Intelligent monitoring system and method for deformation of reclamation site based on InSAR (Interferometric Synthetic Aperture Radar)

Abstract

The invention relates to the technical field of synthetic aperture radar interferometry, in particular to an intelligent sea reclamation site deformation monitoring system and method based on InSAR, and the system comprises a data collection module which is used for collecting a multi-temporal satellite-borne SAR image of a sea reclamation area; the data processing module is used for executing sub-pixel matching, orbit error correction, atmospheric delay correction and residual optimization; the model construction module is used for constructing a segmented orbit error model, a tide-humidity coupling model and a dynamic regression model; the verification feedback module is used for verifying leveling precision and dynamically adjusting model parameters; according to the method, by fusing terrain mutability, tide-humidity coupling and soil rheological characteristics of a reclamation scene, accurate correction of multi-source error distribution of orbit errors, atmospheric delay and creep noise is realized, and the problem of insufficient adaptability of a traditional method in reclamation area deformation monitoring is solved; and millimeter-level reliable data support can be provided for cofferdam safety, hydraulic reclamation consolidation and port engineering.

Description

Technical Field

[0001] The present invention relates to the technical field of synthetic aperture radar interferometry, and in particular to an intelligent monitoring system and method for land reclamation site deformation based on InSAR. Background Art

[0002] With the recent growth of land reclamation, reclamation sites have become a significant form of land development in coastal areas, and subsidence monitoring at these sites is crucial for engineering safety and disaster early warning. Synthetic Aperture Radar Interferometry (InSAR) technology, which inverts surface deformation using phase differences in radar signals, has become a core tool for geological hazard monitoring. However, due to the high compressibility of soft soil foundations, tidal interactions, and spatiotemporal heterogeneity of humidity, reclamation sites exhibit multi-scale variations in surface deformation, posing significant challenges to traditional InSAR technology.

[0003] Current InSAR technology faces the challenge of coupled interference from multiple sources of errors in land reclamation deformation monitoring. Traditional orbit error correction often utilizes global polynomial models. However, in areas of abrupt terrain changes caused by the rapid accumulation of fill in land reclamation areas, such models struggle to capture spatial heterogeneity. Residual errors can reach 5-10 mm / km, compromising the accuracy of spatial trends in the deformation field. Furthermore, existing atmospheric delay correction methods fail to fully account for the coupling effect of high humidity and cyclical tidal fluctuations in land reclamation areas, resulting in residual phase standard deviations generally exceeding 3 mm, with errors exacerbated during the rainy season or during spring tides.

[0004] Existing multi-source data fusion technologies are often limited to a single error source or sensor type, lacking physical correlation modeling for reclamation scenarios. Currently, only corner reflector-based orbit corrections have been proposed, without integrating GNSS data to optimize atmospheric delay models. Alternatively, TAU models are constructed using GNSS alone to compensate for atmospheric disturbances, but without adapting to orbit errors and tidal coupling effects in reclamation areas. Existing verification methods rely on internal consistency checks within InSAR and lack external independent data closed-loop verification, resulting in insufficient absolute accuracy and difficulty meeting the millimeter-level monitoring requirements of reclamation projects.

[0005] Existing InSAR correction methods rely heavily on internal residual statistics and lack closed-loop validation using independent external data. Leveling points in reclamation areas often utilize a uniform grid layout that is not optimized for differential settlement characteristics (such as the interface between dredged fill and undisturbed soil), resulting in insufficiently representative verification results. Furthermore, traditional correction models are static, parameterized designs that are unable to adapt to dynamic load changes during reclamation construction. Deformation inversion exhibits significant lag, making it difficult to meet the real-time monitoring needs of the project.

[0006] Although multi-source data fusion (e.g., GNSS, meteorological stations, and tide gauges) is widely advocated, existing methods focus on correcting single error sources and fail to achieve coordinated optimization of orbital, atmospheric, and tidal loadings. Existing tropospheric elevation correction models, while incorporating multi-source data weighting, fail to integrate soil permeability and tidal modulation effects in reclaimed areas, limiting the applicability of elevation-related models in areas with soft soil foundations. Furthermore, heterogeneous data registration relies on manual control points in island regions, which is costly and risky, hindering the engineering application of high-resolution SAR imagery.

[0007] Existing corner reflector deployments are not optimized for soft soil foundations in reclamation areas. Mechanical deformation errors (such as pile foundation tilt) can reach 0.3-0.5mm / year, and there is a lack of a real-time monitoring and feedback platform. Furthermore, time-series InSAR processing relies on manual intervention, and lacks edge computing and dynamic parameter adjustment capabilities, making it unable to adapt to the rapid deformation monitoring needs during construction. While currently proposed ionospheric correction methods improve accuracy, they lack specialized filters designed for reclamation scenarios, resulting in limited noise suppression.

