Method, device and equipment for determining surface large gradient time sequence deformation and medium

Abstract

The application discloses a surface large-gradient time-series deformation determination method and device, equipment and medium, and relates to the field of data processing. SAR satellite images are processed into SLC images, and a super master image is screened out from all the SLC images; based on the super master image, all the SLC images are registered; based on all the registered SLC images, a grouping threshold of all the SLC images is determined; based on the grouping threshold, all the registered SLC images are grouped to obtain a SLC image subset; short-time sequence deformation of the SLC image subset is determined; based on the short-time sequence deformation of the SLC image subset, a final offset of a master image in adjacent SLC image subsets is determined by using a robust estimation; and based on the final offset of the master image in all the adjacent SLC image subsets, short-time sequence offsets of all the SLC image subsets are connected to obtain surface large-gradient time-series deformation.

Description

Technical Field

[0001] This application relates to the technical field of data processing, and in particular to a method, apparatus, device and medium for determining large gradient temporal deformation of the Earth's surface. Background Technology

[0002] With the continuous development of SAR (Synthetic Aperture Radar) technology and the increasing observation frequency, massive amounts of SAR images provide rich data support for in-depth analysis of dynamic changes on the Earth's surface, bringing unprecedented opportunities for surface monitoring activities. At the same time, it also puts forward higher requirements for efficient and accurate data processing.

[0003] In related technologies, SAR pixel offset tracking (POT) is used to monitor large gradient deformation of SAR images. SAR pixel offset tracking technology calculates the planar displacement (including azimuth and range) of corresponding pixels in two SAR images through cross-correlation, and then obtains the results of two-dimensional deformation changes in the region over time through time series analysis.

[0004] The PO-SBAS (Pixel Offset-Small Baseline Subsets) method, employing a small baseline strategy, is widely used in time series analysis. However, in large gradient deformation monitoring, the PO-SBAS method, influenced by the multi-master image strategy, may experience pixel mismatches (treating non-corresponding points as corresponding points during time series calculations) when the deformation reaches the pixel level (i.e., deformation greater than one pixel). Specifically, in large gradient deformation, when the target offset exceeds one pixel, the target's position changes, and with the accumulation of time series data, the target's position will have a significant offset compared to its initial position. In conventional PO-SBAS time series calculations, it is generally assumed that the pixel position will not change significantly over time. Applying this to pixel offset tracking easily leads to pixel mismatches, resulting in misestimation of the time series calculation results. Meanwhile, during time-series computation, due to the significant pixel offsets, gross errors are unavoidable when performing sequential time-series connections (significant deviations between measured results and true values ​​under the same observation conditions, with values ​​far exceeding the maximum random error that might occur under normal observation conditions). When gross errors from the observation data participate in the computation, the results of the time-series solution will be significantly affected. Therefore, the impact of outliers in the observation data needs to be considered when performing sequential time-series computation. Summary of the Invention

[0005] This application provides a method, apparatus, device, and medium for determining large-gradient temporal deformation of the Earth's surface. The method involves acquiring a set of SAR satellite images; processing the SAR satellite images into SLC images and selecting a super master image from all the SLC images; registering all the SLC images based on the super master image; determining a grouping threshold for all the registered SLC images based on the grouping threshold; and grouping all the registered SLC images based on the grouping threshold to obtain SLC image subsets. This allows SLC images with significant differences to be grouped into different SLC image subsets, subsequently determining the short-time series deformation of individual SLC image subsets. This solves the problem of mismatched pixel targets caused by significant changes in pixel position in the PO-SBAS method. Furthermore, by introducing robust estimation, even if gross errors are mixed in the short-time series deformation of different SLC image subsets, their influence can be minimized through iterative weighting. This ensures high accuracy in obtaining the large-gradient temporal deformation of the Earth's surface after connecting the short-time series deformations of different SLC image subsets.

[0006] In a first aspect, embodiments of this application provide a method for determining temporal deformation of large surface gradients, including: Acquire a set of SAR satellite images, which includes multiple SAR satellite images of the study area taken by satellites in the same repeating orbit and ordered by the time of capture. SAR satellite imagery is processed into SLC imagery; a super master imagery is selected from all the SLC imagery; and all the SLC imagery are registered based on the super master imagery. Based on all registered SLC images, a grouping threshold for all SLC images is determined; based on the grouping threshold, all registered SLC images are grouped to obtain a subset of SLC images; the grouping threshold includes a temporal baseline threshold and / or a spatial baseline threshold. Determine the short-time series deformation of the SLC image subset; Based on the short-time series deformation of the SLC image subset, robust estimation is used to determine the final offset of the main image in adjacent SLC image subsets; Based on the final offset of the main image in all adjacent SLC image subsets, the short time series offsets of all SLC image subsets are concatenated to obtain the large gradient temporal deformation of the land surface.

[0007] In some embodiments, the step of determining the grouping threshold for all SLC images based on all registered SLC images includes: The search window is determined based on the deformation characteristics of the surface study area; Based on the search window, pixel offset tracking is performed on the registered SLC image using a preset step size to obtain tracking results; wherein, the tracking results include determining the influence of the temporal baseline on the number of matching pixels within a preset spatial baseline range, and determining the influence of the spatial baseline on the number of matching pixels within a preset temporal baseline range. Based on the tracking results, a grouping threshold is determined; wherein the grouping threshold includes a temporal baseline threshold and / or a spatial baseline threshold.

[0008] In some embodiments, the step of grouping all the registered SLC images based on a grouping threshold to obtain a subset of SLC images includes: All the registered SLC images are sorted according to the acquisition time of the corresponding SAR satellite images to obtain an image sequence; Using the first SLC image in the image sequence as the starting image of the first SLC image subset, and obtaining the next SLC image of the starting image as the target SLC image, it is determined whether the baseline between the target SLC image and the starting image is greater than a grouping threshold; wherein, when the grouping threshold only includes a temporal baseline threshold, the baseline includes a temporal baseline; when the grouping threshold only includes a spatial baseline threshold, the baseline includes a spatial baseline; when the grouping threshold includes both a temporal baseline threshold and a spatial baseline threshold, the baseline includes both a temporal baseline and a spatial baseline. If the value is not greater than the grouping threshold, the SLC image of the next scene is added to the current SLC image subset, and the SLC image of the next scene is acquired and replaced with the target SLC image. The step of determining whether the baseline between the target SLC image and the starting image is greater than the grouping threshold is executed again. If the value is greater than the grouping threshold, the SLC image of the next scene is used as the starting image of the next SLC image subset; and the process of obtaining the SLC image of the next scene after the starting image is repeated as the target SLC image. It is then determined whether the baseline between the target SLC image and the starting image is greater than the grouping threshold, until all SLC images in the image sequence are divided into the corresponding SLC image subsets.

[0009] In some embodiments, the step of determining the short-time series deformation of the SLC image subset includes: Select a master image from the subset of SLC images; Determine the pixel offset observation values ​​of all SLC images in the SLC image subset and the preset pixel points of the main image; combine all the observation values ​​into an observation value set; Substituting the set of observations into the observation equation and solving it through adjustment, the short-time series deformation of the SLC image subset is obtained; the short-time series deformation of the SLC image subset includes the offsets of the 2nd to nth SLC images in the SLC image subset from the 1st SLC image, and the SLC image subset includes n SLC images.

