Splicing method for maximizing interference area of InSAR main / auxiliary SLC image

By performing phase reference correction and amplitude uniformity in the SLC complex domain, combined with high-fidelity resampling and complex weighted fusion, the seam problem caused by amplitude and phase differences in InSAR stitching is solved, maximizing the interferometric region and improving coherence. It is suitable for InSAR topographic mapping and temporal deformation monitoring.

CN121961837APending Publication Date: 2026-05-01CHINA ACADEMY OF SPACE TECHNOLOGY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202511849562.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-12-09
Publication Date
2026-05-01

AI Technical Summary

Technical Problem

Existing InSAR stitching methods cannot effectively address the issues of brightness discontinuity, phase discontinuity, and reduced coherence at the seams caused by amplitude and phase differences in strip mode, thus affecting the quality of the interferometric region and the accuracy of time series inversion.

Method used

By employing phase reference correction, amplitude gain normalization, unified grid high-fidelity resampling, and transition zone complex weighted fusion, the stitched images are geometrically rigorous, phase-continuous, and amplitude-consistent through unification and fusion in the SLC complex domain.

Benefits of technology

It significantly expands the effective overlap area of ​​the interferometric region, improves coherence and fringe cross-slit traceability, enhances the stability and deformation monitoring accuracy of time-series InSAR, and is suitable for large-scale operational InSAR production.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121961837A_ABST
    Figure CN121961837A_ABST
Patent Text Reader

Abstract

The invention discloses a splicing method for maximizing an InSAR main and auxiliary SLC image interference region, and belongs to the technical field of remote sensing data processing. Comprising the steps of 1, determining a working area and an overlapping area; and 2, correcting the consistency of the phase and the amplitude of the main image. And step 3, main image overlapping region registration is carried out, and a pixel-level geometric corresponding relation is established. And 4, resampling and splicing the main image to obtain a spliced main SLC image. And 5, registering the auxiliary image and the spliced main image. And step 6, re-sampling and splicing the auxiliary image to generate a spliced auxiliary SLC image. According to the method, on the premise of keeping consistency of phase information and amplitude information, the effective coverage area after main / auxiliary splicing is maximized, and seamless splicing of the main / auxiliary SLC images is achieved; the obtained splicing result is natural in strip transition, high in geometric consistency and good in coherence, the precision and robustness of subsequent interference processing and time sequence deformation inversion can be remarkably improved, and the method is particularly suitable for large-range InSAR business production and high-quality splicing of complex scenes.
Need to check novelty before this filing date? Find Prior Art

Description

A mosaicking method for maximizing the interferometric region of InSAR main / secondary SLC images Technical Field

[0001] This invention relates to a stitching method for maximizing the interferometric region of InSAR primary and secondary SLC images, belonging to the field of remote sensing data processing technology. Background Technology

[0002] Interferometric Synthetic Aperture Radar (InSAR), as an active microwave remote sensing technology, can acquire complex echo information of surface scatterers under all-weather and all-time conditions. Its phase observations are extremely sensitive to path differences, making InSAR suitable for both topographic mapping and millimeter-level surface deformation monitoring. In practical operational use, spaceborne SAR often employs stripmap mode to continuously acquire data along the flight azimuth, and then divides it into multiple individual SLC (Single Look Complex) images during ground processing and storage. To improve the spatial coverage and continuity of results in subsequent interferometry and time-series inversion, it is necessary to stitch together multiple primary / secondary SLC images acquired continuously along the azimuth using the same orbit and mode into a single "wide-swath" SLC that is as geometrically, amplitude, and phase-consistent as possible. Unlike stitching across orbits or incident angles, multiple SLC images in stripmap mode are highly consistent in imaging geometry, theoretically making it easier to achieve phase continuity and amplitude uniformity. However, due to factors such as subtle differences in amplitude and phase references, imperfections in registration of overlapping areas, and unreasonable fusion weights and interpolation strategies, simple stitching often results in brightness discontinuities, phase discontinuities, and decreased coherence at the seams, thus affecting the trackability and unwrapping stability of the interferometric fringes. Maximizing the effective interferometric area of ​​the primary / secondary images while maintaining complex phase fidelity has become a key technical challenge in the engineering application of striped InSAR.