[0008] Therefore, to address the problems of multi-source error coupling, insufficient scene adaptability and low verification reliability in land reclamation site deformation monitoring, the present invention proposes an intelligent monitoring system and method for land reclamation site deformation based on InSAR. Summary of the Invention

[0009] In order to overcome the problems of multi-source error coupling and low verification reliability in InSAR deformation monitoring of reclamation sites, the present invention proposes an intelligent deformation monitoring system and method for reclamation sites based on InSAR.

[0010] The technical solution of the present invention is: an intelligent monitoring system for land reclamation site deformation based on InSAR, comprising:

[0011] Data acquisition module, including spaceborne SAR image acquisition unit, dual-frequency GNSS receiving unit, micro-meteorological station and level;

[0012] Data processing module for performing sub-pixel matching, orbit error correction, atmospheric delay correction, and residual optimization;

[0013] Model building module, used to build segmented orbit error models, tide-humidity coupling models, and dynamic regression models;

[0014] Verification and feedback module, including solar power supply module and edge computing module, used for leveling accuracy verification and dynamic adjustment of model parameters;

[0015] The automated processing platform integrates the functions of each module to achieve full-process automated processing.

[0016] Preferably, the satellite-borne SAR image acquisition unit is used to collect multi-phase satellite-borne SAR images of the reclamation area; the dual-frequency GNSS receiving unit is used to collect three-dimensional coordinate data of the reclamation area in real time and monitor the self-settlement and tilt of the base; the micro-meteorological station is used to synchronously collect temperature, humidity and air pressure parameters; the level is used to encrypt and deploy high-precision leveling points in a 50m×50m grid in the fill area.

[0017] Preferably, the system further comprises a GPS timing module, through which the meteorological data collected by the micro-meteorological monitoring unit is strictly synchronized with the SAR image acquisition time.

[0018] Preferably, an intelligent monitoring method for land reclamation site deformation based on InSAR includes the following steps:

[0019] S1, collect reclamation site data and perform preprocessing;

[0020] S2, using corner reflectors for sub-pixel matching, orbit error modeling and correction, and tidal cycle compensation;

[0021] S3, using GNSS to correct InSAR atmospheric delay errors;

[0022] S4, perform closed-loop verification of leveling accuracy;

[0023] S5, using the software platform to automate the entire process;

[0024] S6, using dynamic feedback mechanism to achieve model parameter adaptation and anomaly self-detection.

[0025] Preferably, the step S1 includes:

[0026] Collect multi-temporal satellite-borne SAR images of the reclamation area, covering the entire cycle from construction to operation;

[0027] Anti-subsidence pedestals are deployed every 200m along the axis of the reclamation area's cofferdam, with a depth of ≥3m. Dual-frequency corner reflectors are installed on top, integrating dual-frequency GNSS receiver modules to collect GNSS three-dimensional coordinate data in real time and monitor the subsidence and tilt of the pedestals themselves. The data is transmitted to the edge computing module in real time via 4G / 5G.

[0028] A micro-meteorological station is deployed next to the corner reflector to monitor temperature, humidity, and air pressure parameters. The sampling frequency is ≤ 5 minutes / time. The data is stored synchronously with the GNSS ZTD. The GPS timing module ensures that the meteorological data is strictly synchronized with the SAR image acquisition time.

[0029] High-precision leveling points should be laid out at least 5 per square kilometer at the intersection of the settlement center and the cofferdam in the reclamation area, and measured regularly using a leveling instrument. In the dredger fill area, the leveling points should be laid out in a denser grid of 50m×50m to cover the sensitive areas of soft soil differential settlement;

[0030] Real-time tide height data from tide gauges in the reclamation area is connected, and the compression index and effective stress of the soil in the reclamation area are measured through borehole sampling and consolidation tests for use in track error model correction.

[0031] Preprocessing includes radiometric correction, multi-look processing, sub-pixel registration and deflating phase to generate differential interferograms.

[0032] Preferably, step S2 includes:

[0033] Based on the normalized cross-correlation algorithm, the peak area of ​​the corner reflector is located in the SAR intensity image;

[0034]

[0035] Among them, I1, I2: master and slave image intensity values, Local window mean;

[0036] Perform quadratic surface fitting on the peak area to achieve sub-pixel coordinate extraction;

[0037] Constructing a second-order polynomial model of adaptive track error for land reclamation scenarios:

[0038]

[0039] Among them, h(x,y): DEM elevation data, compensating for the impact of sudden terrain changes;

[0040] C c : compression index obtained from geotechnical tests in the reclamation area, σ′ is the effective stress, which corrects the low-frequency deformation interference caused by soft soil consolidation;

[0041] A piecewise polynomial model is introduced to dynamically divide regions according to the DEM elevation gradient:

[0042]

[0043] The robust least squares estimation is introduced to suppress gross errors caused by pile foundation tilt or local settlement;

[0044] Add tidal harmonics to the model:

[0045]

[0046] Among them, A i ,ω i ,φ i: Tidal harmonic constant calibrated by tide gauge data in the reclamation area.

