Distributed InSAR satellite image speckle suppression method and device

Abstract

This invention discloses a method and apparatus for suppressing speckle in distributed InSAR satellite images, relating to the field of remote sensing image processing technology. The method includes: registering all images of a target area acquired by distributed InSAR satellites; determining the position of each pixel in the registered images; constructing an original complex scattering vector corresponding to the pixel position based on the complex scattering value corresponding to the pixel position; constructing a coherence matrix corresponding to the pixel position based on the original complex scattering vector; performing a decorrelation projection transformation on the original complex scattering vector based on the coherence matrix to obtain a target complex scattering vector corresponding to each pixel position; and performing weighted fusion of the target complex scattering vector corresponding to each pixel position based on the fusion weight of each receiving satellite to obtain a speckle-suppressed target image for the target area. This invention solves the technical problem of ineffective speckle suppression in related technologies.

Description

Technical Field

[0001] This application relates to the field of remote sensing image processing technology, and more specifically, to a distributed InSAR satellite image speckle suppression method and apparatus. Background Technology

[0002] Speckle noise is an inherent multiplicative noise in SAR (Synthetic Aperture Radar) imaging systems. It originates from the coherent superposition of radar echoes, leading to grainy image texture and reduced radiometric resolution, thus severely impacting the accuracy of SAR image interpretation, target recognition, and interferometric measurements. Therefore, speckle suppression is a critical issue in SAR data processing.

[0003] Currently, speckle noise suppression methods mainly include spatial domain filtering, transform domain filtering, multi-temporal averaging filtering, and deep learning methods. Spatial domain filtering smooths noise by designing specific filters and performing convolution operations on SAR images. Transform domain filtering transforms SAR images to the frequency or wavelet domain, suppresses noise components based on the distribution characteristics of speckle noise in the transform domain, and then inversely transforms back to the spatial domain to obtain the despecced result. Multi-temporal averaging filtering statistically averages images of the same area acquired by satellite at different time phases, utilizing the statistical independence of speckle noise to suppress it. Deep learning methods train neural networks using constructed sample datasets to achieve speckle noise suppression.

[0004] However, the aforementioned SAR speckle suppression methods still have significant shortcomings in distributed InSAR (Interferometric Synthetic Aperture Radar) scenarios. While spatial filtering methods can effectively suppress speckle noise, they reduce image resolution and weaken image details. Transform domain filtering methods suffer from poor speckle suppression due to the overlap between the signal and speckle noise in the transform domain, leading to loss of image texture details. Multi-temporal averaging methods require accumulating satellite imagery data from multiple time points, resulting in long acquisition cycles and susceptibility to surface changes, making them unsuitable for real-time or near-real-time applications. Deep learning methods rely on large amounts of training data, and their results are highly dependent on the distribution of training samples, exhibiting insufficient stability and interpretability.

[0005] Distributed InSAR satellite systems, through a single-transmitter-multiple-receiver architecture, can simultaneously acquire multiple complex SAR images of the same area under single-orbit observation conditions, providing a new approach to speckle noise suppression. However, due to the limited decoherence factors among synchronously observed images, speckle noise in each channel is usually highly correlated, making it difficult to achieve ideal noise suppression results through simple averaging.

[0006] Therefore, how to fully utilize multiple SAR image data acquired by a distributed InSAR system while maintaining the spatial resolution and texture details of SAR images, reduce speckle noise correlation, and achieve effective speckle suppression has become an urgent problem to be solved.

[0007] There is currently no effective solution to the above problems. Summary of the Invention

[0008] This application provides a distributed InSAR satellite image speckle suppression method and apparatus to at least solve the technical problem in related technologies that cannot effectively suppress speckle in images.

[0009] According to one aspect of the embodiments of this application, a distributed InSAR satellite image speckle suppression method is provided, comprising: acquiring multiple images of a target region, wherein the images are acquired by a distributed InSAR satellite, the distributed InSAR satellite comprising: multiple receiving satellites, each receiving satellite being used to capture one image, and each pixel position in the image corresponding to a complex scattering value; registering all images to obtain registered multiple images; determining each pixel position in the registered multiple images, constructing an original complex scattering vector corresponding to the pixel position based on the complex scattering value corresponding to the pixel position, and constructing a coherence matrix corresponding to the pixel position based on the original complex scattering vector; performing a decorrelation projection transformation on the original complex scattering vector based on the coherence matrix to obtain a target complex scattering vector corresponding to each pixel position; and performing weighted fusion of the target complex scattering vector corresponding to each pixel position based on the fusion weight of each receiving satellite to obtain a speckle-suppressed target image of the target region.