[0010] In some embodiments, the step of determining the final offset of the master image in adjacent SLC image subsets using robust estimation based on the short-time-series deformation of the SLC image subsets includes: Based on the short-time series deformation of all SLC image subsets, an offset set of the main images in adjacent SLC image subsets is determined. The offset set includes a direct offset and an indirect offset set. The direct offset is the offset of the main image in an adjacent SLC image subset. The indirect offset set includes several indirect offsets, the number of which is the number of SLC images located between two main images. The indirect offset is the difference between the offset of each main image in an adjacent SLC image subset and the offset of an SLC image located between two main images. Using the offset set, robust estimation is performed to obtain the final offset of the main image in the adjacent SLC image subset.

[0011] In some embodiments, the step of selecting a super master image from all the SLC images includes: SLC images that meet the first preset time baseline requirements, the first preset spatial baseline requirements, and the first preset Doppler center frequency requirements are selected from all the SLC images and used as super master images.

[0012] In some embodiments, the step of registering all the SLC images based on the super master image includes: Obtain DEM data of the study area; Based on the DEM data of the super master image and SLC images, all the SLC images are registered based on the super master image.

[0013] Secondly, embodiments of this application provide a device for determining temporal deformation of large surface gradients, comprising: The acquisition unit is used to acquire a SAR satellite image set, which includes multiple SAR satellite images taken by satellites in the same repeating orbit over the research area on the Earth's surface and ordered by the time of capture. A registration unit is used to process SAR satellite images into SLC images; select a super master image from all the SLC images; and register all the SLC images based on the super master image. A grouping unit is used to determine a grouping threshold for all registered SLC images based on all registered SLC images; and to group all registered SLC images based on the grouping threshold to obtain a subset of SLC images; the grouping threshold includes a temporal baseline threshold and / or a spatial baseline threshold. The first determining unit is used to determine the short-time series deformation of the SLC image subset; The second determining unit is used to determine the final offset of the main image in adjacent SLC image subsets based on the short time series deformation of the SLC image subset using robust estimation. The connection unit is used to connect the short time series offsets of all SLC image subsets based on the final offset of the main image in all adjacent SLC image subsets to obtain the large gradient temporal deformation of the land surface.

[0014] Thirdly, embodiments of this application provide a computer device, including a processor and a memory, wherein the memory stores a program or instructions that run on the processor, and the program or instructions, when executed by the processor, implement the steps of the method for determining large gradient temporal deformation of the earth's surface as described above.

[0015] Fourthly, embodiments of this application provide a readable storage medium storing a program or instructions thereon, which, when executed by a processor, implement the steps of the method for determining large gradient temporal deformation of the Earth's surface as described above.

[0016] The above embodiments provide a method, apparatus, device, and medium for determining large-gradient temporal deformation of the Earth's surface. The method involves acquiring a set of SAR satellite images; processing the SAR satellite images into SLC images and selecting a super master image from all the SLC images; registering all the SLC images based on the super master image; determining a grouping threshold for all the registered SLC images based on all the registered SLC images; and grouping all the registered SLC images based on the grouping threshold to obtain SLC image subsets. This allows SLC images with significant differences to be grouped into different SLC image subsets, subsequently determining the short-time series deformation of individual SLC image subsets. This solves the problem of mismatched pixel targets caused by significant changes in pixel position in the PO-SBAS method. Furthermore, by introducing robust estimation, even if there are gross errors mixed in the short-time series deformation of different SLC image subsets, their influence can be minimized through iterative weighting. This ensures high accuracy in obtaining the large-gradient temporal deformation of the Earth's surface after connecting the short-time series deformations of different SLC image subsets. Attached Figure Description

[0017] Figure 1 An exemplary flowchart is shown for a method for determining temporal deformation of large surface gradients according to some embodiments; Figure 2 An example diagram illustrating a tracking result is shown; Figure 3 (a) is a spatial distribution diagram of a time-series deformation field used in a simulation experiment; Figure 3 (b) is the time variation curve of the time series deformation field used in a simulation experiment; Figure 4 (a1)–(a6) are schematic diagrams of the simulated cumulative deformation at different observation times; Figure 4 (b1)–(b6) are schematic diagrams of PO-SBAS inversion results at different observation times; Figure 4 (c1)–(c6) are schematic diagrams of PO-SBAS estimation errors at different observation times; Figure 4 (d1)–(d6) are schematic diagrams of RS-POT inversion results at different observation times; Figure 4 (e1)–(e6) are schematic diagrams of RS-POT estimation errors at different observation time phases; Figure 5 (a) is a schematic diagram simulating cross-sectional deformation; Figure 5(b) is a schematic diagram of a PO-SBAS inversion result; Figure 5 (c) is a schematic diagram of an RS-POT inversion result; Figure 6 (a) is a schematic diagram showing the distribution of points P1, P2, and P3; Figure 6 (b)–(d) are schematic diagrams of the time-series deformation curves of P1, P2, and P3, respectively; Figure 7 (a)–(c) are schematic diagrams of the simulated deformation fields corresponding to traction landslides, shovel-type landslides and homogeneous landslides, respectively. Figure 7 (d)–(f) represent the PO values ​​corresponding to traction landslides, kinetic landslides, and homogeneous landslides, respectively. A schematic diagram illustrating the error estimation using the SBAS method; Figure 7 (g)–(i) represent RS corresponding to traction landslides, kinetic landslides, and homogeneous landslides, respectively. A schematic diagram illustrating the error estimation using the POT method; Figure 8 (a) is a distribution map of the PO-SBAS baseline combination; Figure 8 (b) is a distribution map of RS-POT baseline combinations; Figure 9 (a1)–(a3) are schematic diagrams of azimuth deformation retrieved by PO-SBAS at different times; Figure 9 (b1)–(b3) are schematic diagrams of RS-POT inversion azimuth deformation at different times; Figure 9 (c1)–(c3) are schematic diagrams of the azimuth deformation difference values ​​of the two methods at different times; Figure 9 (d1)–(d3) are schematic diagrams of LOS deformation inverted by PO-SBAS at different times; Figure 9 (e1)–(e3) are schematic diagrams of LOS-directed deformation retrieved by RS-POT at different times; Figure 9 (f1)–(f3) are schematic diagrams of the LOS deformation difference values ​​for the two methods at different times. Figure 10 (a), (c), (e) and (g) are schematic diagrams of the azimuth deformation of points G1-G4; Figure 10 (b), (d), (f), and (h) are schematic diagrams of the LOS-direction deformation of points G1-G4. Detailed Implementation

[0018] To make the objectives and implementation methods of this application clearer, the exemplary implementation methods of this application will be clearly and completely described below with reference to the accompanying drawings of the exemplary embodiments of this application. Obviously, the exemplary embodiments described are only some embodiments of this application, and not all embodiments.

[0019] It should be noted that the brief descriptions of terms in this application are only for the convenience of understanding the embodiments described below, and are not intended to limit the embodiments of this application. Unless otherwise stated, these terms should be understood in their ordinary and common meaning.

[0020] The terms "first," "second," "third," etc., used in the specification and accompanying drawings of this application are used to distinguish similar or related objects or entities, and do not necessarily imply a specific order or sequence, unless otherwise specified. It should be understood that such terms can be used interchangeably where appropriate.

[0021] The terms “comprising” and “having”, and any variations thereof, are intended to cover but not exclude inclusion, for example, a product or device that includes a range of components is not necessarily limited to all of the components that are clearly listed, but may include other components that are not clearly listed or that are inherent to such product or device.