[0047] Preferably, step S3 includes:

[0048] By integrating low-orbit GNSS satellite data with ground-based GNSS station ZTD data, a compressed sensing algorithm was used to invert the three-dimensional vertical distribution of water vapor in the reclamation area, with a resolution of 500m×500m×100m.

[0049] Establish a multiple regression equation:

[0050]

[0051] Where, h: elevation;

[0052] d coast : distance from the coastline;

[0053] Humidity change rate;

[0054] h tide : Real-time tide height;

[0055] Ridge regression is used to solve the multicollinearity problem and dynamic weight allocation:

[0056]

[0057] Where, α: humidity sensitivity coefficient;

[0058] Average humidity of the area;

[0059] The global ocean tide model is used to calculate the tidal deformation components of the reclamation area. The tidal signal is deducted from the GNSS data, and the remaining signal is used for regression model training.

[0060] Preferably, the step S4 includes:

[0061] High gradient area: Benchmark density ≥ 10 / km 2 , using daily measurement frequency;

[0062] Low gradient area: Benchmark density ≥ 3 / km 2 , using weekly measurement frequency;

[0063] The spatial consistency index is used to evaluate the accuracy of the corrected deformation field and leveling data;

[0064] Perform wavelet packet decomposition and separation on the corrected residual phase:

[0065] Low-frequency residual orbit error: fed back to the orbit model to improve the polynomial order;

[0066] Medium frequency atmospheric noise: Optimize regression model weights;

[0067] High-frequency reclamation noise: marked as construction machinery vibration or soft soil creep or dynamic filtering;

[0068] Finally, it integrates a corner reflector, dual-frequency GNSS receiver, micro-weather station, and solar power supply module to support 4G / 5G real-time transmission;

[0069] The edge computing module has a built-in GPU acceleration chip to process SAR images and GNSS data in real time, reducing cloud transmission delays.

[0070] Preferably, step S5 includes:

[0071] Input SAR images, GNSS ZTD, and leveling data;

[0072] Perform sub-pixel matching, orbit correction, atmospheric regression, level verification and residual optimization in sequence;

[0073] Output correction deformation field, accuracy report and abnormal alarm.

[0074] Preferably, step S6 includes:

[0075] If the RMSE is greater than 5mm for three consecutive times, the orbital model order will be automatically increased; the update frequency of the regression model will be adjusted according to the humidity change rate (Δw / Δt), and the update frequency will be between 1 hour and 1 day.

[0076] Perform abnormal self-detection and define the residual abnormal threshold (|φ residual |>3σ), mark abnormal pixels and trigger data re-collection.

[0077] Beneficial effects of the present invention:

[0078] 1. This invention integrates GNSS, micro-meteorological station and low-orbit satellite data to construct a multi-factor regression equation, dynamically correlating tidal height, humidity change rate and atmospheric delay, thereby improving the accuracy of atmospheric delay correction and achieving high-reliability separation of tidal load deformation and true sedimentation signals.

[0079] 2. The present invention introduces geotechnical test parameters of the reclamation area (compression index Cc, effective stress σ′) into the track error model to compensate for the low-frequency deformation interference caused by soft soil consolidation, thereby effectively reducing the low-frequency spatial trend distortion of the track error.

[0080] 3. This invention integrates low-orbit GNSS constellation observation data to fill the monitoring gaps in sparse areas of traditional ground stations, and jointly solves GNSS data with tidal load deformation to avoid spectral confusion between atmospheric signals and tidal components, thereby improving the resolution of atmospheric vertical stratification, enhancing the correction capability of sparse areas, and optimizing data collaboration efficiency.

[0081] 4. The present invention identifies typical noise frequency bands in reclamation areas based on wavelet packet decomposition, thereby dynamically adjusting the filtering threshold and dynamically adjusting the leveling point density according to the settlement gradient, thereby giving priority to covering high-risk areas, thereby improving verification efficiency and pertinence.

[0082] 5. The present invention adopts high-strength materials and real-time deformation monitoring modules to extend the service life of the equipment and reduce the accumulation of mechanical errors. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] Figure 1 Shown is a flow chart of a multi-source error step-by-step correction method suitable for InSAR deformation monitoring of reclamation sites according to the present invention. DETAILED DESCRIPTION

[0084] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of the present invention.

