint time series feature sequence, separate the atmospheric delay phase and the true deformation phase, and obtain the net deformation phase sequence after removing the atmospheric influence.

[0085] The unwrapping integration module 40 is used to perform phase consistency quality control and phase unwrapping processing on the net deformation phase sequence to obtain a continuous and reliable cumulative deformation sequence.

[0086] The matrix construction module 50 is used to perform correlation analysis between the cumulative deformation sequence and the amplitude attenuation characteristics of the radar reflection signal, and to construct a deformation mechanism correlation feature matrix.

[0087] The deviation update module 60 is used to perform deviation calculation and secondary quality control on the deformation mechanism associated feature matrix to obtain the updated deformation mechanism associated feature matrix.

[0088] The early warning output module 70 is used to perform spatiotemporal fusion of the updated deformation mechanism associated feature matrix and the real-time differential interferometric image group, extract the deformation rate and phase gradient changes at the current moment, and generate a set of early warning indicators for local micro-deformation risks in deep foundation pits.

[0089] This application also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when the instructions are executed on a computer, cause the computer to perform the steps of the InSAR remote sensing method for micro-deformation of deep foundation pits.

[0090] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A Deep Foundation Pit Micro-Deformation InSAR Remote Sensing Method, characterized in that, The method includes: S1. By simultaneously acquiring continuous reflection signals and differential interferometric image sets of the deep foundation pit area through ground radar and InSAR satellite, a highly stable set of point targets and original scattering characteristic sequence are obtained. S2. Based on the set of point targets and the original scattering characteristic sequence, perform registration residual correction and phase subtraction processing to obtain a joint time-series feature sequence of fused radar scattering and InSAR phase. S3. Perform phase noise evaluation and weighted fusion on the joint temporal feature sequence to separate the atmospheric delay phase and the true deformation phase, and obtain the net deformation phase sequence after removing the atmospheric influence. S4. Perform phase consistency quality control and phase unwrapping processing on the net deformation phase sequence to obtain a continuous and reliable cumulative deformation sequence. S5. Perform correlation analysis between the cumulative deformation sequence and the amplitude attenuation characteristics of the radar reflection signal to construct a deformation mechanism correlation feature matrix; S6. Perform deviation calculation and secondary quality control on the deformation mechanism associated feature matrix to obtain the updated deformation mechanism associated feature matrix; S7. The updated deformation mechanism associated feature matrix is ​​spatiotemporally fused with the real-time differential interferometric image group to extract the deformation rate and phase gradient change at the current moment, and a set of early warning indicators for local micro-deformation of deep foundation pit is generated.

2. The method according to claim 1, characterized in that, S1 includes: For the retaining structure and soil surface of the deep foundation pit area, the coherence coefficient of each pixel in multiple time phases is calculated based on the differential interferometric image group. The coherence coefficient is compared with a preset coherence threshold, and pixels with a coherence coefficient higher than the preset coherence threshold are selected as candidate permanent scatterers. The candidate permanent scatterers are verified by time series analysis, and the surface permanent scatterers are identified based on the verification results to form the point target set. For each target point in the set of point targets, extract the radar reflection intensity value and radar phase value of the continuous time series from the continuous reflection signal; The radar reflection intensity value and radar phase value at each time point are aligned and integrated in chronological order to form the original scattering characteristic sequence.

3. The method according to claim 1, characterized in that, S2 include: The differential interferometric image group is spatially registered and the residual is corrected using the master-slave image registration method to obtain the registered differential interferometric image group. Based on the reference ellipsoid model and digital elevation model, the registered differential interferometric image group is processed for orbital error and terrain phase subtraction to obtain the corrected InSAR phase sequence. Based on the corrected InSAR phase sequence, the temporal phase stability index is calculated. At the same time, the original radar amplitude sequence corresponding to each point target is extracted from the original scattering characteristic sequence. Based on the aforementioned temporal phase stability index, the original radar amplitude sequence and the corrected InSAR phase sequence are fused to generate the joint temporal feature sequence.

4. The method according to claim 1, characterized in that, S3 include: For each point target in the joint temporal feature sequence, the temporal coherence index of each point target is calculated based on the InSAR phase sequence; The InSAR phase sequences from multiple time phases are fused using a coherence weighting method to obtain a noise-suppressed fused phase sequence. For the fused phase sequence, the atmospheric delayed phase component is separated by spatial low-pass filtering and time series analysis; The net deformation phase sequence is obtained by subtracting the atmospheric delay phase component from the InSAR phase in the joint time series feature sequence.