[0022] To address the aforementioned technical problems, this application provides a method, apparatus, device, and medium for determining large-gradient temporal deformation of the Earth's surface. The method involves acquiring a set of SAR satellite images; processing the SAR satellite images into SLC images and selecting a super master image from all the SLC images; registering all the SLC images based on the super master image; determining a grouping threshold for all the registered SLC images based on all the registered SLC images; and grouping all the registered SLC images based on the grouping threshold to obtain SLC image subsets. This allows SLC images with significant differences to be grouped into different SLC image subsets, subsequently determining short-time series deformation for each individual SLC image subset. This solves the problem of mismatched pixel targets caused by significant changes in pixel position in the PO-SBAS method. Furthermore, by introducing robust estimation, even if gross errors are mixed in the short-time series deformations of different SLC image subsets, their influence can be minimized through iterative weighting. This ensures high accuracy in obtaining the large-gradient temporal deformation of the Earth's surface after connecting the short-time series deformations of different SLC image subsets.

[0023] Figure 1 An exemplary flowchart is shown of a method for determining large gradient temporal deformation of the Earth's surface according to some embodiments, the method comprising steps S100-S600.

[0024] S100: Acquire a collection of SAR satellite images.

[0025] The SAR satellite image collection includes multiple SAR satellite images of the study area taken by satellites in the same repeating orbit and ordered by the time of capture.

[0026] For example, satellites repeatedly observe the same surface study area at a fixed revisit cycle (e.g., 5 days), acquiring a SAR satellite image each time they pass over, thus forming an image sequence covering different points in time, i.e., a SAR satellite image set.

[0027] S200. Process SAR satellite images into SLC images (Single Look Complex); select a super master image from all the SLC images, and register all the SLC images based on the super master image.

[0028] In this embodiment of the application, in order to obtain an image containing complete intensity and phase information and provide high-quality input for subsequent pixel offset determination, the SAR satellite image is processed into an SLC image.

[0029] In some embodiments, the step of selecting a super master image from all the SLC images includes: SLC images that meet the first preset time baseline requirements, the first preset spatial baseline requirements, and the first preset Doppler center frequency requirements are selected from all the SLC images and used as super master images.

[0030] In this embodiment, the time baseline refers to the time interval between the capture of two SAR satellite images. The spatial baseline refers to the spatial distance between the orbital positions of the two SAR satellite images at the time of capture. It is understood that although the satellite operates in the same repeating orbit, its orbital position will change slightly each time it operates. The Doppler center frequency is a parameter in SAR satellite image imaging, reflecting the angular relationship between the radar beam direction and the satellite's flight direction.

[0031] In this embodiment, the first preset time baseline requirement is to minimize the average time baseline between the super master image and other SLC images. The first preset spatial baseline requirement is to center the orbit of the super master image, thereby minimizing the average spatial baseline between the super master image and other SLC images. The first preset Doppler center frequency requirement is to minimize the average Doppler center frequency difference between the super master image and other SLC images. In summary, it can be understood that the selected super master image is an SLC image that is as centered as possible in terms of time, orbit, and Doppler center frequency. It should be noted that the time baseline, spatial baseline, and Doppler center frequency mentioned here refer to the time baseline, spatial baseline, and Doppler center frequency between the SAR satellite image processed into a super master image and the SAR satellite image processed into other SLC images.

[0032] In some embodiments, the step of registering all the SLC images based on the super master image includes: Obtain DEM (Digital Elevation Model) data for the surface study area; In this embodiment of the application, DEM is a digital elevation model, which is a two-dimensional raster matrix containing spatial elevation data. Each pixel unit in the matrix stores the elevation value of the corresponding spatial location on the Earth's surface.

[0033] Based on the DEM data of the super master image and SLC images, all the SLC images are registered based on the super master image.

[0034] In this embodiment, the coordinates of the same pixel in all SLC images are mapped to the same physical point on the ground in the DEM data. This can improve the accuracy of the subsequent determination of the large gradient temporal deformation of the land surface.

[0035] S300. Based on all registered SLC images, determine the grouping threshold for all SLC images; based on the grouping threshold, group all registered SLC images to obtain a subset of SLC images; the grouping threshold includes a temporal baseline threshold and / or a spatial baseline threshold.

[0036] In this embodiment, pixel offset tracking calculation is performed based on the registered SLC images, and the change in the number of matched pixels under the condition of limiting the temporal baseline and spatial baseline is statistically analyzed to determine the grouping threshold of all the SLC images.

[0037] In this embodiment, SLC images with significant differences can be divided into different SLC image subsets. Subsequently, short-time series deformation of individual SLC image subsets can be determined, which solves the problem of mismatch of pixel targets caused by significant changes in pixel position in the PO-SBAS method.

[0038] In some embodiments, the step of determining the grouping threshold for all SLC images based on all registered SLC images includes: The search window is determined based on the deformation characteristics of the surface study area; In this embodiment, the search window refers to the range of pixel offset tracking. In one example, the deformation characteristics of the surface study area include the presence of a rapid landslide zone, which is expected to slide tens of meters per year. The search window can be determined based on the size of the rapid landslide zone. A larger rapid landslide zone results in a relatively larger search window. In this embodiment, determining the search window based on the deformation characteristics of the surface study area allows for a more accurate understanding of the changes in the surface study area.

[0039] In some embodiments, the method further includes: determining an oversampling factor based on the deformation characteristics of the surface study area. The oversampling factor is a factor by which the SLC image is interpolated to make the pixels in the SLC image denser, thereby improving the matching accuracy.

[0040] Based on the search window, pixel offset tracking is performed on the super main image and other SLC images using a preset step size to obtain tracking results; wherein, the tracking results include determining the influence of the temporal baseline on the number of matching pixels within a preset spatial baseline range, and determining the influence of the spatial baseline on the number of matching pixels within a preset temporal baseline range. In this embodiment, due to the large number of pixels in the search window, a preset step size is set to avoid slow processing speed. The preset step size can be understood as the sampling interval. Pixels of pixels are collected from the search window of the super main image at preset step intervals, and pixels of pixels are also collected from the search windows of other SLC images at preset step intervals. Pixel offset tracking is performed on the corresponding pixels collected from the super main image and other SLC images to obtain the tracking result.

[0041] Matched pixels refer to pixels that successfully find the same point when performing pixel offset tracking calculations on two SLC images.

[0042] Figure 2 An exemplary diagram illustrating one tracking result is shown. Figure 2The lower horizontal axis represents the temporal baseline axis, where values ​​represent the temporal baselines between different SLC imagery and the super master imagery. The upper horizontal axis represents the spatial baseline axis, where values ​​represent the spatial baselines between different SLC imagery and the super master imagery. The vertical axis represents the number of matched pixels.

[0043] exist Figure 2 The coordinates of the triangle symbol are determined within a preset time baseline range. The change in the triangle symbol's position reveals how the number of matching pixels changes with the spatial baseline, illustrating the impact of determining the spatial baseline within the preset time baseline range on the number of matching pixels. Figure 2 The coordinates of the circle symbol are determined within a preset spatial baseline. The change in the position of the circle symbol reveals how the number of matched pixels changes with the time baseline; that is, the impact of determining the time baseline within the preset spatial baseline range on the number of matched pixels.

[0044] Based on the tracking results, a grouping threshold is determined; wherein the grouping threshold includes a temporal baseline threshold and / or a spatial baseline threshold.

[0045] like Figure 2 As shown, the impact of determining the time baseline within the preset spatial baseline range on the number of matched pixels shows that the number of matched pixels drops sharply at -180 days and 180 days. Therefore, the time baseline threshold is determined to be 180 days.