[0085] The present invention provides an embodiment: an intelligent monitoring system for land reclamation site deformation based on InSAR, the system comprising a data acquisition module, a data processing module, a model building module, a verification feedback module and an automated processing platform, wherein the data acquisition module comprises a spaceborne SAR image acquisition unit, a dual-frequency GNSS receiving unit, a micro-meteorological station and a level, and is used to collect multi-phase spaceborne SAR images of the land reclamation area; the data processing module is used to perform sub-pixel matching, orbit error correction, atmospheric delay correction and residual optimization; the model building module is used to construct a segmented orbit error model, a tidal-humidity coupling model and a dynamic regression model; the verification feedback module comprises a solar power supply module and an edge computing module, and is used for leveling accuracy verification and dynamic adjustment of model parameters; the automated processing platform integrates the functions of each module to realize full-process automated processing.

[0086] Furthermore, an intelligent monitoring method for land reclamation site deformation based on InSAR includes six steps, which are specifically as follows:

[0087] The first step is to collect and preprocess the reclamation site data, as follows:

[0088] Collect multi-temporal satellite-borne SAR images (C / X band) of the reclamation area, covering the entire cycle from construction to operation;

[0089] Anti-subsidence pedestals are deployed every 200 meters along the axis of the reclamation area's cofferdam, with a depth of ≥3 meters. Dual-frequency corner reflectors compatible with the C / X bands are installed on top, integrating a dual-frequency GNSS receiver module to collect GNSS three-dimensional coordinate data in real time with a sampling frequency of ≥1Hz. The pedestal's own subsidence and tilt are monitored, and the data is transmitted to the edge computing module in real time via 4G / 5G.

[0090] A micro-meteorological station is deployed next to the corner reflector to monitor temperature, humidity, and air pressure parameters. The sampling frequency is ≤ 5 minutes / time. The data is stored synchronously with the GNSS ZTD. The GPS timing module ensures that the meteorological data is strictly synchronized with the SAR image acquisition time.

[0091] High-precision leveling points should be laid out at least 5 per square kilometer at the intersection of the settlement center and the cofferdam in the reclamation area, and measured regularly using a leveling instrument. In the dredger fill area, the leveling points should be laid out in a denser grid of 50m×50m to cover the sensitive areas of soft soil differential settlement;

[0092] Real-time tide height data from tide gauges in the reclamation area is connected, and the compression index and effective stress of the soil in the reclamation area are measured through borehole sampling and consolidation tests for use in track error model correction.

[0093] Preprocessing includes radiometric correction, multi-look processing, sub-pixel registration and deflating phase to generate differential interferograms.

[0094] The second step is to use corner reflectors for sub-pixel matching, orbit error modeling and correction, and tidal period compensation, as follows:

[0095] Based on the normalized cross-correlation algorithm, the peak area of ​​the corner reflector is located in the SAR intensity image;

[0096]

[0097] Among them, I1, I2: master and slave image intensity values, Local window mean;

[0098] Perform quadratic surface fitting on the peak area to achieve sub-pixel coordinate extraction;

[0099] Constructing a second-order polynomial model of adaptive track error for land reclamation scenarios:

[0100]

[0101] Among them, h(x,y): DEM elevation data, compensating for the impact of sudden terrain changes;

[0102] C c : compression index obtained from geotechnical tests in the reclamation area, σ′ is the effective stress, which corrects the low-frequency deformation interference caused by soft soil consolidation;

[0103] A piecewise polynomial model is introduced to dynamically divide regions according to the DEM elevation gradient:

[0104]

[0105] The robust least squares estimation is introduced to suppress gross errors caused by pile foundation tilt or local settlement;

[0106] Add tidal harmonics to the model:

[0107] Δφ tide =∑ i＝1 A i sin(ω i t+φ i );

[0108] Among them, A i ,ω i ,φ i : Tidal harmonic constant calibrated by tide gauge data in the reclamation area.

[0109] The third step is to use GNSS to correct the InSAR atmospheric delay error, as follows:

[0110] By integrating low-orbit GNSS satellite data with ground-based GNSS station ZTD data, a compressed sensing algorithm was used to invert the three-dimensional vertical distribution of water vapor in the reclamation area, with a resolution of 500m×500m×100m.

[0111] Establish a multiple regression equation:

[0112]

[0113] Where, h: elevation;

[0114] d coast : distance from the coastline;

[0115] Humidity change rate;

[0116] h tide : Real-time tide height;

[0117] Ridge Regression is used to solve the multicollinearity problem and dynamic weight allocation:

[0118]

[0119] Where, α: humidity sensitivity coefficient;

[0120] Average humidity of the area;

[0121] The global ocean tide model is used to calculate the tidal deformation components of the reclamation area. The tidal signal is deducted from the GNSS data, and the remaining signal is used for regression model training.