[0010] Further, the step of registering all images to obtain multiple registered images includes: acquiring the satellite orbit parameters and imaging geometric relationships of each receiving satellite; using a preset imaging model to perform geometric mapping on all images based on the satellite orbit parameters and imaging geometric relationships of each receiving satellite to obtain the initial pixel correspondence between each image and the target area; performing pixel-level registration on all images based on the initial pixel correspondence to obtain multiple images after initial registration; and performing sub-pixel-level registration on all images after initial registration to obtain multiple registered images.

[0011] Further, the step of performing sub-pixel-level registration on all images after the initial registration to obtain multiple registered images includes: determining a local window; selecting the master image from all images and determining all remaining images as images to be registered; sliding the local window, and for each image to be registered, calculating the correlation coefficient between the master image and the image to be registered under the local window; determining the residual offset at the position indicated by the maximum correlation coefficient, and performing sub-pixel-level displacement on the image to be registered based on the residual offset to obtain the target pixel correspondence between the image to be registered and the master image; and registering each image to be registered with the master image based on the target pixel correspondence to obtain multiple registered images.

[0012] Further, the steps of constructing the original complex scattering vector corresponding to the pixel position based on the complex scattering value corresponding to the pixel position, and constructing the coherence matrix corresponding to the pixel position based on the original complex scattering vector, include: processing the complex scattering value of the pixel position in each image to obtain the original complex scattering vector corresponding to the pixel position; constructing a desired operation window centered on the pixel position; and statistically averaging the desired operation window based on the original complex scattering vector to obtain the coherence matrix corresponding to the pixel position.

[0013] Furthermore, the step of performing a decorrelation projection transformation on the original complex scattering vector based on the coherence matrix to obtain the target complex scattering vector corresponding to each pixel position includes: performing eigenvalue decomposition on the coherence matrix to obtain a unitary matrix and an initial diagonal matrix, wherein each element in the initial diagonal matrix corresponds to an original eigenvalue; performing regularization processing on the initial diagonal matrix to obtain the target diagonal matrix; constructing a transformation matrix based on the target diagonal matrix and the unitary matrix; and performing spatial projection on the original complex scattering vector based on the transformation matrix to obtain the target complex scattering vector.

[0014] Further, the step of regularizing the initial diagonal matrix to obtain the target diagonal matrix includes: determining the coherence coefficient, the scaling constant, and the minimum regularization factor; determining the target regularization factor based on the coherence coefficient, the scaling constant, and the minimum regularization factor; determining the largest eigenvalue in the initial diagonal matrix, and determining a reference value based on the target regularization factor and the largest eigenvalue; for each element in the initial diagonal matrix, comparing the original eigenvalue corresponding to the element with the reference value, and taking the larger value between the original eigenvalue and the reference value as the target eigenvalue of the element; and adjusting the initial diagonal matrix based on the target eigenvalue corresponding to each element to obtain the target diagonal matrix.

[0015] Furthermore, before weighted fusing the target complex scattering vector corresponding to each pixel location based on the fusion weight of each receiving satellite to obtain the target image after speckle suppression for the target region, the process further includes: determining the spatial position vector of each receiving satellite in the distributed InSAR satellites; determining the spatial geometric center position of the distributed InSAR satellites based on the spatial position vector of each receiving satellite and the number of receiving satellites; calculating the spatial baseline length of each receiving satellite based on the spatial geometric center position and the spatial position vector; and calculating the fusion weight of each receiving satellite based on the spatial baseline length of each receiving satellite.

[0016] According to another aspect of the embodiments of this application, a distributed InSAR satellite image speckle suppression device is also provided, comprising: an acquisition unit for acquiring multiple images of a target area, wherein the images are acquired by a distributed InSAR satellite, the distributed InSAR satellite comprising: multiple receiving satellites, each receiving satellite being used to capture one image, and each pixel position in the image corresponding to a complex scattering value; a registration unit for registering all images to obtain registered multiple images; a construction unit for determining each pixel position in the registered multiple images, constructing an original complex scattering vector corresponding to the pixel position based on the complex scattering value corresponding to the pixel position, and constructing a coherence matrix corresponding to the pixel position based on the original complex scattering vector; a transformation unit for performing a decorrelation projection transformation on the original complex scattering vector based on the coherence matrix to obtain a target complex scattering vector corresponding to each pixel position; and a fusion unit for performing weighted fusion of the target complex scattering vector corresponding to each pixel position based on the fusion weight of each receiving satellite to obtain a speckle-suppressed target image of the target area.