[0046] The impact of determining the spatial baseline within the preset time baseline range on the number of matching pixels shows that there is no sharp decline in the number of matching pixels, therefore, based on Figure 2 The content of this study does not specify the spatial baseline threshold. If a sharp decline in matching image data occurs, the spatial baseline can also be determined. The specific determination method is the same as that for determining the temporal baseline, and will not be elaborated further.

[0047] In some embodiments, the step of grouping all the registered SLC images based on a grouping threshold to obtain a subset of SLC images includes: All the registered SLC images are sorted according to the acquisition time of the corresponding SAR satellite images to obtain an image sequence; Using the first SLC image in the image sequence as the starting image of the first SLC image subset, and obtaining the next SLC image of the starting image as the target SLC image, it is determined whether the baseline between the target SLC image and the starting image is greater than a grouping threshold; wherein, when the grouping threshold only includes a temporal baseline threshold, the baseline includes a temporal baseline; when the grouping threshold only includes a spatial baseline threshold, the baseline includes a spatial baseline; when the grouping threshold includes both a temporal baseline threshold and a spatial baseline threshold, the baseline includes both a temporal baseline and a spatial baseline. If the value is not greater than the grouping threshold, the SLC image of the next scene is added to the current SLC image subset, and the SLC image of the next scene is acquired and replaced with the target SLC image. The step of determining whether the baseline between the target SLC image and the starting image is greater than the grouping threshold is executed again. If the value is greater than the grouping threshold, the SLC image of the next scene is used as the starting image of the next SLC image subset; and the process of obtaining the SLC image of the next scene after the starting image is repeated as the target SLC image. It is then determined whether the baseline between the target SLC image and the starting image is greater than the grouping threshold, until all SLC images in the image sequence are divided into the corresponding SLC image subsets.

[0048] For example, the image sequence includes 7 SLC images ordered according to the capture time of the corresponding SAR satellite images, namely SLC image A to SLC image G.

[0049] SLC image A is the first SLC image and serves as the starting image for the first SLC image subset a. The next SLC image, SLC image B, is then acquired as the target SLC image. The baseline between SLC image B and SLC image A is determined to be greater than a grouping threshold. It should be noted that if the grouping threshold only includes a temporal baseline threshold, then the baseline between SLC image B and SLC image A is the temporal baseline. If the grouping threshold only includes a spatial baseline threshold, then the baseline between SLC image B and SLC image A is the spatial baseline. If the grouping threshold includes both a temporal and spatial baseline threshold, then the baseline between SLC image B and SLC image A includes both temporal and spatial baselines.

[0050] If the baseline between SLC image B and SLC image A is not greater than the grouping threshold, then SLC image B is added to the current SLC image subset, i.e., the first SLC image subset a. Then, the next SLC image C is acquired and SLC image C is used as the target SLC image. It is then determined whether the baseline between SLC image C and the starting image A is greater than the grouping threshold. If the baseline between SLC image C and SLC image A is greater than the grouping threshold, then SLC image C is used as the starting image of the second SLC image subset b; the next scene of SLC image C, SLC image D, is obtained as the target image, and it is determined whether the baseline between SLC image D and SLC image C is greater than the grouping threshold. If the baseline between SLC image D and SLC image C is not greater than the grouping threshold, then SLC image D is added to the current SLC image subset, i.e., the second SLC image subset b. Then, the next SLC image E is acquired and SLC image E is used as the target SLC image. It is then determined whether the baseline between SLC image E and the starting image C is greater than the grouping threshold. If the baseline between SLC image E and SLC image C is not greater than the grouping threshold, then SLC image E is added to the second SLC image subset b. Then, the next SLC image F is acquired and SLC image F is used as the target SLC image. It is then determined whether the baseline between SLC image F and the starting image C is greater than the grouping threshold. If the baseline between SLC image F and SLC image C is greater than the grouping threshold, then SLC image F is used as the starting image of the third SLC image subset c; the next scene of SLC image F is obtained as the target image, and it is determined whether the baseline between SLC image G and SLC image F is greater than the grouping threshold. If the baseline between SLC image G and SLC image F is not greater than the grouping threshold, then SLC image G is added to the third SLC image subset c. Since the last SLC image in the image sequence is SLC image G, all SLC images are now assigned to their respective SLC image subsets. Ultimately, the first SLC image subset includes SLC images A and B; the second SLC image subset includes SLC images C, D, and E; and the third SLC image subset includes SLC images F and G.

[0051] S400, Determine the short-time series deformation of the SLC image subset; In this embodiment, since the baselines between SLC images in the SLC image subset are not greater than the grouping threshold, the temporal and spatial changes of the SLC images in the SLC image subset are small, and the deformation is not too large. This avoids the situation where pixel offset tracking is performed directly on the time series corresponding to all ungrouped SLC images, and the pixel offset becomes larger and larger due to the accumulation of deformation, which may eventually lead to pixel mismatch.

[0052] In some embodiments, the step of determining the short-time series deformation of the SLC image subset includes: Select a master image from the subset of SLC images; In some embodiments, the step of selecting a master image from the subset of SLC images includes: selecting SLC images from the subset of SLC images that meet the second preset time baseline requirement, the second preset spatial baseline requirement, and the second preset Doppler center frequency requirement as the master image.

[0053] In this embodiment, the second preset time baseline requirement, the second preset spatial baseline requirement, and the second preset Doppler center frequency requirement are similar to the first preset time baseline requirement, the first preset spatial baseline requirement, and the first preset Doppler center frequency requirement mentioned above. The first preset time baseline requirement, the first preset spatial baseline requirement, and the first preset Doppler center frequency requirement mentioned above are intended to ensure that the super main image is as centered as possible in terms of time, orbit, and Doppler center frequency. The second preset time baseline requirement, the second preset spatial baseline requirement, and the second preset Doppler center frequency requirement here are also intended to ensure that the main image is as centered as possible in terms of time, orbit, and Doppler center frequency.

[0054] Determine the pixel offset observation values ​​of all SLC images in the SLC image subset and the preset pixel points of the main image; combine all the observation values ​​into an observation value set; In some embodiments, the preset pixels can be all pixels in the SLC image.

[0055] In one example, the number of SLC image subsets is n. L This represents the set of observed pixel offsets of all SLC images in the SLC image subset, respectively, relative to a preset pixel point of the main image.

[0056] ; in, The observed pixel offset of the preset pixel point between the first SLC image and the main image in the SLC image subset; The observed pixel offset values ​​for preset pixel points in the second SLC image and the main image within the SLC image subset; and so on. This refers to the observed pixel offset of a preset pixel point between the nth SLC image and the main image within the SLC image subset. The SLC image subset includes n SLC images.

[0057] Substituting the set of observations into the observation equation and solving it through adjustment, the short-time series deformation of the SLC image subset is obtained; the short-time series deformation of the SLC image subset includes the offsets of the 2nd to nth SLC images in the SLC image subset from the 1st SLC image, and the SLC image subset includes n SLC images.

[0058] In this embodiment of the application, the observation equation is: ; in, A represents the set of observed pixel offsets of all SLC images in the SLC image subset and the preset pixel points of the main image; A is the first coefficient matrix, consisting of 0, -1, and 1; F is the short-time series deformation of the SLC image subset; and V is the error term.