[0122] The fourth step is to conduct closed-loop verification of leveling accuracy, as follows:

[0123] High gradient area (sedimentation rate>5mm / 30m): Benchmark density ≥10 / km 2 , using daily measurement frequency;

[0124] Low gradient area (sedimentation rate <2mm / 30m): Benchmark density ≥3 / km 2 , using a weekly measurement frequency;

[0125] The spatial consistency index is used to evaluate the accuracy of the corrected deformation field and leveling data;

[0126] Perform wavelet packet decomposition and separation on the corrected residual phase:

[0127] Low-frequency residual orbit error (0-0.001Hz): fed back to the orbit model to improve the polynomial order;

[0128] Medium frequency atmospheric noise (0.001-0.01Hz): Optimize the regression model weights;

[0129] High-frequency reclamation noise (>1Hz): marked as construction machinery vibration or soft soil creep or dynamic filtering;

[0130] Finally, it integrates a corner reflector, dual-frequency GNSS receiver, micro-weather station, and solar power supply module to support 4G / 5G real-time transmission;

[0131] The edge computing module has a built-in GPU acceleration chip to process SAR images and GNSS data in real time, reducing cloud transmission delays.

[0132] The fifth step is to use GNSS to correct the InSAR atmospheric delay error, as follows:

[0133] Input SAR images, GNSS ZTD, and leveling data;

[0134] Perform sub-pixel matching, orbit correction, atmospheric regression, level verification and residual optimization in sequence;

[0135] Output correction deformation field, accuracy report and abnormal alarm.

[0136] The sixth step is to use the dynamic feedback mechanism to achieve model parameter adaptation and anomaly self-detection, as follows:

[0137] If the RMSE is greater than 5mm for three consecutive times, the orbital model order will be automatically increased; the update frequency of the regression model will be adjusted according to the humidity change rate (Δw / Δt), and the update frequency will be between 1 hour and 1 day.

[0138] Perform abnormal self-detection and define the residual abnormal threshold (|φ residual |>3σ), mark abnormal pixels and trigger data re-collection.

[0139] See also Figure 1 Furthermore, the present invention provides an embodiment of a dam deformation monitoring and prediction method based on InSAR and deep learning, which is as follows:

[0140] The first step is to collect multi-temporal images of the reclamation area from the construction period to the operation period and perform radiometric correction, multi-look processing, registration, and ground-leveling phase removal to generate a differential interferogram. This step specifically includes the following sub-steps:

[0141] The range-Doppler geometry model is used to eliminate the radiation distortion of SAR images and ensure the backscatter coefficient (σ 0 )’s physical consistency.

[0142] Perform multi-look processing (e.g., 4 views in azimuth and 1 view in range) on single-look complex (SLC) images to reduce speckle noise and improve the image signal-to-noise ratio (SNR).

[0143] Based on the improved coherent registration algorithm, sub-pixel registration of slave images is achieved by maximizing the interference coherence with the master image as a reference, and the registration accuracy is ≤0.1 pixel.

[0144] Using digital elevation model (DEM) data, the flat ground phase caused by terrain is removed using the formula:

[0145]

[0146] B ⊥ : vertical baseline, h: DEM elevation, R: slant range, θ: radar incident angle.

[0147] By subtracting the two temporal phases, the terrain phase is removed and the deformation and error phases are retained:

[0148] φ diff =φ interf -φ flat ;

[0149] The second step is to design the anti-subsidence base and corner reflector layout, as follows:

[0150] The anti-subsidence base consists of a prestressed concrete pile foundation (diameter ≥ 0.5m, depth ≥ 3m) and a titanium alloy bracket. Four vertical corner reflectors (supporting C / X bands) are installed on the top of the bracket, and the inclination angle of the reflector matches the radar incident angle.

[0151] A pedestal is arranged every 200m along the axis of the cofferdam in the reclamation area to cover the area with changing settlement trends. In the fill area, the pedestal is arranged in a denser grid of 50m×50m to meet the needs of monitoring differential settlement of soft soil consolidation. A dual-frequency GNSS receiver (compatible with GPS / Beidou) is integrated on the top of the pedestal to collect three-dimensional coordinate data in real time (frequency ≥1Hz) to monitor the settlement and tilt of the pedestal itself.

[0152] The third step is to integrate the micro-meteorological station with low-orbit GNSS data and deploy high-precision leveling points, as follows:

[0153] A micro-meteorological station is set up next to the corner reflector to monitor temperature, humidity, and air pressure. The sampling frequency is ≤5 minutes / time. The data is stored synchronously with the GNSS ZTD. Then, the original observation data (L1 / L2 frequency band) of low-orbit GNSS constellations such as Starlink is connected and jointly processed with ground GNSS data to improve the vertical resolution of atmospheric tomography. Finally, high-precision leveling points are set up at the settlement center of the reclamation area and the junction of the cofferdam, with a density of ≥5 / km 2 , regular measurements are carried out using a level, with a measurement accuracy of ±0.3mm / km.