[0003] The core of InSAR processing lies in the accurate registration of complex SLC images of the same ground feature acquired at different times or under multiple landscape surveys, and the calculation of complex conjugate products to form interferograms. The interferometric phase contains various components such as topography, deformation, atmosphere, and noise; errors introduced by any of these factors will be amplified and reflected in the phase. Similar to phase filtering or unwrapping problems, the stitching problem can also be viewed as error control and energy redistribution in the complex domain: that is, in the overlapping region, how to achieve statistical consistency in amplitude energy between the primary and secondary images while maintaining phase continuity, thereby maximizing the effective overlap area and average coherence of the interferometric pair. Traditional image stitching research focuses on visual continuity in the amplitude domain, achieving a seamless appearance through normalized brightness and setting transition band weights; however, these methods neglect the strict continuity requirement of the complex phase. For InSAR, any uncontrollable phase offset or local slope at the seam may lead to fringe distortion, breakage, or even non-closure errors during unwrapping, ultimately affecting the stability and accuracy of deformation inversion. On the other hand, if multiple SLC images are orthorectified to geographic coordinates before stitching, although geometrically unified, the phase errors introduced by orthorectification and interpolation in the complex domain are often difficult to fully compensate for, especially in overlapping zones and areas with topographic relief, where path differences, which are extremely sensitive to phase, will manifest as localized decreases in coherence. Therefore, a stitching strategy aimed at maximizing the interferometric region must be directly modeled and processed at the SLC level, ensuring a smooth and synchronous transition of amplitude and phase at the seams through unification, registration, resampling, and weighted fusion in the complex domain.

[0004] In strip mode, the differences between multiple consecutive SLC images along the azimuth direction mainly fall into two categories: amplitude differences and phase differences. Amplitude differences arise from the slow drift of system gain and noise floor, slight inconsistencies in calibration parameters between processed slices, and subtle variations in surface scattering statistics. Without amplitude uniformity, direct stitching will result in uneven brightness at the seams, affecting subsequent amplitude-based weight estimation and coherence assessment. Phase differences are more critical, manifesting as overall phase offset, gradually varying slopes, or local high-frequency discontinuities. Their causes can be simplified as a combination of insufficient registration and a lack of unified reference. Once phase continuity is disrupted at the seams, even if the amplitudes appear seamless, the interferometric quality will still significantly decrease. Engineering practice shows that estimating and eliminating the overall phase offset between master images using robust methods in the overlapping region, while correcting minor geometric deviations based on the complex correlation field of the overlapping region, is a prerequisite for successful subsequent complex-domain fusion.

[0005] Existing stitching methods in InSAR scenarios have three main limitations. First, amplitude-domain stitching emphasizes visual transitions, typically using grayscale normalization and fixed-weight windows for fusion, neglecting complex phase continuity, resulting in a significant reduction in the usable interferometric area near the seams. Second, while orthorectification followed by stitching provides geometric uniformity, the phase damage is significant, especially when the terrain of the overlapping area varies considerably. The phase distortion introduced by orthorectification interpolation causes discontinuous interferometric fringes across the seams, contradicting the goal of maximizing the interferometric area. Third, a few attempts at complex stitching at the SLC level often fail to systematically address the consistency of amplitude and phase, or lack targeted design in weight and interpolation kernel selection, leading to fringe stretching, coherence suppression, or local unwrapping failures at the seams. It can be seen that the key to achieving "maximizing the interferometric area" lies not in whether a certain form of stitching is completed, but in whether the stitching strategy is truly designed collaboratively around the three constraints of complex phase fidelity, amplitude consistency, and geometric rigor.

[0006] In summary, while multiple primary / secondary SLC images acquired consecutively along the azimuth in strip imaging mode exhibit high geometric consistency, subtle differences in amplitude and phase are still unavoidable. Without targeted complex-domain unification and fusion of these differences, brightness discontinuities, phase discontinuities, and decreased coherence will appear at the seams, directly weakening the interferometric region and reducing the quality of unwrapping and temporal inversion. To address this issue, it is necessary to propose a stitching strategy that maximizes the interferometric region for InSAR primary / secondary SLC images. Summary of the Invention

[0007] The technical problem solved by this invention is to overcome the shortcomings of the prior art and provide a stitching method for maximizing the interferometric region of InSAR primary and secondary SLC images. In strip mode, multiple SLC images continuously acquired along the azimuth direction are subjected to phase reference correction, amplitude gain normalization, high-fidelity resampling with a unified grid, and complex weighted fusion of the transition zone to form geometrically rigorous, phase-continuous, and amplitude-consistent stitched primary / secondary SLC data, thereby significantly expanding the interferometric common field of view and improving the coherence preservation and deformation monitoring capabilities of temporal InSAR.

