Unmanned aerial vehicle repeat-pass insar residual baseline error estimation method and system
Patent Information
- Application Number
- CN202611251707.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-18
- Publication Date
- 2026-11-03
AI Technical Summary
Existing UAV SAR systems are severely affected by airflow disturbances at high frequencies, resulting in significant residual baseline errors in imaging processing. Existing methods lack robustness, are difficult to directly extend to multi-baseline data, and are computationally intensive and time-consuming, making them unsuitable for rapid emergency response.
A virtual time-varying baseline equivalent model is adopted, and an overdetermined set of equations is constructed based on imaging geometry. The equations are solved directly through robust estimation, achieving single/multi-baseline residual baseline error phase compensation without sub-aperture division and original echo refocusing. A reference DEM is generated using dual-antenna cross-track InSAR data, terrain phase is removed, and a robust loss function is constructed for baseline estimation.
It has strong anti-phase entanglement and decoherence capabilities, significantly improved robustness, high computational efficiency, adaptability to single/multi-baseline data, supports real-time/near real-time processing, is suitable for rapid emergency response by UAVs, and has stable compensation accuracy.
Smart Images
Figure CN123029532A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of remote sensing science and technology, specifically relating to a method and system for estimating residual baseline errors in InSAR during repeated UAV flights. Background Technology
[0002] Synthetic Aperture Radar Interferometry (InSAR) technology, with its unique advantages of all-weather, all-time, and non-contact observation, has become an indispensable technical means in fields such as surface deformation monitoring and topographic mapping. Currently, InSAR data mainly comes from spaceborne and airborne SAR systems. Spaceborne SAR systems have the capability of wide-area coverage and long-term sequential observation, but are limited by factors such as long revisit periods, insufficient sensitivity of fixed observation geometry to displacement in specific directions, and limited spatial resolution, making it difficult to meet the needs of high-frequency, high-resolution monitoring in local areas. Airborne SAR systems, with their advantages of mobility, short revisit periods, and the ability to acquire sub-meter or even higher resolution data, have become an effective supplement to spaceborne systems. In recent years, small unmanned aerial vehicle (UAV) SAR systems have developed rapidly. Their convenient deployment, low cost, and flexible flight trajectories enable high-frequency repeated observations at the minute level, providing a new technical path for high-precision deformation monitoring in local areas. However, small UAV platforms are significantly affected by airflow disturbances, and their flight trajectories deviate greatly from the ideal routes. The navigation and positioning systems they carry have limited accuracy, resulting in significant time-varying baseline errors remaining after imaging processing and motion compensation. These errors manifest as phase distortion in the interferograms, becoming the most significant source of error restricting the accuracy of SAR interferometry for small UAVs.
[0003] To address these challenges, scholars both domestically and internationally have developed various residual baseline error estimation methods. The multisquint (MS) method, which estimates the time-varying baseline change rate by analyzing the interferometric phase differences of sub-apertures at different azimuths, has been successfully applied in large airborne SAR systems in the mid-to-long wavelength band. However, the MS method relies on two key assumptions: the residual error changes slowly within the sub-aperture and can be approximated as a low-order model, and the system wavelength is long enough to control the impact of platform error on the phase. These two assumptions do not hold true in high-frequency SAR scenarios involving small UAVs—high-frequency wavelengths are short, and centimeter-level platform errors can induce multi-cycle phase entanglement; UAVs are subject to severe airflow disturbances, resulting in significant attitude changes within a single aperture, making it difficult to guarantee sub-aperture coherence. Furthermore, phase tracking methods rely on stable, strong scattering points in the scene, leading to poor universality in areas with natural surface coverage; self-focusing methods are sensitive to the assumption of slowly varying errors, and the estimation error increases significantly in nonlinear motion scenarios. Existing methods either rely on specific types of ground features, aperture coherence, or require human intervention, and generally suffer from insufficient robustness and incomplete error compensation in high-frequency, low-altitude, and strongly disturbed small UAV SAR data.
[0004] Furthermore, most existing methods are designed around single-baseline interferometric data processing workflows, making them difficult to directly extend to multi-baseline data stacks. Multi-baseline InSAR, through multi-track and multi-baseline observations, can improve the reliability of elevation inversion and phase unwrapping, representing an important development direction for refined UAV monitoring. However, traditional methods can only process each baseline independently and cannot utilize the redundancy information of multiple baselines to improve accuracy and robustness. Some methods require switching processing workflows between single-baseline and multi-baseline scenarios, which presents significant engineering challenges.