[0154] The fourth step is corner reflector sub-pixel matching and track error correction, as follows:

[0155] First, locate the NCC peak:

[0156] Based on the normalized cross correlation (NCC) algorithm, the corner reflector peak area is searched in the SAR intensity image:

[0157]

[0158] I1, I2: master and slave image intensity values, Local window mean.

[0159] Quadratic surface fitting:

[0160] In the peak region (e.g., a 5×5 pixel window), the quadratic surface model is used to fit the intensity distribution:

[0161] f(x,y)=ax 2 +by 2 +cxy+dx+ey+f;

[0162] The coefficients were solved by the least squares method, and the sub-pixel peak coordinates were calculated (with an accuracy of 0.1 pixel).

[0163] Second-order polynomial model of orbit error:

[0164]

[0165] h(x,y): DEM elevation data, compensating for the impact of sudden terrain changes; C c : compression index obtained from geotechnical tests in the reclamation area, σ′: effective stress (determined by borehole sampling and consolidation tests), σ′0: initial effective stress.

[0166] Robust least squares estimation:

[0167] The Huber loss function is used to suppress the influence of gross errors, and the objective function is:

[0168]

[0169] ri: residual, k: adjustment parameter (usually 1.345σ).

[0170] Tidal harmonic modeling:

[0171]

[0172] A i ,ω i ,φ i : The tidal harmonic constants of M2, S2, K1 and O1 calibrated by the data of tide gauge station in the reclamation area.

[0173] The orbit error and tidal parameters are jointly estimated using the least squares method to ensure that only deformation and atmospheric signals are retained in the phase residual.

[0174] The fifth step is to use GNSS to correct the atmospheric delay error, as follows:

[0175] The 3D tomographic observation equation is constructed by fusing ground-based GNSS ZTD data with the slant path wet delay (SWD) of the GNSS signal from a low-orbit satellite:

[0176] SWD=∫ path S·PWV·ds;

[0177] S: projection function, PWV: atmospheric water vapor content.

[0178] The basis pursuit denoising (BPDN) algorithm is used to solve the sparse solution, and the objective function is:

[0179] min\|x\|1s.t.\|Ax-y\|2≤∈;

[0180] A: observation matrix, x: water vapor density vector, y: delayed observation value.

[0181] The sixth step is the humidity-tide dynamic regression model, which is as follows:

[0182] Multiple regression equation:

[0183]

[0184] h: elevation, d coast : Distance from the coastline (calculated by GIS), Humidity change rate (differential calculation), h tide :Real-time tide height (tide gauge station data).

[0185] Ridge regression solution:

[0186] The L2 regularization term is introduced to prevent overfitting, and the objective function is:

[0187] min||y-Xβ|| 2 +λ||β|| 2 ;

[0188] λ: Regularization coefficient (determined by cross-validation).

[0189]

[0190] The tidal component is deducted from the GNSS vertical displacement data, and the remaining signal is used to train the atmospheric delay regression model.

[0191] The seventh step is leveling accuracy closed-loop verification and residual optimization, as follows:

[0192] Dynamic layout of leveling points: High gradient area (sedimentation rate>5mm / 30m): leveling point density ≥10 / km 2 , using daily measurement frequency. Low gradient area (sedimentation rate <2mm / 30m): leveling point density ≥3 / km 2 , with a weekly measurement frequency.

[0193] Accuracy evaluation index: The spatial consistency index (SCI) is used to evaluate the accuracy of the corrected deformation field and leveling data.

[0194]

[0195] Residual phase wavelet packet decomposition: Perform wavelet packet decomposition on the corrected residual phase to separate:

[0196] Low-frequency residual track error (0-0.001Hz): Feedback to the track model to increase the polynomial order (e.g., second order → third order); Medium-frequency atmospheric noise (0.001-0.01Hz): Optimize the regression model weights; High-frequency land reclamation noise (>1Hz): Marked as construction machinery vibration or soft soil creep, dynamically filtered out; Dynamically adjust the threshold based on the noise standard deviation σ:

[0197]

[0198] Step 8: Full-process automated processing and dynamic feedback, as follows:

[0199] It integrates a corner reflector, a dual-frequency GNSS receiver, a micro-meteorological sensor, and a solar power supply module, supports 4G / 5G real-time transmission, and has a data sampling interval of ≤1 minute.

[0200] Built-in NVIDIA Jetson AGX Xavier GPU, it runs sub-pixel matching, orbit correction, and atmospheric regression algorithms in real time with a latency of ≤10 seconds.