[0059] In this embodiment, the adjustment calculation employs the least squares principle. The short-time series deformation of the SLC image subset obtained according to the least squares principle can be understood as the subset time-series inversion result under the constraints of temporal and spatial baseline thresholds. The adjustment calculation uses the following formula: ; Where F represents the short-time series deformation of the SLC image subset; A is the first coefficient matrix; and P1 represents the weights of different observations in the observation set.

[0060] In one example, for a subset of SLC images comprising n SLC images, starting with the first SLC image, the short-time series deformation F of the obtained SLC image subset is: ; Wherein, F represents the short-time series deformation of the SLC image subset; This indicates the offset between the SLC image of the second scene and the SLC image of the first scene; This indicates the offset between the SLC image of the third scene and the SLC image of the first scene; and so on. This represents the offset between the SLC image of the nth scene and the SLC image of the first scene.

[0061] S500. Based on the short time series deformation of the SLC image subset, use robust estimation to determine the final offset of the main image in the adjacent SLC image subset; In this embodiment of the application, by introducing robust estimation, even if there are gross errors mixed in the short time series deformations of different SLC image subsets, their influence can be minimized by iterative weight reduction. This ensures that the accuracy of the large gradient time series deformation of the land surface obtained after connecting the short time series deformations of different SLC image subsets is high.

[0062] In some embodiments, the step of determining the final offset of the master image in adjacent SLC image subsets using robust estimation based on the short-time-series deformation of the SLC image subsets includes: Based on the short-time series deformation of all SLC image subsets, an offset set of the main images in adjacent SLC image subsets is determined. The offset set includes a direct offset and an indirect offset set. The direct offset is the offset of the main image in an adjacent SLC image subset. The indirect offset set includes several indirect offsets, the number of which is the number of SLC images located between two main images. The indirect offset is the difference between the offset of each main image in an adjacent SLC image subset and the offset of an SLC image located between two main images. In this embodiment, each SLC image subset can determine the corresponding short time series deformation, and based on all SLC image subsets, the offset set of the main image in adjacent SLC image subsets is determined.

[0063] In some embodiments, the offset set can be represented as: ; Where G is the set of offsets; where, The direct offset represents the offset between the main image m1 in the SLC image subset and the main image m2 in the adjacent SLC image subset; the remaining offsets in the offset set can form the indirect offset set, where, , representing the indirect offset, indicates the difference between the offsets of the main image m1 in the SLC image subset and the main image m2 in the adjacent SLC image subset, and the offset of the SLC image s1 located between the main image m1 and the adjacent main image m2 in the SLC image subset; and so on. , is the indirect offset, representing the difference between the offset of the main image m1 in the SLC image subset and the main image m2 in the adjacent SLC image subset, and the offset of the SLC image si-1 located between the main image m1 in the SLC image subset and the main image m2 in the adjacent SLC image subset; wherein the number of SLC images located between the main image m1 in the SLC image subset and the main image m2 in the adjacent SLC image subset is i-1.

[0064] For example, a subset of SLC images includes SLC images A, B, C, D, and E, where SLC image B is the principal image. The adjacent subset of SLC images includes SLC images D, E, F, G, and H, where SLC image G is the principal image. In this case, the SLC images located between two adjacent principal images are identified as SLC images C, D, E, and F. The corresponding number of SLC images located between two adjacent principal images is 4.

[0065] Using the offset set, robust estimation is performed to obtain the final offset of the main image in the adjacent SLC image subset.

[0066] In this embodiment of the application, a robust estimation algorithm is used to iteratively estimate the optimized offset (i.e., the final offset) between adjacent SLC image subsets.

[0067] Specifically, for a single iteration of robust estimation, we have: ; Where K represents the difference between the offset set and the approximate offset of the main image in the adjacent SLC image subset obtained by robust estimation; G is the offset set; x0 represents the approximate offset of the main image in the adjacent SLC image subset obtained by robust estimation, which is determined iteratively by the formula: new x0 = previous x0 + x. Where x is the correction value for x0, which can be calculated using the following formula: ; Where B represents the second coefficient matrix; P2 is the weight matrix composed of the weight values ​​of each element (i.e. each offset) in G; and K represents the difference between the offset set and the approximate offset of the main image in the adjacent SLC image subset obtained by robust estimation.

[0068] Where u is the residual; B represents the second coefficient matrix; x is the correction value of x0; and K represents the difference between the offset set and the approximate offset of the main image in the adjacent SLC image subset obtained by robust estimation.

[0069] ; in, denoted as unit weight error; u is the residual; P2 is the weight matrix composed of the weight values ​​of each element (i.e., each offset) in G; i-1 is the number of SLC images located between the main image m1 in the SLC image subset and the main image m2 in the adjacent SLC image subset.

[0070] ; Among them, Q u Let be the covariance matrix of the residual u, and Q be the covariance matrix of G. For the q-th element in set G, its standardized residual can be expressed as: ; in, For standardized residuals; For residuals; Q represents the unit weighted error.u Let be the covariance matrix of the residual u.

[0071] For the j-th iteration: ; in, This is the noise amplification factor; For the first j In the nth iteration q The standardized residuals of each element; c is a constant (usually 2.0–2.5). When the standardized residuals exceed this constant, their weights are reduced by a noise amplification factor. At this point, the second weight matrix... P The diagonal elements of 2 are: ; in, Weight matrix P The diagonal element of 2; This is the noise amplification factor; The iteration is complete when the standardized residuals obtained from two consecutive calculations remain unchanged. At this point, the final offset after robust estimation can be calculated based on the elements and their corresponding weights.

[0072] S600. Based on the final offset of the main image in all adjacent SLC image subsets, the short time series offsets of all SLC image subsets are connected to obtain the large gradient temporal deformation of the land surface.

[0073] For example, there are five SLC image subsets: SLC image subset a, SLC image subset b, SLC image subset c, SLC image subset d, and SLC image subset e. Adjacent SLC image subsets include SLC image subset a and SLC image subset b, SLC image subset b and SLC image subset c, SLC image subset c and SLC image subset d, and SLC image subset d and SLC image subset e. Therefore, the final offsets of the main images of SLC image subsets a and b, SLC image subset b and SLC image subset c, SLC image subset c and SLC image subset d, and SLC image subset d and SLC image subset e can be obtained.

[0074] In this embodiment of the application, based on the final offset obtained after all robust estimations, all short-time series deformations can be connected to reconstruct the long-time series deformation results, namely, the large-gradient temporal deformation of the land surface.

[0075] In this embodiment, when a new SAR satellite image is subsequently acquired, if the baseline between it and the starting image in the last SLC image subset is not greater than the grouping threshold, it can be added to the last SLC image subset; if the baseline between it and the starting image in the last SLC image subset is greater than the grouping threshold, a new SLC image subset can be formed, and the short-time series deformation of the new SLC image subset is determined. Based on the short-time series deformation of the new SLC image subset, robust estimation is used to determine the final offset of the main image in the new adjacent SLC image subset (that is, the final offset between the new SLC image subset and the main image in the original last SLC image subset). Based on the final offset of the main image in the new adjacent SLC image subset and all the original final offsets, the short-time series offset of the new SLC image subset and the original large-gradient temporal deformation of the land surface are connected to obtain the new large-gradient temporal deformation of the land surface.

[0076] In other words, in this embodiment, acquiring new SAR satellite imagery does not require modifying the existing surface gradient temporal deformation. Instead, it only requires concatenating the short-time series offsets of the new SLC image subset using the newly obtained final offset, based on the existing surface gradient temporal deformation. This eliminates the need to recalculate historical SAR satellite imagery, improving computational efficiency. It solves the problem in related technologies where introducing new data requires recalculating historical data, resulting in low computational efficiency and difficulty in meeting timeliness requirements. The method in this embodiment can meet the needs of large-scale, high-frequency dynamic monitoring tasks.