[0005] Meanwhile, existing methods mostly use physical baseline or actual trajectory errors as the estimation object, and the estimation results need to be substituted back into the imaging process to achieve error correction through refocusing. This mode relies on raw echo data, which is computationally intensive and time-consuming, making it unsuitable for rapid emergency response. Moreover, refocusing is still difficult to effectively suppress phase distortion when there are gross errors in the navigation data. There is currently a lack of a method for directly extracting residual baseline errors that does not require sub-aperture division, is compatible with single / multi-baseline data, and does not require raw echo refocusing. Summary of the Invention
[0006] To address the aforementioned technical problems, this invention provides a method and system for estimating residual baseline errors in InSAR during repeated UAV flights. Based on imaging geometry, a virtual time-varying baseline equivalent model is constructed. An overdetermined set of equations is built using all range pixels and solved directly using robust estimation. This achieves one-time compensation of single / multi-baseline residual baseline error phase without the need for sub-aperture division and original echo refocusing.
[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows:
[0008] A method for estimating residual baseline error in InSAR during repeated UAV flights, the method comprising:
[0009] Step 1: Acquire single-baseline or multi-baseline SAR single-look complex data. Using the selected main image as the reference, perform registration and resampling on all auxiliary images to form a registered multi-baseline data stack. Perform phase connection processing on the multi-baseline data stack to obtain a differential interferometric phase map.
[0010] Step 2: Using the dual-antenna cross-track InSAR data acquired synchronously with the SAR single-look complex data in the same sortie, a reference DEM is generated by inversion. Combined with the orbit parameters of the repeated track interferometry, the terrain phase is simulated group by group and subtracted from the differential interferometric phase map to obtain the differential interferometric phase map after removing the terrain phase.
[0011] Step 3: For each azimuth time, based on the range-Doppler imaging principle, the positions of the two passing platforms at the same zero Doppler time are used to form an equivalent observation geometry, and the horizontal and vertical components of the virtual time-varying baseline are defined as parameters to be estimated; based on all range pixels in the differential interferometric phase map after terrain phase removal, a robust loss function is used to construct the objective function and solve it to obtain the residual time-varying baseline estimates for each interferometric pair in the corresponding azimuth direction.
[0012] Step 4: Calculate the residual baseline interferometric phase based on the residual time-varying baseline estimate, and subtract it from the differential interferometric phase diagram after removing the topographic phase to complete the residual baseline error compensation.
[0013] Furthermore, the registration in step 1 includes: coarse registration by calculating the overall pixel offset of the main and auxiliary images based on the two-dimensional phase correlation method; then solving the local offset grid by grid by grid using spatial sliding window cross-correlation; removing abnormal offsets based on statistical criteria; obtaining the theoretical pixel-by-pixel offset of the whole image by polynomial fitting of the effective offset; and achieving fine registration by bilinear interpolation.
[0014] Furthermore, in step 1, when performing phase connection processing on the multi-baseline data stack, a local window is selected centered on each pixel, the complex covariance matrix is estimated using all samples within the window, the complex covariance matrix is decomposed into eigenvalues, and the eigenvector corresponding to the largest eigenvalue is taken as the optimal phase representation of the pixel in the multi-temporal image. An equivalent single-reference interferogram is constructed with the selected master image as a reference.
[0015] Furthermore, in step 2, the dual-antenna cross-track InSAR data and the SAR data are acquired synchronously in the same flight. The surface elevation of the observation area is calculated by cross-track interferometric geometry, and the elevation of water bodies and shadowed area pixels is repaired to generate a reference digital elevation model.
[0016] Furthermore, in step 2, when simulating the terrain phase group by group, the terrain phase is calculated pixel by pixel based on the vertical baseline, slant distance and incident angle corresponding to each repeated track interferometer pair, combined with the reference digital elevation model, and subtracted from the original interferometric phase to obtain the differential interferometric phase after removing the terrain influence.