[0201] Finally, input SAR image, GNSS ZTD, and leveling data, and execute the following steps in sequence:

[0202] Sub-pixel matching → orbit correction → atmospheric regression → level verification → residual optimization.

[0203] Outputs the corrected deformation field, accuracy report (PDF / Excel) and abnormality alarm (JSON format).

[0204] Define the residual abnormality threshold |φresidual|>3σ to automatically mark abnormal pixels and trigger data re-collection.

[0205] The present invention provides a comparative example, comparing the performance of the traditional method and the present solution in a reclamation area. For specific data, please refer to Table 1:

[0206] Table 1 Performance comparison between traditional method and this scheme in reclamation area

[0207]

[0208] The global polynomial model of the traditional scheme does not take into account the low-frequency deformation caused by the consolidation of soft soil in the reclamation area, such as the change of effective stress, resulting in a residual error of up to 8.7mm in the rapid loading area of ​​the dredger fill. The item compensates for the rheological properties of the soil and adapts to the sudden change of elevation gradient through segmented modeling, with the residual reduced to 1.8 mm.

[0209] The single GNSS-ZTD model of the traditional solution does not integrate the tidal cycle and humidity change rate, which causes the error to soar during the rainy season. For example, the RMSE measured during the typhoon period is 4.5mm. The tidal harmonic term of the present invention Separation of tidal load deformation and humidity change rate The weights are adjusted dynamically to keep the RMSE stable within 1.5mm under extreme weather conditions.

[0210] The traditional solution has 3-5 leveling points evenly distributed per square kilometer, which makes it impossible to capture the differential settlement of dredger fill soil, with a missed detection rate of >30%. The high gradient area of ​​the present invention is encrypted to 10 / km 2 , and measured daily, and verified by the Spatial Consistency Index (SCI), the correlation increased from 0.65 to 0.93.

[0211] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the present invention may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. An intelligent monitoring system for land reclamation site deformation based on InSAR, characterized in that: Includes: Data acquisition module, including spaceborne SAR image acquisition unit, dual-frequency GNSS receiving unit, micro-meteorological station and level; Data processing module for performing sub-pixel matching, orbit error correction, atmospheric delay correction, and residual optimization; Model building module, used to build segmented orbit error models, tide-humidity coupling models, and dynamic regression models; Verification and feedback module, including solar power supply module and edge computing module, used for leveling accuracy verification and dynamic adjustment of model parameters; The automated processing platform integrates the functions of each module to achieve full-process automated processing.

2. The InSAR-based intelligent monitoring system for land reclamation site deformation according to claim 1, characterized in that: The satellite-borne SAR image acquisition unit is used to collect multi-phase satellite-borne SAR images of the reclamation area; the dual-frequency GNSS receiving unit is used to collect three-dimensional coordinate data of the reclamation area in real time and monitor the self-settlement and inclination of the base; the micro-meteorological station is used to synchronously collect temperature, humidity and air pressure parameters; the level is used to encrypt and deploy high-precision leveling points in a 50m×50m grid in the fill area.

3. The InSAR-based intelligent land reclamation site deformation monitoring system according to claim 2, characterized in that: The system also includes a GPS timing module, which strictly synchronizes the meteorological data collected by the micro-meteorological monitoring unit with the SAR image collection time.

4. An intelligent monitoring method for land reclamation site deformation based on InSAR, according to the intelligent monitoring system for land reclamation site deformation based on InSAR according to any one of claims 1 to 3, characterized in that: The following steps are included: S1, collect reclamation site data and perform preprocessing; S2, using corner reflectors for sub-pixel matching, orbit error modeling and correction, and tidal cycle compensation; S3, using GNSS to correct InSAR atmospheric delay errors; S4, perform closed-loop verification of leveling accuracy; S5, using the software platform to automate the entire process; S6, using dynamic feedback mechanism to achieve model parameter adaptation and anomaly self-detection.

5. The method for intelligent monitoring of land reclamation site deformation based on InSAR according to claim 4, characterized in that: The step S1 comprises: Collect multi-temporal satellite-borne SAR images of the reclamation area, covering the entire cycle from construction to operation; Anti-subsidence pedestals are deployed every 200m along the axis of the reclamation area's cofferdam, with a depth of ≥3m. Dual-frequency corner reflectors are installed on top, integrating dual-frequency GNSS receiver modules to collect GNSS three-dimensional coordinate data in real time and monitor the subsidence and tilt of the pedestals themselves. The data is transmitted to the edge computing module in real time via 4G / 5G. A micro-meteorological station is deployed next to the corner reflector to monitor temperature, humidity, and air pressure parameters. The sampling frequency is ≤ 5 minutes / time. The data is stored synchronously with the GNSS ZTD. The GPS timing module ensures that the meteorological data is strictly synchronized with the SAR image acquisition time. High-precision leveling points should be laid out at least 5 per square kilometer at the intersection of the settlement center and the cofferdam in the reclamation area, and measured regularly using a leveling instrument. In the dredger fill area, the leveling points should be laid out in a denser grid of 50m×50m to cover the sensitive areas of soft soil differential settlement; Real-time tide height data from tide gauges in the reclamation area is connected, and the compression index and effective stress of the soil in the reclamation area are measured through borehole sampling and consolidation tests for use in track error model correction. Preprocessing includes radiometric correction, multi-look processing, sub-pixel registration and deflating phase to generate differential interferograms.