[0008] The technical solution of this invention is:

[0009] A mosaicking method for maximizing the interferometric region of InSAR primary and secondary SLC images includes:

[0010] Step 1: Determine the effective working area and overlap area: Based on precise orbital data, imaging geometric parameters and digital elevation model, calculate the overlap area and effective working area of ​​two adjacent master SLC images on the same date under the surface projection;

[0011] Step 2: Perform phase and amplitude consistency correction of the main image: Based on the estimation of complex phase difference in the overlapping area, apply phase correction to the entire main image to be aligned; based on the statistical estimation of the radiation gain coefficient in the overlapping area, perform overall normalization processing on the amplitude of the main image to be aligned.

[0012] Step 3: Perform registration of the overlapping area of ​​the main image and establish pixel-level geometric correspondence: use a window-based cross-correlation method to estimate the azimuth and range displacements in the overlapping area, and robustly fit to obtain a geometric mapping model from the main image to be aligned to the reference main image.

[0013] Step 4: Perform main image resampling and stitching: The main image to be aligned, after phase and amplitude consistency correction, is resampled to a unified main image grid according to the geometric mapping model. A phase-preserving interpolation kernel and a transition band of a set width are used to perform complex domain weighted fusion in the overlapping area to obtain the stitched main SLC image.

[0014] Step 5: Register the auxiliary images with the stitched master image: Perform coarse registration to fine registration and geometric mapping fitting on the two adjacent auxiliary SLC images of the same date with the stitched master SLC image, and resample them to a unified master image grid.

[0015] Step 6: Perform auxiliary image resampling and stitching: Take the two auxiliary SLC images that have been resampled to the same main image grid in step 5, set a transition zone in the overlapping area of ​​the two auxiliary images, and perform multi-domain weighted fusion in the overlapping area; directly copy the multi-pixel of the corresponding auxiliary image in the non-overlapping area to generate the stitched auxiliary SLC image.

[0016] Furthermore, step 2 involves performing phase and amplitude consistency correction on the main image, specifically including:

[0017] Step 21: In the overlapping area of ​​the two main SLC images, calculate the complex interferometric phase difference by pixel, and statistically average the complex interferometric phase difference to obtain the average phase difference Δφ in the overlapping area.

[0018] Step 22: Use the average phase difference Δφ as the overall phase correction amount and apply it to the complex value of the main image to be aligned. Remove the overall difference of the complex value phase to obtain the phase-corrected main image to be aligned.

[0019] Step 23: Within the overlapping area, based on the amplitude samples of the two main images, calculate the median of the amplitude ratio, and use robust linear regression to estimate the radiation gain coefficient g;

[0020] Step 24: Normalize the amplitude of the phase-corrected main image to be aligned by multiplying its complex amplitude by the radiation gain coefficient g, so that the amplitude ratio of the two main images in the overlapping area is close to 1.

[0021] Furthermore, step 3 involves registering the overlapping areas of the main images and establishing pixel-level geometric correspondences, specifically including:

[0022] Step 31: Set up regular grid control points in the overlapping area of ​​the main image and extract relevant windows of fixed size;

[0023] Step 32: Calculate the complex cross-correlation or coherence cross-correlation for each window, and use a two-dimensional parabola or a two-dimensional Gaussian to fit the sub-pixel peak values ​​to obtain the range and azimuth displacements.

[0024] Step 33: Use RANSAC or M-estimation to remove mismatches, and use affine, low-order polynomial or thin-plate spline models to robustly fit the displacement field to obtain a geometric mapping model from the master image to be aligned to the reference master image.

[0025] Furthermore, step 4 involves resampling and stitching the main image, specifically including:

[0026] Step 41: Select a unified main image output grid and projection parameters, and resample the corrected main image to be aligned to the unified grid using the geometric mapping model;

[0027] Step 42: Resampling is performed using sinc interpolation or sinc interpolation kernel with Kaiser window, and terrain correction is performed in conjunction with digital elevation model to reduce geometric distortion caused by terrain.