[0077] This application presents a method for determining large-gradient temporal deformation of the Earth's surface, which can acquire long-term deformation data of large-gradient deformation regions, thereby achieving high-resolution deformation monitoring of the Earth's surface. The advantage of this method is that it can fully utilize the intensity information in SAR imagery to calculate the time-series offset with low time and storage costs, thus reconstructing the evolution of the large-gradient deformation field over time. Simultaneously, it can accurately record the spatial location of the monitored target's displacement trajectory within the time series, solving the problems of mismatch and offset estimation of target points in traditional time-series methods for long-term monitoring of large-gradient deformation. This work is of great significance for obtaining high-precision temporal deformation data of the Earth's surface through pixel offset tracking technology.

[0078] The following specific embodiments demonstrate the computational performance of the methods described in this application.

[0079] To verify the temporal performance of the proposed method, 54 Cosmo-SkyMed down-orbit data images from June 1, 2014 to December 18, 2016 were acquired for the experiment. Cosmo-SkyMed is an X-band SAR satellite system. This system consists of multiple satellites and possesses high resolution, high temporal resolution, and all-weather, day-and-night observation capabilities, making it suitable for fields such as surface deformation monitoring, disaster emergency response, and environmental protection. Its high-precision spatial resolution (up to 1 meter) and rapid revisit capability (16 days) make it important in SAR research and remote sensing applications. The SAR image parameters used in this study are shown in Table 1.

[0080] Table 1 SAR Image Parameter Information

[0081] Subsequently, simulation experiments and real-data experiments were conducted based on the acquired COSMO-SkyMed data. In the simulation experiments, the deformation field was obtained through simulation. In the real-data experiments, the surface deformation field, temporal baseline, and spatial baseline were all based on real data.

[0082] Simulated data comparison experiment: First, an elliptical deformation field was simulated, with a deformation of 0 pixels at the top edge and 55 pixels at the bottom edge, exhibiting uniform spatial variation. Then, a time-series deformation field was generated based on the temporal variation curve. The spatial distribution and temporal variation of the deformation field are shown below. Figure 3 As shown. Figure 3 (a) is a spatial distribution diagram of a time-series deformation field used in a simulation experiment; Figure 3 (b) is a time-varying curve of a time-series deformation field used in a simulation experiment, with the vertical axis representing the ratio between the displacement at the current moment and the final moment.

[0083] For the generated temporal deformation field, pairwise offset data were obtained, and then temporal inversion was performed using the PO-SBAS method and the method proposed in this paper (hereinafter referred to as the RS-POT method). The original deformation field, the temporal inversion results (i.e., the large gradient temporal deformation of the land surface), and the estimation error are shown below. Figure 4 As shown. Figure 4 (a1)–(a6) are schematic diagrams of the simulated cumulative deformation at different observation times; Figure 4 (b1)–(b6) are schematic diagrams of PO-SBAS inversion results at different observation times; Figure 4 (c1)–(c6) are schematic diagrams of PO-SBAS estimation errors at different observation times; Figure 4 (d1)–(d6) are schematic diagrams of RS-POT inversion results at different observation times; Figure 4Figures (e1)–(e6) illustrate the RS-POT estimation errors at different observation phases: phases 5, 10, 15, 20, 25, and 30. The time-series inversion results show that the PO-SBAS method exhibits a larger spatial deformation range, while showing ambiguity at the lower edge of the deformation field. The estimation error diagrams show that the PO-SBAS results significantly underestimate the deformation in the central deformation region, while exhibiting severe overestimation outside the lower boundary. Conversely, the RS-POT method shows smaller estimation errors, with localized errors only appearing at the lower edge of the deformation region.

[0084] To further analyze the spatial distribution of the error, Figure 5 Draw along Figure 4 Time series distortion estimated by mid-profile Figure 5 (a) is a schematic diagram simulating cross-sectional deformation; Figure 5 (b) is a schematic diagram of a PO-SBAS inversion result; Figure 5 (c) is a schematic diagram of an RS-POT inversion result. Significant deformation underestimation exists within the main deformation zone along the PO-SBAS profile, while overestimation occurs outside the lower boundary of the deformation zone. In contrast, the proposed RS-POT method achieves robustness and high accuracy in deformation estimation, with only minor errors outside the deformation zone.

[0085] To quantify the deformation estimation accuracy of RS-POT and PO-SBAS, Table 2 summarizes the statistical data on the number and proportion of underestimated, overestimated, and correctly estimated pixels. Pixels with an estimation error of less than 0.5 pixels are considered correctly estimated. The PO-SBAS method has a correct estimation rate of 72.60%, an underestimation rate of approximately 7.38%, and an overestimation rate of 20.02%. The RS-POT method achieves a correct estimation rate of 96.70%, with an underestimation rate of only 2.00% and an overestimation rate of only 1.30%. Compared to PO-SBAS, RS-POT improves accuracy by 24.10%. The root mean square error (RMSE) of PO-SBAS is 5.35 pixels, while RS-POT significantly reduces it to 1.95 pixels, a 63.55% reduction in error compared to PO-SBAS.

[0086] Table 2. Statistics on the number and proportion of pixels that are underestimated, overestimated, and correctly estimated.

[0087] Figure 6 The time-series variation curves of the monitoring points in the simulation experiment are shown. Figure 6 (a) is a schematic diagram showing the distribution of points P1, P2, and P3; Figure 6(b)–(d) are schematic diagrams of the time-series deformation curves for P1, P2, and P3, respectively. The results show that for P1 and P2, the PO-SBAS method underestimates the deformation, while the proposed RS-POT method achieves accurate estimation. For the undeformed P3 point, PO-SBAS introduces erroneous deformation. The RS-POT method achieves relatively accurate estimation at all three monitoring points. From the time-varying characteristics of the monitoring points, the PO-SBAS method exhibits only slight errors in the initial stage, but as the accumulated deformation increases over time, the estimation error significantly expands due to the spatial heterogeneity of the deformation field and the exacerbation of pixel matching failure.

[0088] To evaluate the applicability of the RS-POT method in different scenarios, three representative landslide models were constructed for simulation: traction landslide (small deformation at the rear edge, large deformation at the front edge), push-type landslide (large deformation at the rear edge, small deformation at the front edge), and homogeneous landslide. Figure 7 The diagram shows applicability in three simulated scenarios, using PO. SBAS and RS Two methods, POT, were used to simulate deformation field inversion and error analysis for three types of landslides. The columns, from left to right, correspond to the three landslide types: traction landslides, push-moving landslides, and homogeneous landslides. The rows are: (a)–(c) simulated deformation field; (d)–(f) POT. SBAS method for estimating error; (g)–(i)RS The POT method estimates the error. Figure 7 (a)–(c) are schematic diagrams of the simulated deformation fields corresponding to traction landslides, shovel-type landslides and homogeneous landslides, respectively. Figure 7 (d)–(f) represent the PO values ​​corresponding to traction landslides, kinetic landslides, and homogeneous landslides, respectively. A schematic diagram illustrating the error estimation using the SBAS method; Figure 7 (g)–(i) represent RS corresponding to traction landslides, kinetic landslides, and homogeneous landslides, respectively. A schematic diagram illustrating the estimation error of the POT method. RS-POT consistently produces small estimation errors in all three scenarios, demonstrating its superior performance in temporal deformation reconstruction.