[0017] Furthermore, in step 3, when constructing the objective function using the robust loss function, the robust loss function is the Huber loss function. The transition between the quadratic loss and the linear loss is controlled by the scale parameter, suppressing the interference of low coherence regions, registration errors, and abnormal observations on the baseline estimation. The solution of the horizontal and vertical components of the virtual time-varying baseline is transformed into a robust nonlinear least squares problem with boundary constraints. The boundary constraints are set according to the platform positioning error range provided by the navigation system.
[0018] Furthermore, in step 4, when calculating the residual baseline interferometric phase based on the estimated residual time-varying baseline, the slant range deviation is first calculated from the estimated virtual baseline parameters and the observation geometry, and then the residual baseline interferometric phase is calculated from the slant range deviation.
[0019] On the other hand, the present invention provides a system for estimating residual baseline errors in InSAR during repeated flights of unmanned aerial vehicles, comprising:
[0020] The differential module is used to acquire single-baseline or multi-baseline SAR single-look complex data. Using the selected main image as a reference, it performs registration and resampling on all auxiliary images to form a registered multi-baseline data stack. Phase connection processing is performed on the multi-baseline data stack to obtain a differential interferometric phase map.
[0021] The terrain removal module is used to invert and generate a reference DEM using dual-antenna cross-track InSAR data acquired synchronously with the SAR single-look complex data. Combined with the orbit parameters of the repeated track interferometry, the terrain phase is simulated group by group and subtracted from the differential interferometric phase map to obtain the terrain-removed differential interferometric phase map.
[0022] The estimation module is used to construct an equivalent observation geometry for the positions of the two passing platforms at the same zero Doppler time for each azimuth time based on the range-Doppler imaging principle, and define the horizontal and vertical components of the virtual time-varying baseline as parameters to be estimated; based on all range pixels in the differential interferometric phase map after terrain phase removal, a robust loss function is used to construct an objective function and solve it to obtain the residual time-varying baseline estimates for each interferometric pair in the corresponding azimuth direction;
[0023] The compensation module is used to calculate the residual baseline interferometric phase based on the residual time-varying baseline estimate and subtract it from the differential interferometric phase diagram after removing the topographic phase, thereby completing the residual baseline error compensation.
[0024] Thirdly, the present invention provides an electronic device, comprising: one or more processors; and a memory for storing one or more programs; wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the aforementioned method for estimating residual baseline errors of repeated UAV flights in InSAR.
[0025] Fourthly, the present invention provides a computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, enable the processor to implement the aforementioned method for estimating residual baseline errors of repeated UAV flights in InSAR.
[0026] The beneficial effects of this invention are as follows:
[0027] Strong resistance to phase entanglement and decoherence, and significantly improved robustness: This invention does not require sub-aperture division, does not rely on aperture coherence and error gradual variation assumptions, and directly constructs an estimation model based on the full aperture interferometric phase. It fundamentally avoids the method failure problem caused by strong phase entanglement and sub-aperture decoherence in high-frequency SAR, and has significantly better robustness than traditional methods for residual baseline error estimation under strong disturbance conditions.
[0028] Adaptive Compatibility with Single / Multi-Baseline Data: The unified estimation model constructed in this invention can adaptively handle single-baseline and multi-baseline data stacks without switching processing flows. In multi-baseline mode, redundant observations can be fully utilized to suppress the influence of noise and outliers, improving estimation accuracy and robustness; in single-baseline mode, it still maintains its advantages of speed and efficiency, and has wide engineering adaptability.
[0029] High computational efficiency without the need for refocusing the original echo: This invention uses virtual time-varying baseline modeling to directly complete residual phase estimation and subtraction at the level of SLC data and interferograms. There is no need to substitute the estimation results back to the imaging stage to trigger refocusing. It does not rely on the original echo data throughout the process, which greatly reduces the amount of computation and storage requirements. It can support real-time / near real-time processing and is suitable for rapid emergency response scenarios of UAVs.
[0030] Robust estimation ensures stability in complex scenarios: This invention uses the Huber loss function to construct a robust nonlinear least squares objective function, which effectively suppresses the interference of low coherence regions, registration errors and abnormal observations on baseline estimation, and can still maintain stable estimation performance in complex surface coverage areas and terrain undulation areas. Attached Figure Description
[0031] Figure 1 This is a flowchart of the InSAR residual baseline error estimation method for repeated UAV flights according to the present invention;
[0032] Figure 2 Interference phase diagrams for 15 sets of interference pairs after registration and phase connection;
[0033] Figure 3 This is a schematic diagram of the geometric relationship between two intersecting antennas in InSAR.