[0017] Furthermore, the registration unit includes: a first acquisition module, used to acquire the satellite orbit parameters and imaging geometric relationships of each receiving satellite; a first mapping module, used to perform geometric mapping on all images based on the satellite orbit parameters and imaging geometric relationships of each receiving satellite using a preset imaging model to obtain the initial pixel correspondence between each image and the target area; a first registration module, used to perform pixel-level registration on all images based on the initial pixel correspondence to obtain multiple images after initial registration; and a second registration module, used to perform sub-pixel-level registration on all images after initial registration to obtain multiple images after registration.

[0018] Further, the second registration module includes: a first determination submodule for determining a local window; a first selection submodule for selecting a master image from all images and determining all remaining images as images to be registered; a first calculation submodule for sliding the local window and calculating the correlation coefficient between the master image and the image to be registered under the local window for each image to be registered; a second determination submodule for determining the residual offset at the position indicated by the maximum correlation coefficient and performing subpixel-level displacement on the image to be registered based on the residual offset to obtain the target pixel correspondence between the image to be registered and the master image; and a first registration submodule for registering each image to be registered with the master image based on the target pixel correspondence to obtain multiple registered images.

[0019] Furthermore, the construction unit includes: a first processing module, used to process the complex scattering value of the pixel position in each image to obtain the original complex scattering vector corresponding to the pixel position; a first construction module, used to construct a desired operation window centered on the pixel position; and a first averaging module, used to perform statistical averaging on the desired operation window based on the original complex scattering vector to obtain the coherence matrix corresponding to the pixel position.

[0020] Furthermore, the transformation unit includes: a first decomposition module for performing eigenvalue decomposition on the coherence matrix to obtain a unitary matrix and an initial diagonal matrix, wherein each element in the initial diagonal matrix corresponds to an original eigenvalue; a second processing module for performing regularization processing on the initial diagonal matrix to obtain a target diagonal matrix; a second construction module for constructing a transformation matrix based on the target diagonal matrix and the unitary matrix; and a first projection module for performing spatial projection on the original complex scattering vector based on the transformation matrix to obtain a target complex scattering vector.

[0021] Further, the second processing module includes: a third determining submodule for determining the coherence coefficient, the scaling constant, and the minimum regularization factor; a fourth determining submodule for determining the target regularization factor based on the coherence coefficient, the scaling constant, and the minimum regularization factor; a fifth determining submodule for determining the largest eigenvalue in the initial diagonal matrix and determining a reference value based on the target regularization factor and the largest eigenvalue; a first comparison submodule for comparing the original eigenvalue corresponding to each element in the initial diagonal matrix with the reference value, and taking the larger value between the original eigenvalue and the reference value as the target eigenvalue of the element; and a first adjusting submodule for adjusting the initial diagonal matrix based on the target eigenvalue corresponding to each element to obtain the target diagonal matrix.

[0022] Furthermore, the speckle suppression device also includes: a first determining module, used to determine the spatial position vector of each receiving satellite in the distributed InSAR satellites before performing weighted fusion of the target complex scattering vector corresponding to each pixel position based on the fusion weight of each receiving satellite to obtain a speckle-suppressed target image for the target region; a second determining module, used to determine the spatial geometric center position of the distributed InSAR satellites based on the spatial position vector of each receiving satellite and the number of receiving satellites; a first calculation module, used to calculate the spatial baseline length of each receiving satellite based on the spatial geometric center position and the spatial position vector; and a second calculation module, used to calculate the fusion weight of each receiving satellite based on the spatial baseline length of each receiving satellite.

[0023] According to another aspect of the embodiments of this application, a computer program product is also provided, including a non-volatile computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the distributed InSAR satellite image speckle suppression method described above.

[0024] According to another aspect of the embodiments of this application, an electronic device is also provided, including one or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by one or more processors, the one or more processors cause the one or more processors to implement any of the above-described distributed InSAR satellite image speckle suppression methods.

[0025] In this application, multiple images of a target region are acquired, and all images are registered to obtain registered multiple images. The position of each pixel in the registered multiple images is determined. Based on the complex scattering value corresponding to the pixel position, an original complex scattering vector corresponding to the pixel position is constructed. Based on the original complex scattering vector, a coherence matrix corresponding to the pixel position is constructed. Based on the coherence matrix, the original complex scattering vector is subjected to a decorrelation projection transformation to obtain the target complex scattering vector corresponding to each pixel position. Based on the fusion weight of each receiving satellite, the target complex scattering vector corresponding to each pixel position is weighted and fused to obtain a target image with speckle suppression for the target region. This solves the technical problem in related technologies that cannot effectively suppress speckle in images.