[0089] Example: This embodiment focuses on a landslide. This landslide is a giant bedding-plane landslide located on the southeast side of a mine pit, adjacent to the pit's boundary, with distinct geomorphological features. The landslide crest rises approximately 200 meters above the mine pit bottom, exhibiting a stepped distribution with steep slopes and local angles exceeding 40°. The landslide was primarily formed by mining disturbance. The overlying strata consist of loose Quaternary sediments, as well as Mesozoic sandstone and shale, exhibiting relatively fractured lithology, low shear strength, and a loose structure. Long-term mining activities have weakened the rock mass's stability, and combined with rainfall and groundwater seepage, have exacerbated the landslide's deformation and instability risk. The landslide boundary is relatively clear, with significant tension cracks at the upper boundary, some reaching tens of centimeters in width, indicating that the landslide is expanding outwards at a significant rate. Monitoring data from 2012 to 2019 shows that the landslide's maximum horizontal displacement reached 96.01 meters, and its maximum subsidence reached 56.65 meters, with a significant increase in deformation rate during periods of heavy rainfall.

[0090] SAR images are processed into SLC images. First, a super master image is selected, and the SLC images are registered based on the super master image. Under the constraint of temporal and spatial baselines, a larger step size is used for pixel offset tracking calculation, and the number of matching pixels is counted. The impact of temporal and spatial baselines on the number of matching pixels is analyzed based on the statistical results to determine the grouping threshold. Subsequently, all the SLC images are grouped according to the temporal and spatial baseline thresholds. The resulting temporal and spatial baseline maps (horizontal axis: temporal baseline, vertical axis: spatial baseline) are shown below. Figure 8 As shown, where Figure 8 (a) is a distribution map of the PO-SBAS baseline combination; Figure 8 (b) is a distribution map of the RS-POT baseline combination.

[0091] Figure 9 The cumulative deformation of the landslide at three different time points is estimated using two methods, PO-SBAS and RS-POT, and the differences between the two methods are shown. Figure 9 Each column corresponds to a different imaging time, namely December 2014, December 2015, and December 2016. The rows from top to bottom are: (a1)–(a3) schematic diagrams of azimuth deformation retrieved from PO-SBAS at different times; Figure 9 (b1)–(b3) are schematic diagrams of RS-POT inversion azimuth deformation at different times; Figure 9 (c1)–(c3) are schematic diagrams of the azimuth deformation difference (RS-POT minus PO-SBAS) of the two methods at different times; Figure 9 (d1)–(d3) are schematic diagrams of LOS deformation inverted by PO-SBAS at different times; Figure 9(e1)–(e3) are schematic diagrams of LOS-directed deformation retrieved by RS-POT at different times; Figure 9 (f1)–(f3) are schematic diagrams of the LOS deformation difference (RS-POT minus PO-SBAS) for the two methods at different times. All deformation results are cumulative deformations based on the initial observation time. In the azimuth deformation map: positive values ​​(blue) represent surface displacement in the same direction as the satellite flight, and negative values ​​(red) represent surface displacement in the opposite direction to the satellite flight. In the line-of-sight deformation map: positive values ​​(blue) represent surface displacement away from the satellite, and negative values ​​(red) represent surface displacement towards the satellite. Overall, the deformation estimated by PO-SBAS and RS-POT shows high consistency. Spatially, the deformation is larger in the western region of the landslide and smaller in the eastern region. In areas with larger deformation, the difference between PO-SBAS and RS-POT is usually more significant, indicating that the increase in deformation gradient leads to an increased risk of pixel matching in PO-SBAS. The maximum difference between the two deformation results is approximately 5 meters (located in the high-gradient deformation zone). In regions with high gradients and large deformations, the proposed RS-POT method achieves a more significant accuracy improvement compared to PO-SBAS.

[0092] To evaluate the deformation estimation accuracy of RS-POT and PO-SBAS, this study cross-compares deformation time series data with GNSS measurements. Figure 10 (a), (c), (e) and (g) are schematic diagrams of the azimuth deformation of points G1-G4; Figure 10 (b), (d), (f), and (h) are schematic diagrams of the LOS-direction deformation of points G1-G4. Figure 10 The time-series deformation data at four points (G1, G2, G3, G4) show that the RS-POT (red curve) and PO-SBAS (blue curve) results generally exhibit a deformation trend consistent with GNSS measurements (green curve). However, in the later stages of deformation (after the summer of 2015), PO-SBAS tends to significantly underestimate the deformation magnitude. In contrast, RS-POT remains consistent with GNSS measurements throughout the observation period, without any significant underestimation.

[0093] The temporal deformation evolution at the monitoring points indicates that the landslide was reactivated by heavy rainfall in August 2013, thus initiating a large-scale deformation phase. From August 2013 to December 2014, the cumulative deformation in the azimuth direction was approximately 40-60 meters, and the cumulative deformation in the LOS direction was approximately 30 meters. The deformation rate temporarily slowed in the first half of 2015, followed by a slight acceleration during the summer (rainy season). From June 2015 to December 2016, the cumulative deformation in both the azimuth and LOS directions was approximately 20 meters, indicating that the deformation intensity was weaker than in the previous two years. Similar to the experimental results from the simulation data, after August 2015, the PO-SBAS results showed a significant underestimation as the cumulative deformation intensified—increased pixel mismatches led to a continuous decline in measurement accuracy.

[0094] To further quantify the estimation accuracy, the root mean square error (RMSE) of the difference between the time-series deformation estimated by the two methods and the GNSS data was calculated, and the results are shown in Table 3. Compared with PO-SBAS, RS-POT achieves a significant improvement in measurement accuracy. The average RMSE of PO-SBAS in the azimuth direction is 1.73 meters, while that of RS-POT is only 0.81 meters, representing an improvement in estimation accuracy of 51.51%. Similarly, RS-POT improves accuracy by 43.61% in the range direction. Compared with PO-SBAS, the RS-POT method is more adaptable to deformation areas with high spatial heterogeneity and large displacement, reducing the risk of mismatch and improving measurement accuracy.

[0095] Table 3. Root mean square error of the time series distortion and the difference between GNSS data estimated by the two methods.

[0096] This application provides a method for determining the temporal deformation of a large surface gradient. The method involves acquiring a set of SAR satellite images; processing the SAR satellite images into SLC images and selecting a super master image from all the SLC images; registering all the SLC images based on the super master image; determining a grouping threshold for all the registered SLC images based on the grouping threshold; and grouping all the registered SLC images based on the grouping threshold to obtain SLC image subsets. This allows SLC images with significant differences to be grouped into different SLC image subsets, and subsequent determination of short-time series deformation for individual SLC image subsets solves the problem of mismatched pixel targets caused by significant changes in pixel position in the PO-SBAS method. Furthermore, by introducing robust estimation, even if there are gross errors mixed in the short-time series deformation of different SLC image subsets, their influence can be minimized through iterative weighting. This ensures high accuracy in obtaining the temporal deformation of a large surface gradient after connecting the short-time series deformations of different SLC image subsets.