[0034] Figure 4 For cross-track InSAR inversion DEM, coherence map and coherence histogram;
[0035] Figure 5 The interferometric phase map after removing the terrain phase;
[0036] Figure 6 The figure shows the results of residual time-varying baseline estimation;
[0037] Figure 7 Interferometric phase diagram after removing residual baseline phase and topographic phase. Detailed Implementation
[0038] The present invention will be further described below with reference to the accompanying drawings and embodiments.
[0039] The core of this invention lies in proposing the concept of a virtual time-varying baseline. Based on the range-Doppler imaging principle, it constructs an equivalent observation geometry by taking the positions of two passing platforms at the same zero-Doppler moment. This allows for direct estimation of the residual motion error (RME) at the SLC (Single Look Complex) and interferogram levels. The virtual time-varying baseline is an equivalent geometric parameter that does not correspond to the real physical baseline. It is used only to characterize phase distortion and observation geometric deviation, without needing to be substituted back into the imaging process. Error compensation is achieved by directly calculating the residual baseline phase based on the estimation results and subtracting it from the interferometric phase, without relying on the original echo, triggering refocusing, or compromising the quality of the focused image. This method significantly reduces computational load and processing time, supports real-time / near-real-time processing, and adapts to the rapid response requirements of UAVs. Simultaneously, it avoids error accumulation introduced by refocusing, resulting in more stable compensation accuracy. It is significantly superior to traditional methods in terms of engineering feasibility, lightweight design, and real-time performance.
[0040] like Figure 1 As shown, the UAV repeated flight InSAR residual baseline error estimation method provided by the present invention includes:
[0041] Step 1, Data Acquisition and Preprocessing: Acquire single-baseline or multi-baseline SAR single-look complex data of main and auxiliary images. Using the selected main image as the reference, perform coarse registration, fine registration and resampling on all auxiliary images to form a registered multi-baseline data stack. Perform phase connection processing on the multi-baseline data stack to obtain a differential interferometric phase map.
[0042] Step 2, Terrain Phase Removal: Using the dual-antenna cross-track InSAR data acquired synchronously with the SAR data, a reference DEM is generated by inversion. Combined with the orbit parameters of the repeated orbit interferometry, the terrain phase is simulated group by group and subtracted from the differential interferometric phase map to obtain the differential interferometric phase map after terrain phase removal.
[0043] Step 3: Direct estimation of residual baseline error: For each azimuth time, based on the range-Doppler imaging principle, the positions of the two passing platforms at the same zero Doppler time are used to form an equivalent observation geometry, and the horizontal and vertical components of the virtual time-varying baseline are defined as parameters to be estimated; based on all range pixels in the differential interferometric phase map after terrain phase removal, an overdetermined nonlinear equation system containing N observation equations and 2 unknown parameters is established, and a robust loss function is used to construct the objective function and solve it to obtain the residual time-varying baseline estimates for each interferometric pair in the corresponding azimuth direction;
[0044] Step 4, Residual Phase Compensation: Based on the estimated residual time-varying baseline and observation geometry, calculate the slant range deviation and residual baseline interferometric phase, and subtract the residual baseline interferometric phase from the differential interferometric phase diagram after removing the topographic phase to complete the residual baseline error compensation.
[0045] Furthermore, in step 1:
[0046] This invention utilizes InSAR flight experiment data from a UAV conducted at a test site. The experiment used a fixed-wing UAV as the platform, equipped with a self-developed SAR imaging system. The system operates at a wavelength of λ=0.03124m and supports two observation modes: dual-antenna single-flyover and single-antenna repeated-flyover. It can simultaneously acquire raw SAR echo data of the same observation area and perform high-precision imaging processing.
[0047] Among them, the dual-antenna single-flyover mode relies on a fixed spatial baseline to achieve synchronous observation, which can quickly acquire short baseline interferometric data and is suitable for the rapid construction of regional DEMs; the single-antenna repeated-flyover mode has the ability to flexibly configure flight trajectories, which can not only acquire long baseline data by shortening the revisit cycle to support high-precision DEM inversion, but also extract surface deformation information through multi-temporal repeated orbit observations, providing core data support for local high-precision surface deformation monitoring.