[0028] Step 43: Set a transition zone along the seam normal direction in the overlapping area of ​​the main image, and use linear, cosine or Hanning weights to perform fade-in and fade-out weighted fusion in the complex domain to obtain the stitched main SLC image.

[0029] Furthermore, step 6 involves auxiliary image resampling and stitching, specifically as follows:

[0030] Step 61: Select a unified auxiliary image output grid and projection parameters, and resample the two auxiliary SLC images registered in Step 5 to the unified grid according to their respective geometric mappings;

[0031] Step 62: Resampling uses sinc interpolation or sinc interpolation kernel with Kaiser window, and combines it with digital elevation model to perform terrain correction in order to reduce residual geometric distortion caused by terrain;

[0032] Step 63: Set a transition zone along the seam normal direction in the overlapping area of ​​the two auxiliary images, and perform weighted fusion in the complex domain using linear, cosine, or Hanning weights. Directly assign values ​​to the corresponding pixels in the non-overlapping area to obtain the stitched auxiliary SLC image.

[0033] Furthermore, when estimating azimuth and range displacements using a window-based cross-correlation method, the cross-correlation window size is 32×32 to 128×128 pixels, and the step size is 1 / 2 to 1 / 4 of the window side length.

[0034] Furthermore, when performing complex-domain weighted fusion in the overlapping region, the weighting function is a linear, cosine, or Hanning window function, and the transition band width is 50 to 200 pixels; fusion is performed in the complex domain to ensure phase continuity and reduce amplitude jumps.

[0035] Furthermore, the resampling interpolation kernel is a sinc kernel or a sinc kernel with a Kaiser window, and the window function parameter is 4 to 8, in order to maintain phase continuity and reduce ringing effects.

[0036] In a second aspect, the present invention also proposes a non-volatile storage medium comprising: a computer program product, wherein the method is executed when the computer program product is executed.

[0037] Thirdly, the present invention also proposes a computer program product, which includes a computer program that, when executed by a processor, implements the method described.

[0038] The advantages of this invention compared to the prior art are:

[0039] (1) The stitching method proposed in this invention, which aims to maximize the interferometric region of InSAR primary / secondary SLC images, completes phase reference correction, amplitude gain normalization, high-fidelity resampling of unified grid, and complex weighted fusion of transition zones for primary / secondary images based on SLC complex domain information and overlapping area statistics. The fundamental difference between this method and traditional methods that only perform brightness equalization in the amplitude domain or perform orthorectification before stitching lies in the fusion mechanism and reference setting.

[0040] This invention achieves phase continuity and amplitude consistency in the SLC complex domain, employs monotonic weighting along the seam normal direction for fusion, and uses a high-fidelity interpolation kernel to reduce phase distortion introduced by interpolation, thus preventing seam coherence degradation and fringe breakage from the source. Through a processing sequence of "primary followed by secondary, secondary followed by primary," a stitched primary image with a unified reference is first obtained, and then multiple secondary images are coarse-to-fine registered and resampled with a unified grid, significantly improving fringe continuity and coherence preservation in the seam neighborhood and maximizing the primary / secondary co-interference field of view.

[0041] (2) This invention not only effectively solves the engineering bottlenecks such as brightness discontinuity, phase discontinuity, and coherence reduction that easily occur at the seams of strip SLC, but also designs the stitching process from the perspective of complex signal processing and interferometric availability, which is more in line with the essential requirements of InSAR imaging and interferometric measurement for phase fidelity. The algorithm process is clear, simple to implement, and highly efficient. It can run automatically without relying on external DEM fine correction or complex priors. Only the main / auxiliary SLCs of the same track and mode and conventional track and geometric information are required to produce geometrically strict, phase-continuous, and amplitude-consistent stitched main / auxiliary images.

[0042] (3) The method of the present invention significantly expands the effective overlap area of ​​the interferometric pairs, improves coherence and fringe cross-slit traceability, thereby improving the spatial integrity and accuracy of the density of stable scatterers and deformation estimation in time-series InSAR, which is convenient for large-scale and engineering-based promotion and application, and promotes the operational implementation of InSAR in surface deformation monitoring, geological disaster early warning and urban infrastructure safety assessment. Attached Figure Description

[0043] Figure 1 is a flowchart illustrating the splicing method provided by the present invention;

[0044] Figure 2 is an example diagram of the splicing method provided by the present invention;

[0045] Figure 3 is a schematic diagram comparing the main image before and after stitching, showing that after complex domain phase correction and amplitude consistency, the seams disappear, the texture is continuous, and the grayscale transition is smooth.