[0026] In this application, a distributed InSAR satellite synchronous acquisition of multiple images is adopted. After high-precision registration of the images, a pixel-level coherence matrix is ​​constructed. Based on eigenvalue decomposition, a decorrelation projection transformation is performed to eliminate noise correlation between channels. Combined with the method of weighted fusion of target complex scattering vectors by virtual reference point baseline, the statistical correlation of multi-channel coherence speckle noise is effectively reduced. Thus, the technical effect of improving the equivalent number of views is achieved without losing spatial resolution and ground texture. Attached Figure Description

[0027] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0028] Figure 1 This is a flowchart of an optional distributed InSAR satellite image speckle suppression method according to an embodiment of this application;

[0029] Figure 2 This is a schematic diagram of an optional distributed InSAR satellite data acquisition method according to an embodiment of this application;

[0030] Figure 3 This is an optional SAR image data decorrelation comparison diagram based on the coherence matrix according to an embodiment of this application;

[0031] Figure 4 This is an image comparison before and after speckle suppression according to an embodiment of this application;

[0032] Figure 5 This is a schematic diagram of an optional distributed InSAR image speckle suppression process based on coherence matrix feature solution correlation according to an embodiment of this application.

[0033] Figure 6 This is a schematic diagram of an optional distributed InSAR satellite image speckle suppression device according to an embodiment of this application;

[0034] Figure 7 This is a hardware structure block diagram of an electronic device (or mobile device) for a distributed InSAR satellite image speckle suppression method according to an embodiment of this application. Detailed Implementation

[0035] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0036] It should be noted that the terms "first," "second," etc., used in this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0037] To facilitate understanding of this application by those skilled in the art, the following explanations are provided for some terms or nouns involved in the various embodiments of this application:

[0038] Distributed InSAR: An interferometric synthetic aperture radar (InSAR) system employing multi-satellite formation, one-to-many transmission or dual-station / multi-station configurations can simultaneously acquire multiple complex SAR images of the same area.

[0039] Complex SAR image: A SAR image that contains both amplitude and phase information, and whose pixel values ​​are in complex form.

[0040] Coherence matrix: A second-order matrix that describes the coherence and statistical characteristics of multiple SAR images. It can quantify the degree of linear correlation between signals and noise among multiple images. For example, a 2×2 coherence matrix is ​​used for distributed InSAR dual-receiver images.

[0041] Decorrelation transformation: Based on the linear orthogonal transformation of the coherence matrix, the core purpose is to eliminate the linear correlation of speckle noise between images, so that the noise components approximately satisfy the independent distribution condition, laying the foundation for subsequent fusion and speckle suppression.

[0042] Equivalent number of views: The core quantitative indicator for measuring the speckle suppression effect of SAR images. The higher the equivalent number of views, the better the speckle noise suppression effect.

[0043] It should be noted that all information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, and displayed data) collected and involved in this application are information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of this data comply with the relevant laws, regulations, and standards of the relevant regions, and necessary measures have been taken to ensure compliance with public order and good morals. Corresponding access points are provided for users to choose whether to authorize or refuse. For example, this system has interfaces with relevant users or organizations. Before obtaining relevant information, a request to obtain the information needs to be sent to the aforementioned user or organization through the interface, and the relevant information is obtained only after receiving consent from the aforementioned user or organization.

[0044] This application belongs to the field of Synthetic Aperture Radar (SAR) remote sensing image processing technology, specifically relating to a multi-baseline image speckle suppression method for distributed Interferometric Synthetic Aperture Radar (InSAR) satellite systems.

[0045] This application addresses the characteristic of distributed InSAR satellite systems that can simultaneously acquire multiple complex SAR images of the same area under single-orbit conditions. It proposes a distributed InSAR speckle suppression method based on coherence matrix feature decorrelation and multi-image fusion. The method constructs a coherence matrix using multiple synchronous observation data, performs speckle noise decorrelation processing through eigenvalue decomposition, and performs weighted fusion of the decorrelated images. This effectively suppresses speckle noise while maintaining image spatial resolution and texture details.