[0097] This application also provides a device for determining the temporal deformation of a large surface gradient, comprising: The acquisition unit is used to acquire a SAR satellite image set, which includes multiple SAR satellite images taken by satellites in the same repeating orbit over the research area on the Earth's surface and ordered by the time of capture. A registration unit is used to process SAR satellite images into SLC images; select a super master image from all the SLC images; and register all the SLC images based on the super master image. A grouping unit is used to determine a grouping threshold for all registered SLC images based on all registered SLC images; and to group all registered SLC images based on the grouping threshold to obtain a subset of SLC images; the grouping threshold includes a temporal baseline threshold and / or a spatial baseline threshold. The first determining unit is used to determine the short-time series deformation of the SLC image subset; The second determining unit is used to determine the final offset of the main image in adjacent SLC image subsets based on the short time series deformation of the SLC image subset using robust estimation. The connection unit is used to connect the short time series offsets of all SLC image subsets based on the final offset of the main image in all adjacent SLC image subsets to obtain the large gradient temporal deformation of the land surface.

[0098] This application also provides a computer device, including a processor and a memory, wherein the memory stores a program or instructions that run on the processor, and the program or instructions, when executed by the processor, implement the steps of the method for determining large gradient temporal deformation of the earth's surface as described above.

[0099] This application embodiment also provides a readable storage medium storing a program or instructions thereon, which, when executed by a processor, implements the steps of the method for determining the large gradient temporal deformation of the Earth's surface.

[0100] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0101] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the scope of protection of this application, and all of these forms are within the protection scope of this application.

Claims

1. A method for determining temporal deformation of large surface gradients, characterized in that, include: Acquire a set of SAR satellite images, which includes multiple SAR satellite images of the study area taken by satellites in the same repeating orbit and ordered by the time of capture. SAR satellite imagery is processed into SLC imagery; a super master imagery is selected from all the SLC imagery; and all the SLC imagery are registered based on the super master imagery. Based on all registered SLC images, a grouping threshold for all SLC images is determined; based on the grouping threshold, all registered SLC images are grouped to obtain a subset of SLC images; the grouping threshold includes a temporal baseline threshold and / or a spatial baseline threshold. Determine the short-time series deformation of the SLC image subset; Based on the short-time series deformation of the SLC image subset, robust estimation is used to determine the final offset of the main image in adjacent SLC image subsets; Based on the final offset of the main image in all adjacent SLC image subsets, the short time series offsets of all SLC image subsets are concatenated to obtain the large gradient temporal deformation of the land surface.

2. The method according to claim 1, characterized in that, The step of determining the grouping threshold for all SLC images based on all registered SLC images includes: The search window is determined based on the deformation characteristics of the surface study area; Based on the search window, pixel offset tracking is performed on the registered SLC image using a preset step size to obtain tracking results; wherein, the tracking results include determining the influence of the temporal baseline on the number of matching pixels within a preset spatial baseline range, and determining the influence of the spatial baseline on the number of matching pixels within a preset temporal baseline range. Based on the tracking results, a grouping threshold is determined; wherein the grouping threshold includes a temporal baseline threshold and / or a spatial baseline threshold.

3. The method according to claim 2, characterized in that, The step of grouping all the registered SLC images based on a grouping threshold to obtain a subset of SLC images includes: All the registered SLC images are sorted according to the acquisition time of the corresponding SAR satellite images to obtain an image sequence; Using the first SLC image in the image sequence as the starting image of the first SLC image subset, and obtaining the next SLC image of the starting image as the target SLC image, it is determined whether the baseline between the target SLC image and the starting image is greater than a grouping threshold; wherein, when the grouping threshold only includes a temporal baseline threshold, the baseline includes a temporal baseline; when the grouping threshold only includes a spatial baseline threshold, the baseline includes a spatial baseline; when the grouping threshold includes both a temporal baseline threshold and a spatial baseline threshold, the baseline includes both a temporal baseline and a spatial baseline. If the value is not greater than the grouping threshold, the SLC image of the next scene is added to the current SLC image subset, and the SLC image of the next scene is acquired and replaced with the target SLC image. The step of determining whether the baseline between the target SLC image and the starting image is greater than the grouping threshold is executed again. If the value is greater than the grouping threshold, the SLC image of the next scene is used as the starting image of the next SLC image subset; and the process of obtaining the SLC image of the next scene after the starting image is repeated as the target SLC image. It is then determined whether the baseline between the target SLC image and the starting image is greater than the grouping threshold, until all SLC images in the image sequence are divided into the corresponding SLC image subsets.

4. The method according to claim 1, characterized in that, The step of determining the short-time series deformation of the SLC image subset includes: Select a master image from the subset of SLC images; Determine the pixel offset observation values ​​of all SLC images in the SLC image subset and the preset pixel points of the main image; combine all the observation values ​​into an observation value set; Substituting the set of observations into the observation equation and solving it through adjustment, the short-time series deformation of the SLC image subset is obtained; the short-time series deformation of the SLC image subset includes the offsets of the 2nd to nth SLC images in the SLC image subset from the 1st SLC image, and the SLC image subset includes n SLC images.

5. The method according to claim 4, characterized in that, The step of determining the final offset of the main image in adjacent SLC image subsets using robust estimation based on the short time series deformation of the SLC image subsets includes: Based on the short-time series deformation of all SLC image subsets, an offset set of the main images in adjacent SLC image subsets is determined. The offset set includes a direct offset and an indirect offset set. The direct offset is the offset of the main image in an adjacent SLC image subset. The indirect offset set includes several indirect offsets, the number of which is the number of SLC images located between two main images. The indirect offset is the difference between the offset of each main image in an adjacent SLC image subset and the offset of an SLC image located between two main images. Using the offset set, robust estimation is performed to obtain the final offset of the main image in the adjacent SLC image subset.

6. The method according to claim 1, characterized in that, The step of selecting a super master image from all the SLC images includes: SLC images that meet the first preset time baseline requirements, the first preset spatial baseline requirements, and the first preset Doppler center frequency requirements are selected from all the SLC images and used as super master images.

7. The method according to claim 1, characterized in that, The step of registering all the SLC images based on the super master image includes: Obtain DEM data of the study area; Based on the DEM data of the super master image and SLC images, all the SLC images are registered based on the super master image.

8. A device for determining temporal deformation of large surface gradients, characterized in that, include: The acquisition unit is used to acquire a SAR satellite image set, which includes multiple SAR satellite images taken by satellites in the same repeating orbit over the research area on the Earth's surface and ordered by the time of capture. A registration unit is used to process SAR satellite images into SLC images; select a super master image from all the SLC images; and register all the SLC images based on the super master image. A grouping unit is used to determine a grouping threshold for all registered SLC images based on all registered SLC images; and to group all registered SLC images based on the grouping threshold to obtain a subset of SLC images; the grouping threshold includes a temporal baseline threshold and / or a spatial baseline threshold. The first determining unit is used to determine the short-time series deformation of the SLC image subset; The second determining unit is used to determine the final offset of the main image in adjacent SLC image subsets based on the short time series deformation of the SLC image subset using robust estimation. The connection unit is used to connect the short time series offsets of all SLC image subsets based on the final offset of the main image in all adjacent SLC image subsets to obtain the large gradient temporal deformation of the land surface.

9. A computer device, characterized in that, It includes a processor and a memory, the memory storing a program or instructions that run on the processor, the program or instructions being executed by the processor to implement the steps of the method for determining the temporal deformation of large gradients of the earth's surface as described in any one of claims 1 to 7.

10. A readable storage medium having a program or instructions stored thereon, characterized in that, When the program or instructions are executed by the processor, they implement the steps of the method for determining the temporal deformation of large gradients of the earth's surface as described in any one of claims 1 to 7.

Images