[0046] Figure 4 is a schematic diagram comparing the auxiliary images before and after stitching, showing that after applying the process of the present invention, the stitching is natural, the details are coherent, and the radiation consistency is significantly improved. Detailed Implementation

[0047] The specific embodiments of the present invention will now be described in further detail with reference to the accompanying drawings.

[0048] This invention provides a mosaicking strategy for maximizing the interferometric region of InSAR primary / secondary SLC images. This method can be applied to InSAR topographic mapping, temporal deformation monitoring, and other fields. The method maximizes the effective coverage area after primary / secondary mosaicking while maintaining consistency in phase and amplitude information, achieving seamless mosaicking of primary / secondary SLC images. The resulting mosaicking results exhibit natural stripe transitions, high geometric consistency, and good coherence, significantly improving the accuracy and robustness of subsequent interferometric processing and temporal deformation inversion. It is particularly suitable for high-quality mosaicking in large-scale InSAR operational production and complex scenarios.

[0049] Referring to Figures 1 and 2, a flowchart illustrating the stitching method for maximizing the interferometric region of InSAR primary / secondary SLC images provided by this invention is shown. The method mainly includes the following steps:

[0050] Step 1: Determine the working area and overlap area. Based on precise orbital data, imaging geometry parameters, and digital elevation model, calculate the overlap area and effective working area of ​​two adjacent master SLC images from the same date under the surface projection;

[0051] Step 2: Perform phase and amplitude consistency correction of the main image: Based on the estimation of complex phase difference in the overlapping area, apply phase correction to the entire main image to be aligned; based on the statistical estimation of the radiation gain coefficient in the overlapping area, perform overall normalization processing on the amplitude of the main image to be aligned.

[0052] Step 2.1: In the overlapping area of ​​the two main SLC images, calculate the complex interferometric phase difference by pixel, and statistically average the complex interferometric phase difference to obtain the average phase difference Δφ in the overlapping area.

[0053] Step 2.2: Use the average phase difference Δφ as the overall phase correction amount and apply it to the complex values ​​of the main image to be aligned. Remove the overall difference of the complex phase values ​​to obtain the phase-corrected main image to be aligned.

[0054] Step 2.3: Within the overlapping area, based on the amplitude samples of the two main images, calculate the median of the amplitude ratio, and use robust linear regression to estimate the radiation gain coefficient g;

[0055] Step 2.4: Normalize the amplitude of the phase-corrected main image to be aligned by multiplying its complex amplitude by the radiation gain coefficient g, so that the amplitude ratio of the two main images in the overlapping area is close to 1.

[0056] Step 3: Perform master image overlap region registration and establish pixel-level geometric correspondence. A window-based cross-correlation method is used to estimate the azimuth and range displacements in the overlap region, and a robust fitting is performed to obtain a geometric mapping model from the master image to the reference master image. Specifically, this includes:

[0057] Step 3.1: Set up regular grid control points in the overlapping area of ​​the main image and extract relevant windows of fixed size;

[0058] Step 3.2: Calculate the complex cross-correlation or coherence cross-correlation for each window, and use a two-dimensional parabola or a two-dimensional Gaussian to fit the sub-pixel peak values ​​to obtain the range and azimuth displacements;

[0059] Step 3.3: Use RANSAC or M-estimation to remove mismatches, and use affine, low-order polynomial or thin plate spline models to robustly fit the displacement field to obtain a geometric mapping model from the master image to be aligned to the reference master image.

[0060] Step 4: Perform master image resampling and stitching. The corrected master image to be aligned is resampled to a unified master image grid according to the described geometric mapping. A phase-preserving interpolation kernel and a transition band of a set width are used to perform complex-domain weighted fusion in the overlapping area to obtain the stitched master SLC image; specifically including:

[0061] Step 4.1: Select a unified main image output grid and projection parameters, and resample the corrected main image to be aligned to the unified grid using the geometric mapping model;

[0062] Step 4.2: Resampling uses sinc interpolation or sinc interpolation kernel with Kaiser window, and combines it with digital elevation model to perform terrain correction to reduce geometric distortion caused by terrain;

[0063] Step 4.3: Set a transition zone along the seam normal in the overlapping area of ​​the main image, and use linear, cosine or Hanning weights to perform fade-in and fade-out weighted fusion in the complex domain to obtain the stitched main SLC image.