[0048] This experiment involved 16 flight missions with a total operation time of approximately 4 hours. The average flight speed of the UAV platform was approximately 30.5 m / s, and the radar incident angle covered 35°~60°. All SAR data were acquired using the right-side look-ahead method. In the data processing stage, the azimuth and range resolutions of the SAR imaging were uniformly set to 0.5 m to ensure the consistency of the spatial resolution of the observation data.
[0049] The processing method employed in this invention has high requirements for the coherence of SAR data, necessitating high-precision registration of the single-look complex data after focusing processing. For single-baseline data, master-slave image registration can be performed directly; for multi-baseline data, one image must be selected as the master image, and the remaining images as slave images, each of which must be registered with the master image.
[0050] First, based on the two-dimensional phase correlation method, the overall pixel offset of the primary and secondary SAR amplitude images is solved to achieve large-scale coarse registration. Assume the amplitudes of the primary and secondary images are as follows: and x and y represent pixel coordinates in the spatial domain, which, after a two-dimensional Fourier transform, are: as well as u and v represent two-dimensional frequency domain coordinates, and the cross-power spectrum is calculated using formula (1). :
[0051] (1)
[0052] It represents the conjugate of the frequency domain signal of the auxiliary image.
[0053] The cross-power spectrum is normalized and inverse Fourier transform is performed to obtain the correlation function. The peak coordinates of the correlation function are used as the global offset of the entire image to complete the large-scale coarse registration.
[0054] Based on this, the local offset is solved by grid-by-grid cross-correlation using a spatial sliding window: all offsets are traversed within the search window, and the offset corresponding to the maximum correlation coefficient is taken as the optimal local offset of the current grid; abnormal offsets are eliminated based on the 3σ criterion, and polynomial fitting is performed on the effective offsets to obtain the theoretical offset per pixel of the whole image; finally, high-precision registration of master and slave SAR images is achieved through bilinear interpolation.
[0055] For multi-baseline data stacks, to reduce speckle noise and the impact of temporal decorrelation on the interferometric phase in multi-temporal interferometric data, phase concatenation processing is performed on the stacked data after accurate registration of all images. This is done by selecting a point centered on each image pixel. A local window is used to estimate the complex covariance matrix using all samples within the window. Then, eigenvalue decomposition is performed on the covariance matrix:
[0056] (2)
[0057] V represents the eigenvector matrix. Indicates the conjugate of V. This indicates the conjugate transpose.
[0058] The eigenvector corresponding to the largest eigenvalue is taken as the optimal phase representation of that pixel in the multi-temporal imagery. Using the selected master image as a reference, an equivalent single-reference (ESR) interferogram is constructed. After phase concatenation processing, random phase noise in the multi-baseline interferometric data is significantly suppressed, and coherence is effectively improved. The preprocessed interferometric phase map is shown below. Figure 2 As shown, 1-2...1-16 in the diagram correspond to the indices of the interference phase diagram.
[0059] Furthermore, in step 2:
[0060] This invention uses dual-antenna cross-track InSAR data inversion to generate a reference DEM, and simulates terrain phase components based on the DEM to remove terrain phase from the interferometric phase, thereby better estimating the residual time-varying baseline.
[0061] Dual-antenna cross-track InSAR uses two antennas to simultaneously observe the Earth, acquiring SAR echo data of the same area and imaging it separately. The two single-look complex images are conjugate-multiplied to obtain an interferometric phase map. The surface elevation of the observation area is then calculated using the dual-antenna observation geometry. Its observation geometry is as follows: Figure 3 As shown in the figure. H represents the SAR platform's flight altitude, and B represents the baseline length, i.e., the distance between the phase centers of the two antennas. The distances from point P on the ground to the first and second antennas are respectively... and Let θ be the angle of incidence of the first antenna relative to point P, h represent the target height, and β represent the angle between the baseline and the vertical direction. Then, the phase difference between the signals received by the two antennas is... satisfy:
[0062] (3)
[0063] On the other hand, according to the observation geometry, the target elevation and the angle of incidence satisfy:
[0064] (4)
[0065] Then, the angle of incidence of the target can be obtained from equation (3):
[0066] (5)
[0067] Substituting this into equation (4) yields the target height h:
[0068] (6)
[0069] Due to weak or absent echo signals from water bodies and shaded areas, the corresponding interferometric phase exhibits strong noise characteristics, leading to distortion in direct elevation inversion. Therefore, after completing the elevation calculation, elevation restoration processing is performed on pixels in water bodies and shaded areas to ultimately obtain a high-quality reference DEM. The cross-track InSAR inverted DEM, coherence map, and coherence histogram are shown below. Figure 4 As shown.