5. The method according to claim 1, characterized in that, S4 include: For the group of point targets within the enclosure structure area, based on the net deformation phase sequence, the standard deviation of the phase sequence corresponding to each point target is calculated as the phase consistency quality control index of the point targets. When the phase consistency quality control index of any target point is higher than the preset index threshold, jump point detection is performed on the phase sequence corresponding to any target point to identify possible mutations or abnormal data during the deformation process. The minimum cost flow method or statistical test method is used to perform differential interference phase unwrapping processing on the detected phase jump points to eliminate the influence of phase jumps and obtain the cumulative deformation sequence.

6. The method according to claim 1, characterized in that, S5 include: Based on the cumulative deformation sequence and satellite geometric parameters, deformation decomposition projection is performed to obtain the projection results of the deformation components; From the projection results, the horizontal displacement component sequence of the retaining structure and the vertical heave component sequence of the soil are extracted respectively. Extract the radar reflection intensity value sequence corresponding to each point target from the original scattering characteristic sequence, and calculate the amplitude attenuation ratio sequence based on the radar reflection intensity value sequence; The horizontal displacement component sequence, the vertical bulge component sequence, and the amplitude attenuation ratio sequence are time-series aligned and correlated to establish the correspondence between deformation components and amplitude attenuation characteristics. Based on the correspondence and the projection results, a deformation mechanism association feature matrix containing multiple deformation component association relationships is constructed.

7. The method according to claim 1, characterized in that, S6 include: The deformation feature values ​​of each target point in the deformation mechanism associated feature matrix are compared with the corresponding phase stability reference values ​​in the historical time-series phase stability sequence to calculate the feature deviation value of each target point. When the feature deviation value exceeds the preset deviation threshold, the corresponding point target is marked as an abnormal point target; Based on the differential interferometric image group, a polynomial fitting removal process for orbital error is performed on each abnormal point target to eliminate the influence of orbital error on phase data and obtain a corrected phase data sequence. Based on the corrected phase data sequence, the phase consistency index of each target point is obtained by calculating the phase change gradient between adjacent time nodes. Valid point targets with a phase consistency index less than a preset quality threshold are selected, and an updated deformation mechanism association feature matrix is ​​generated based on the valid point targets.

8. The method according to claim 1, characterized in that, S7 includes: The updated deformation mechanism associated feature matrix is ​​spatiotemporally aligned and fused with the latest acquired real-time differential interferometric image group to obtain the feature data field; Based on the feature data field, the line-of-sight deformation rate of each target point at the current moment is calculated, and in the feature data field, the local region formed by each target point and its spatial neighborhood is used as the analysis unit to calculate the phase gradient change rate within the corresponding local region. The line-of-sight deformation rate and the phase gradient change rate are compared with a preset rate threshold and a preset change rate threshold, respectively. Based on the comparison results, the risk levels of multiple local areas in the deep foundation pit are classified. Based on risk level classification and spatial location information of local areas, a set of early warning indicators for local micro-deformation risks in deep foundation pits is generated, which includes risk level information of multiple local areas.

9. A deep foundation pit micro-deformation InSAR remote sensing system, used to implement the method as described in any one of claims 1 to 8, characterized in that, The system includes: The data acquisition module is used to simultaneously collect continuous reflection signals and differential interferometric image sets of the deep foundation pit area through ground radar and InSAR satellite, and obtain a highly stable set of point targets and the original scattering characteristic sequence. The correction and fusion module is used to perform registration residual correction and phase subtraction processing based on the point target set and the original scattering characteristic sequence to obtain a joint time-series feature sequence of fused radar scattering and InSAR phase. The noise reduction and separation module is used to perform phase noise evaluation and weighted fusion on the joint temporal feature sequence, separate the atmospheric delay phase and the true deformation phase, and obtain the net deformation phase sequence after removing the atmospheric influence. The unwrapping integration module is used to perform phase consistency quality control and phase unwrapping processing on the net deformation phase sequence to obtain a continuous and reliable cumulative deformation sequence. The matrix construction module is used to perform correlation analysis between the cumulative deformation sequence and the amplitude attenuation characteristics of the radar reflection signal, and to construct a deformation mechanism correlation feature matrix. The deviation update module is used to perform deviation calculation and secondary quality control on the deformation mechanism associated feature matrix to obtain the updated deformation mechanism associated feature matrix. The early warning output module is used to perform spatiotemporal fusion of the updated deformation mechanism associated feature matrix with the real-time differential interferometric image group, extract the deformation rate and phase gradient changes at the current moment, and generate a set of early warning indicators for local micro-deformation risks in deep foundation pits.

10. A computer-readable storage medium storing instructions thereon, characterized in that, When the instructions are executed by the processor, they implement an InSAR remote sensing method for micro-deformation of deep foundation pits as described in any one of claims 1 to 8.

Images