6. The method for intelligent monitoring of land reclamation site deformation based on InSAR according to claim 4, characterized in that: The step S2 comprises: Based on the normalized cross-correlation algorithm, the peak area of ​​the corner reflector is located in the SAR intensity image; Among them, I1, I2: master and slave image intensity values, Local window mean; Perform quadratic surface fitting on the peak area to achieve sub-pixel coordinate extraction; Constructing a second-order polynomial model of adaptive track error for land reclamation scenarios: Among them, h(x,y): DEM elevation data, compensating for the impact of sudden terrain changes; C c : Compression index obtained from geotechnical tests in the reclamation area, σ ′ For effective stress, correct the low-frequency deformation interference caused by soft soil consolidation; A piecewise polynomial model is introduced to dynamically divide regions according to the DEM elevation gradient: The robust least squares estimation is introduced to suppress gross errors caused by pile foundation tilt or local settlement; Add tidal harmonics to the model: Among them, A i ,ω i ,φ i : Tidal harmonic constant calibrated by tide gauge data in the reclamation area.

7. The method for intelligent monitoring of land reclamation site deformation based on InSAR according to claim 4, characterized in that: The step S3 comprises: By integrating low-orbit GNSS satellite data with ground-based GNSS station ZTD data, a compressed sensing algorithm was used to invert the three-dimensional vertical distribution of water vapor in the reclamation area, with a resolution of 500m×500m×100m. Establish a multiple regression equation: Where, h: elevation; d coast : distance from the coastline; Humidity change rate; h tide : Real-time tide height; Ridge regression is used to solve the multicollinearity problem and dynamic weight allocation: Where, α: humidity sensitivity coefficient; Average humidity of the area; The global ocean tide model is used to calculate the tidal deformation components of the reclamation area. The tidal signal is deducted from the GNSS data, and the remaining signal is used for regression model training.

8. The method for intelligent monitoring of land reclamation site deformation based on InSAR according to claim 4, characterized in that: The step S4 comprises: High gradient area: Benchmark density ≥ 10 / km 2 , using daily measurement frequency; Low gradient area: Benchmark density ≥ 3 / km 2 , using a weekly measurement frequency; The spatial consistency index is used to evaluate the accuracy of the corrected deformation field and leveling data; Perform wavelet packet decomposition and separation on the corrected residual phase: Low-frequency residual orbit error: fed back to the orbit model to improve the polynomial order; Medium frequency atmospheric noise: Optimize regression model weights; High-frequency reclamation noise: marked as construction machinery vibration or soft soil creep or dynamic filtering; Finally, it integrates a corner reflector, dual-frequency GNSS receiver, micro-weather station, and solar power supply module to support 4G / 5G real-time transmission; The edge computing module has a built-in GPU acceleration chip to process SAR images and GNSS data in real time, reducing cloud transmission delays; Among them, the track error model of compression index Cc and effective stress 'σ' first introduced soil mechanics parameters (C c ,σ'), compensate for the soft soil consolidation interference, and model the elevation gradient segmentally; the multiple regression equation of tidal harmonic term and humidity change rate integrates the tidal harmonic term (h tide ) and humidity change rate Dynamically correlated tidal-humidity coupling effects.

9. The method for intelligent monitoring of land reclamation site deformation based on InSAR according to claim 4, characterized in that: The step S5 comprises: Input SAR images, GNSS ZTD, and leveling data; Perform sub-pixel matching, orbit correction, atmospheric regression, level verification and residual optimization in sequence; Output correction deformation field, accuracy report and abnormal alarm.

10. The method for intelligent monitoring of land reclamation site deformation based on InSAR according to claim 4, characterized in that: The step S6 comprises: If the RMSE is greater than 5mm for three consecutive times, the orbital model order will be automatically increased; the update frequency of the regression model will be adjusted according to the humidity change rate (Δw / Δt), and the update frequency will be between 1 hour and 1 day. Perform abnormal self-detection and define the residual abnormal threshold (|φ residual |>3σ), mark abnormal pixels and trigger data re-collection.

Images