[0070] Based on the reference DEM obtained from the cross-track InSAR inversion, and combined with the orbital parameters corresponding to each repeating orbit interferometer pair, the terrain phase is simulated group by group. For any pair of primary and secondary SAR images, the vertical baseline corresponding to the target point can be calculated based on parameters such as the platform's position, attitude, and target slant range during the two observations. The distance between point P on the ground and the first antenna and angle of incidence The terrain phase of the target point can then be expressed as:
[0071] (7)
[0072] Because different repeating orbit interferograms correspond to different orbital positions, each set of interferograms corresponds to different observation geometric parameters. Based on the orbital parameters acquired for each set of SAR data, the vertical baseline, slant range, and incident angle for each interferogram pair can be calculated. Then, the corresponding terrain phase is simulated pixel-by-pixel using the DEM to obtain the simulated terrain phase map for each interferogram. Subsequently, the simulated terrain phase is subtracted from the original interferometric phase to obtain the differential interferometric phase after removing the terrain influence. The interferometric phase map after removing the terrain phase is shown below. Figure 5 As shown.
[0073] Furthermore, in step 3:
[0074] Even after terrain phase removal, the differential interferometric phase still contains deformation phase, residual baseline error phase, and noise phase. To suppress the impact of platform motion error on the accuracy of interferometry, this invention constructs a residual baseline error estimation model based on SAR imaging geometry, enabling direct calculation of the residual baseline error corresponding to each set of interferograms.
[0075] In repeated-track InSAR observations, the two tracks change continuously with azimuth and time, and the spatial baseline between the antenna phase centers is a time-varying baseline. Based on the range-Doppler imaging principle, each pixel after focusing corresponds to a unique zero-Doppler moment, and the scattered echo can be equivalently represented by the platform position observation at that moment. Therefore, the equivalent observation geometry of the two passing platform positions at the same zero-Doppler moment is defined as a virtual time-varying baseline. This virtual baseline is an equivalent model constructed based on the imaging geometry, used to estimate the residual baseline error (RME) and its induced interferometric phase distortion per azimuth, and is not the actual physical baseline.
[0076] For any given time and orientation, assume that the vertical and horizontal components of the virtual baseline in the local coordinate system are respectively... and Then the baseline length can be expressed as: .
[0077] Baseline Inclination satisfy:
[0078] (8)
[0079] according to Figure 3 Geometric relationships in:
[0080] (9)
[0081] This represents the difference in slope distance due to the existence of the baseline, i.e. .
[0082] Substituting equation (9) into equation (4) yields:
[0083] (10)
[0084] but:
[0085] (11)
[0086] According to equation (11), it can be seen that in the case of and In a rectangular coordinate system established with the coordinate axes, this equation represents a circle whose center coordinates are: , radius is .
[0087] For each range cell, a circle can be constructed based on the corresponding slant range and observation geometry. Since the same azimuth row corresponds to the same zero Doppler time, it can be assumed that all range cells within that azimuth row share the same virtual baseline parameters. and Let M be the number of pixels in the azimuth direction and N be the number of pixels in the range direction of the SAR image. Then, for any azimuth row, an overdetermined nonlinear equation system containing N observation equations and 2 unknown parameters can be established from all N range pixels.
[0088] Define the geometric residual corresponding to the i-th observed pixel in the azimuth row. for:
[0089] (12)
[0090] in, Representing the baseline vector:
[0091] (13)
[0092] Then the residual vector It can be represented as:
[0093] (14)
[0094] Considering that local decoherence, registration errors, and anomalous observations may cause some residuals to deviate from the Gaussian distribution, this paper uses the Huber loss function to construct the objective function to improve the robustness of parameter estimation. Its expression is as follows:
[0095] (15)
[0096] in, Huber robust loss function The scaling parameter controls the transition between quadratic and linear losses. Ultimately, the optimal estimate of the virtual baseline parameters is... This can be obtained by solving the following robust nonlinear least squares problem with boundary constraints:
[0097] (16)
[0098] in, and Representing baseline vectors respectively The lower and upper bound vectors for the values are set, and the parameter boundaries are set according to the platform positioning error range provided by the IMU to ensure that the estimation results meet the actual flight conditions. Finally, the residual time-varying baseline of each interferometric pair is obtained. and like Figure 6 As shown.