[0064] Step 5: Register the auxiliary images with the stitched main image. Perform coarse-to-fine registration and geometric mapping fitting between the two adjacent auxiliary SLC images from the same date and the stitched main SLC image, and resample them to a unified main image grid;

[0065] Step 6: Perform auxiliary image resampling and stitching. For the two auxiliary SLC images resampled to the same master image grid in Step 5, a transition band is set in their overlapping areas, and multi-domain weighted fusion is performed in the overlapping areas; in the non-overlapping areas, the multi-pixel counts of the corresponding auxiliary images are directly copied to generate the stitched auxiliary SLC image. Specifically:

[0066] Step 6.1: Select a unified auxiliary image output grid and projection parameters, and resample the two auxiliary SLC images registered in Step 5 to the unified grid according to their respective geometric mappings;

[0067] Step 6.2: Resampling uses sinc interpolation or sinc interpolation kernel with Kaiser window, and combines it with digital elevation model to perform terrain correction in order to reduce residual geometric distortion caused by terrain;

[0068] Step 6.3: Set a transition zone along the seam normal direction in the overlapping area of ​​the two auxiliary images, and perform weighted fusion in the complex domain using linear, cosine, or Hanning weights. Directly assign values ​​to the corresponding pixels in the non-overlapping area to obtain the stitched auxiliary SLC image.

[0069] Preferably, when estimating the azimuth and range displacements using a window-based cross-correlation method, the cross-correlation window size is 32×32 to 128×128 pixels, and the step size is 1 / 2 to 1 / 4 of the window side length.

[0070] Preferably, when performing complex-domain weighted fusion in the overlapping region, the weighting function is a linear, cosine, or Hanning window function, and the transition band width is 50 to 200 pixels; the fusion is performed in the complex domain to ensure phase continuity and reduce amplitude jumps.

[0071] Preferably, the resampling interpolation kernel is a sinc kernel or a sinc kernel with a Kaiser window, and the window function parameter is 4 to 8, in order to maintain phase continuity and reduce ringing effect.

[0072] Example:

[0073] When performing phase and amplitude consistency correction of the main image, this invention can be further implemented according to the following steps:

[0074] Phase coherence correction:

[0075] In the main image phase correction process of this invention, a concise phase correction model is constructed based on the interferometric phase of the stitched SLC in the overlapping region:

[0076] First, the interference quantity is calculated within a unified grid or common coverage area. Under ideal registration, the phase of this quantity represents the residual phase offset and gradual slope between image 1 and image 2 to be stitched.

[0077] Subsequently, the complex samples of the interferogram are summed along the azimuth direction in the range direction and their arguments are taken to obtain the azimuth phase modulation sequence. Then, phase unwrapping is performed on the one-dimensional phase sequence to eliminate the 2π jump and obtain a continuous phase sequence.

[0078] The continuous phase sequence is fitted with a first-order polynomial using least squares to obtain the azimuth residual phase slope m and intercept b. This linear term physically corresponds to the residual phase slope and overall phase offset caused by slight deviations in the azimuth spectrum or Doppler center of the two images to be stitched.

[0079] After obtaining the parameters m and b, a complex exponential compensation is applied to the image to be stitched, multiplying the pixels at the azimuth index k by exp(j·(m·k+b)) to align the phase references of the two images. The correction effect can be verified by reconstructing the cross-interferogram and checking that its phase is close to zero (the phase residual near the seam should converge significantly).

[0080] Amplitude consistency correction:

[0081] In the main image amplitude correction process of this invention, the present invention constructs a simple and robust amplitude consistency model based on the amplitude statistics of the complex SLC in the overlapping area:

[0082] First, the amplitude ratio between the two images to be stitched is calculated pixel by pixel in the public coverage area. This ratio describes the radiometric scale deviation caused by differences in system gain, noise floor, and processing slices between the two images.

[0083] Data cleaning is then performed to avoid outliers interfering with statistical robustness; the statistical mean and standard deviation are calculated on the cleaned data, and the statistical mean is used as an unbiased estimate of the global amplitude gain coefficient, which physically represents the scaling factor required to scale the amplitude of image 1 to the amplitude reference of image 2.

[0084] After obtaining the statistical mean, the complex data of the image to be stitched 1 are linearly calibrated only in terms of amplitude, so as to ensure that the complex phase is not affected and the amplitude statistics are consistent with the main image.

[0085] In this embodiment, based on satellite data, the stitching example result is as follows:

[0086] Figure 3 shows a comparison between the main image before (left) and after (right) stitching: Before stitching, there are obvious brightness discontinuities and texture discontinuities at the strip seams. After stitching, through complex domain phase reference correction and amplitude uniformity, the grayscale transition in the seam area is smooth, the texture structure is continuous, and no perceptible seam traces are observed. The overall radiation field consistency is significantly improved. Figure 4 shows the effect of the auxiliary image before and after stitching: The original seam area is affected by gain difference and geometric micro-mismatch, resulting in bright and dark bands and detail misalignment. After using the method of this invention, the auxiliary image is consistent with the reference of the main image on a unified grid, the details at the seam are coherent, the contrast is stable, and the spatial frequency components between strips are naturally connected. Combining Figures 3 and 4, it can be seen that this invention can simultaneously ensure the geometric, radiation, and phase uniformity of the main / auxiliary SLC, significantly expand the effective common field of view, and provide high-quality input for subsequent interferometric construction and temporal deformation inversion.

[0087] In summary, this invention proposes an integrated stitching and unification method for the complex domain of InSAR main / auxiliary SLC, which maximizes the common field of view of the interferometer while maintaining fringe continuity across slits. Analyzing from the perspective of complex signals, this invention presents a strategy for constructing interferograms in the overlapping region to estimate the residual phase model and global amplitude ratio. High-fidelity resampling and complex-weighted fusion of the transition zone are implemented on a unified grid, fundamentally eliminating phase discontinuities and brightness breaks at the seams. Unlike traditional amplitude domain equalization or orthorectification followed by stitching methods, this invention directly performs phase reference alignment and amplitude unification in the SLC complex domain, filtering out linear phase ramps. Simultaneously, a global gain factor is used to achieve amplitude scale unification, obtaining stable results without iterative optimization. Experimental verification using main / auxiliary strip stitching of Gaofen-3 measured data demonstrates that this invention can significantly expand the effective interferometric region, improve seam coherence and texture continuity, providing high-quality input for subsequent interferometric generation and temporal InSAR deformation inversion. This method is simple in process, has high computational efficiency, and is easy to deploy in engineering. It is suitable for large-scale, multi-temporal data monitoring of surface deformation and operational topographic mapping applications.

[0088] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

[0089] The parts of this invention not described in detail are common knowledge to those skilled in the art.

Claims

1. A mosaicking method for maximizing the interferometric region of InSAR primary and secondary SLC images, characterized in that, The process includes: Step 1: Determining the effective working area and overlapping area: Based on precise orbital data, imaging geometric parameters, and digital elevation models, calculate the overlapping area and effective working area of ​​two adjacent master SLC images from the same date under the surface projection; Step 2: Performing phase and amplitude consistency correction for the master images: Based on the estimation of complex phase difference in the overlapping area, apply phase correction to the entire master image to be aligned; estimate the radiation gain coefficient based on the amplitude statistics of the overlapping area, and normalize the amplitude of the master image to be aligned as a whole; Step 3: Performing master image overlapping area registration and establishing pixel-level geometric correspondence: Use a window-based cross-correlation method to estimate the azimuth and range displacements in the overlapping area, and robustly fit to obtain a geometric mapping model from the master image to be aligned to the reference master image; Step 4: Performing master image resampling and stitching: Resample and stitch the master images based on phase and amplitude... Step 5: Register the auxiliary images with the stitched main image: Perform coarse-to-fine registration and geometric mapping fitting on two adjacent auxiliary SLC images of the same date with the stitched main SLC image, and resample them to the unified main image grid. Step 6: Resample and stitch the auxiliary images: Set a transition band in the overlapping area of ​​the two auxiliary SLC images resampled to the unified main image grid in Step 5, and perform complex-domain weighted fusion in the overlapping area. Copy the complex pixels of the corresponding auxiliary images directly in the non-overlapping area to generate the stitched auxiliary SLC image.