[0099] Furthermore, in step 4:
[0100] The slope deviation is calculated by estimating the virtual baseline parameters and the observation geometry. :
[0101] (17)
[0102] This allows us to obtain the interferometric phase corresponding to the residual baseline error. :
[0103] (18)
[0104] Subtracting this residual phase from the differential interference phase, the remaining phase is the deformation phase and the noise phase. For example... Figure 7 As shown, the processing results indicate that the compensated phase approaches zero under short-time repeated orbit observations, verifying that the method of this patent can effectively compensate for residual baseline errors and significantly improve the quality of the interferometric phase.
[0105] On the other hand, the present invention provides a UAV repeated flight InSAR residual baseline error estimation system, which includes modules capable of implementing the steps of the aforementioned method, specifically including:
[0106] The differential module is used to acquire single-baseline or multi-baseline SAR single-look complex data. Using the selected main image as a reference, it performs registration and resampling on all auxiliary images to form a registered multi-baseline data stack. Phase connection processing is performed on the multi-baseline data stack to obtain a differential interferometric phase map.
[0107] The terrain removal module is used to invert and generate a reference DEM using dual-antenna cross-track InSAR data acquired synchronously with the SAR data. Combined with the orbit parameters of the repeated orbit interferometer, the terrain phase is simulated group by group and subtracted from the differential interferometric phase map to obtain the terrain-removed differential interferometric phase map.
[0108] The estimation module is used to construct an equivalent observation geometry for the positions of the two passing platforms at the same zero Doppler time for each azimuth time based on the range-Doppler imaging principle, and define the horizontal and vertical components of the virtual time-varying baseline as parameters to be estimated; based on all range pixels in the differential interferometric phase map after terrain phase removal, a robust loss function is used to construct an objective function and solve it to obtain the residual time-varying baseline estimates for each interferometric pair in the corresponding azimuth direction;
[0109] The compensation module is used to calculate the residual baseline interferometric phase based on the residual time-varying baseline estimate and subtract it from the differential interferometric phase diagram after removing the topographic phase, thereby completing the residual baseline error compensation.
[0110] Thirdly, the present invention provides an electronic device, comprising: one or more processors; and a memory for storing one or more programs; wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the aforementioned method for estimating residual baseline errors of repeated UAV flights in InSAR.
[0111] Fourthly, the present invention provides a computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, enable the processor to implement the aforementioned method for estimating residual baseline errors of repeated UAV flights in InSAR.
[0112] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A method for estimating residual baseline errors in InSAR during repeated flights of unmanned aerial vehicles (UAVs), characterized in that, The method includes: Acquire single-baseline or multi-baseline SAR single-look complex data, and perform registration and resampling on all auxiliary images based on the selected main image to form a registered multi-baseline data stack; perform phase connection processing on the multi-baseline data stack to obtain a differential interferometric phase map; Using dual-antenna cross-track InSAR data acquired synchronously with the SAR single-look complex data in the same sortie, a reference DEM is generated by inversion. Combined with the orbit parameters of the repeated track interferometry, the terrain phase is simulated group by group and subtracted from the differential interferometric phase map to obtain the differential interferometric phase map after removing the terrain phase. For each azimuth time, based on the range-Doppler imaging principle, the positions of the two passing platforms at the same zero Doppler time are used to form an equivalent observation geometry, and the horizontal and vertical components of the virtual time-varying baseline are defined as parameters to be estimated. Based on all range pixels in the differential interferometric phase map after terrain phase removal, a robust loss function is used to construct the objective function and solve it to obtain the residual time-varying baseline estimates for each interferometric pair in the corresponding azimuth direction. The residual baseline interferometric phase is calculated based on the residual time-varying baseline estimate and subtracted from the differential interferometric phase diagram after removing the topographic phase, thus completing the residual baseline error compensation.