2. The stitching method for maximizing the interferometric region of InSAR primary and secondary SLC images according to claim 1, characterized in that: Step 2, which performs phase and amplitude consistency correction of the main image, specifically includes: Step 21: In the overlapping area of ​​the two main SLC images, the complex interferometric phase difference is calculated per pixel, and the complex interferometric phase difference is statistically averaged to obtain the average phase difference Δφ in the overlapping area; Step 22: The average phase difference Δφ is used as the overall phase correction amount and applied to the complex values ​​of the main image to be aligned, and the overall difference of the complex value phase is removed to obtain the phase-corrected main image to be aligned; Step 23: In the overlapping area, based on the amplitude samples of the two main images, the median of the amplitude ratio is calculated, and the radiation gain coefficient g is estimated using robust linear regression; Step 24: The amplitude of the phase-corrected main image to be aligned is normalized, and its complex amplitude is multiplied by the radiation gain coefficient g, so that the amplitude ratio of the two main images in the overlapping area is close to 1.

3. The stitching method for maximizing the interferometric region of InSAR primary and secondary SLC images according to claim 1, characterized in that: Step 3 involves registering the overlapping area of ​​the main image and establishing pixel-level geometric correspondence. Specifically, this includes: Step 31: Setting up regular grid control points in the overlapping area of ​​the main image and extracting relevant windows of fixed size; Step 32: Calculating complex cross-correlation or coherence cross-correlation for each window, and using a two-dimensional parabola or two-dimensional Gaussian to fit sub-pixel peak values ​​to obtain range and azimuth displacements; Step 33: Using RANSAC or M-estimation to remove mismatches, and using affine, low-order polynomial, or thin-plate spline models to robustly fit the displacement field to obtain a geometric mapping model from the main image to be aligned to the reference main image.

4. The stitching method for maximizing the interferometric region of InSAR primary and secondary SLC images according to claim 1, characterized in that: Step 4, which involves resampling and stitching the main image, specifically includes: Step 41: Selecting a unified main image output grid and projection parameters, and resampling the corrected main image to be aligned to the unified grid using the geometric mapping model; Step 42: Resampling uses sinc interpolation or a sinc interpolation kernel with a Kaiser window, and combines it with a digital elevation model to perform terrain correction to reduce geometric distortion caused by terrain; Step 43: Setting a transition zone along the seam normal direction in the overlapping area of ​​the main image, and using linear, cosine, or Hanning weights to perform gradual in-and-out weighted fusion in the complex domain to obtain the stitched main SLC image.

5. A stitching method for maximizing the interferometric region of InSAR primary and secondary SLC images according to claim 1, characterized in that: Step 6 involves resampling and stitching the auxiliary images, specifically: Step 61: Select a unified auxiliary image output grid and projection parameters, and resample the two auxiliary SLC images registered in Step 5 to the unified grid according to their respective geometric mappings; Step 62: Resampling uses sinc interpolation or a sinc interpolation kernel with a Kaiser window, and combines it with a digital elevation model to perform terrain correction in order to reduce residual geometric distortion caused by terrain; Step 63: Set a transition zone along the seam normal direction in the overlapping area of ​​the two auxiliary images, and use linear, cosine, or Hanning weights to perform weighted fusion in the complex domain, and directly assign values ​​to the corresponding pixels in the non-overlapping area to obtain the stitched auxiliary SLC image.

6. A stitching method for maximizing the interferometric region of InSAR primary and secondary SLC images according to any one of claims 1-5, characterized in that: When estimating azimuth and range displacements using a window-based cross-correlation method, the cross-correlation window size is 32×32 to 128×128 pixels, and the step size is 1 / 2 to 1 / 4 of the window side length.

7. A stitching method for maximizing the interferometric region of InSAR primary and secondary SLC images according to any one of claims 1-5, characterized in that: When performing complex-domain weighted fusion in the overlapping region, the weighting function is a linear, cosine, or Hanning window function, and the transition band width is 50 to 200 pixels; fusion is performed in the complex domain to ensure phase continuity and reduce amplitude jumps.

8. A stitching method for maximizing the interferometric region of InSAR primary and secondary SLC images according to any one of claims 1-5, characterized in that: The resampling interpolation kernel is a sinc kernel or a sinc kernel with a Kaiser window, with a window function parameter of 4 to 8, to maintain phase continuity and reduce ringing effects.

9. A non-volatile storage medium, characterized in that, include: A computer program product that, when executed, performs the method described in any one of claims 1 to 8.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the steps of the method according to any one of claims 1 to 8.