2. The method for estimating residual baseline error of InSAR during repeated UAV flights according to claim 1, characterized in that, The registration process includes: coarse registration by calculating the overall pixel offset of the main and auxiliary images based on the two-dimensional phase correlation method; then solving the local offset by grid-by-grid using spatial sliding window cross-correlation; removing abnormal offsets based on statistical criteria; obtaining the theoretical pixel-by-pixel offset of the entire image by polynomial fitting of the effective offsets; and achieving fine registration by bilinear interpolation.
3. The method for estimating residual baseline error of UAV repeated flights under InSAR according to claim 1, characterized in that, When performing phase connection processing on the multi-baseline data stack, a local window is selected centered on each pixel, the complex covariance matrix is estimated using all samples within the window, the complex covariance matrix is decomposed into eigenvalues, and the eigenvector corresponding to the largest eigenvalue is taken as the optimal phase representation of the pixel in the multi-temporal image. An equivalent single-reference interferogram is constructed with the selected master image as a reference.
4. The method for estimating residual baseline error of InSAR during repeated UAV flights according to claim 1, characterized in that, The dual-antenna cross-track InSAR data and the SAR single-look complex data are acquired synchronously in the same flight. The surface elevation of the observation area is calculated by cross-track interferometric geometry, and the elevation of water bodies and shadowed pixels is restored to generate a reference digital elevation model.
5. The method for estimating residual baseline error of UAV repeated flights under InSAR according to claim 1, characterized in that, When simulating the terrain phase in groups, the terrain phase is calculated pixel by pixel based on the vertical baseline, slant distance and incident angle corresponding to each repeated track interferometer pair, combined with the reference digital elevation model, and then subtracted from the original interferometric phase to obtain the differential interferometric phase after removing the terrain influence.
6. The method for estimating residual baseline error of InSAR during repeated UAV flights according to claim 1, characterized in that, When constructing the objective function using the robust loss function, the robust loss function is the Huber loss function. The transition between quadratic loss and linear loss is controlled by the scale parameter to suppress the interference of low coherence regions, registration errors and abnormal observations on baseline estimation. The solution of the horizontal and vertical components of the virtual time-varying baseline is transformed into a robust nonlinear least squares problem with boundary constraints. The boundary constraints are set according to the platform positioning error range provided by the navigation system.
7. The method for estimating residual baseline error of InSAR during repeated UAV flights according to claim 1, characterized in that, When calculating the residual baseline interferometric phase based on the estimated residual time-varying baseline, the slant range deviation is first calculated from the estimated virtual baseline parameters and the observation geometry, and then the residual baseline interferometric phase is calculated from the slant range deviation.
8. A system for estimating residual baseline errors in InSAR during repeated flights of unmanned aerial vehicles (UAVs), characterized in that, include: The differential module is used to acquire single-baseline or multi-baseline SAR single-look complex data. Using the selected main image as a reference, it performs registration and resampling on all auxiliary images to form a registered multi-baseline data stack. Phase connection processing is performed on the multi-baseline data stack to obtain a differential interferometric phase map. The terrain removal module is used to invert and generate a reference DEM using dual-antenna cross-track InSAR data acquired synchronously with the SAR single-look complex data. Combined with the orbit parameters of the repeated track interferometry, the terrain phase is simulated group by group and subtracted from the differential interferometric phase map to obtain the terrain-removed differential interferometric phase map. The estimation module is used to construct an equivalent observation geometry for the positions of the two passing platforms at the same zero Doppler time for each azimuth time based on the range-Doppler imaging principle, and define the horizontal and vertical components of the virtual time-varying baseline as parameters to be estimated; based on all range pixels in the differential interferometric phase map after terrain phase removal, a robust loss function is used to construct an objective function and solve it to obtain the residual time-varying baseline estimates for each interferometric pair in the corresponding azimuth direction; The compensation module is used to calculate the residual baseline interferometric phase based on the residual time-varying baseline estimate and subtract it from the differential interferometric phase diagram after removing the topographic phase, thereby completing the residual baseline error compensation.
9. An electronic device, characterized in that, include: One or more processors; Memory, used to store one or more programs; When one or more programs are executed by the one or more processors, the one or more processors implement the UAV repeated flight InSAR residual baseline error estimation method according to any one of claims 1-7.
10. A computer-readable storage medium, characterized in that, It stores executable instructions that, when executed by a processor, enable the processor to implement the UAV repeated flight InSAR residual baseline error estimation method as described in any one of claims 1-7.