[0046] This application utilizes multiple synchronous observation data for fusion processing, effectively suppressing speckle noise while maintaining the spatial resolution and texture structure information of SAR images. Furthermore, by employing multiple SAR images synchronously acquired under single-orbit conditions using a distributed InSAR system, speckle suppression can be achieved without long-term repetitive observation data, improving processing efficiency and real-time performance. By introducing coherence matrix feature decorrelation processing, the statistical correlation between multiple SAR images can be reduced, significantly increasing the equivalent number of views in the fused image, thus achieving a superior speckle suppression effect. In addition, by constructing a virtual reference point for a multi-satellite formation and using the baseline length between each satellite and this reference point to determine the multi-image fusion weights, the fusion process can reflect the spatial geometric relationships between multiple observation data. Compared to equal-weighted averaging or empirical weighting methods, this application can more rationally utilize multiple SAR image observation data, improve the equivalent number of views in the fused image, and further enhance the speckle noise suppression effect.

[0047] The present application will now be described in detail with reference to various embodiments.

[0048] Example 1

[0049] According to an embodiment of this application, an embodiment of a distributed InSAR satellite image speckle suppression method is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0050] Figure 1 This is a flowchart of an optional distributed InSAR satellite image speckle suppression method according to an embodiment of this application, such as... Figure 1 As shown, the method includes the following steps:

[0051] Step S101: Acquire multiple images of the target area. The images are acquired by distributed InSAR satellites, which include multiple receiving satellites. Each receiving satellite is used to capture one image, and each pixel position in the image corresponds to a complex scattering value.

[0052] In this embodiment, multiple complex SAR images of the same area can be acquired synchronously under single-orbit conditions using a distributed InSAR satellite system. Here, multiple receiving satellites in the distributed InSAR system simultaneously observe the same target area at the same imaging time, with each satellite independently acquiring images. Each image is a single-view complex image, where the value of each pixel is in complex form, consisting of a real part and an imaginary part, corresponding to the amplitude and phase information of the radar echo, i.e., the complex scattering value, expressed as I=A. exp(jΦ), where A is the amplitude, Φ is the phase, and j is the imaginary unit. This value reflects the backscattering characteristics of ground features under a specific observation geometry.

[0053] Figure 2 This is a schematic diagram of an optional distributed InSAR satellite data acquisition according to an embodiment of this application, such as... Figure 2 As shown, a distributed InSAR satellite system includes a launching satellite and multiple receiving satellites (e.g., receiving satellite 1 and receiving satellite 2). The launching satellite can transmit radar pulse signals to the target area, and at the same time, multiple receiving satellites synchronously receive the echoes of the signals after they are scattered by the ground surface, thereby obtaining multiple complex SAR images from multiple independent perspectives at the same imaging time and under the same radar wave illumination conditions.

[0054] Step S102: Register all images to obtain multiple registered images.

[0055] In this embodiment, because the distributed InSAR system acquires SAR images simultaneously from multiple satellites, the differences in imaging geometry, orbital position, and observation angle among the different observation satellites result in geometric offsets between the acquired SAR images. Without precise registration, the accuracy of subsequent coherence matrix estimation and multi-image fusion processing will be affected. Therefore, high-precision registration processing of the multiple SAR images is required.

[0056] Here, a combination of coarse and fine registration is used to eliminate pixel spatial offsets caused by differences in the orbital positions, imaging geometry, and attitudes of the receiving satellites. Coarse registration calculates the theoretical pixel mapping relationship based on the ephemeris and range-Doppler model of each satellite; fine registration calculates the cross-correlation function between the main image and the sub-image within a local window, and determines the sub-pixel level residual offset through interpolation, achieving pixel-level alignment of all images. After registration, the spatial coordinates of the same ground features in each image are precisely consistent, ensuring that corresponding pixels come from the same ground feature unit when constructing the subsequent coherence matrix.

[0057] Step S103: Determine the position of each pixel in the registered multiple images, construct the original complex scattering vector corresponding to the pixel position based on the complex scattering value corresponding to the pixel position, and construct the coherence matrix corresponding to the pixel position based on the original complex scattering vector.

[0058] In this embodiment, after high-precision registration of multiple SAR images is completed, it is necessary to model the statistical correlation between the images. To this end, a complex scattering vector of multiple SAR images is constructed at each pixel location, and the coherence matrix is ​​calculated based on the scattering vector.

[0059] Here, for each pixel location in the registered image, its complex scattering values ​​from each receiving satellite are extracted to form an N-dimensional complex vector, i.e., the original complex scattering vector, where N is the number of receiving satellites. Using this vector as the basic unit, statistical averaging is performed within a local window (e.g., 7×7) centered on the target pixel to calculate its covariance matrix, i.e., the coherence matrix. Its diagonal elements represent the power information of each image, and the off-diagonal elements represent the correlation between different images. This matrix characterizes the statistical correlation characteristics of multiple images at this pixel location, providing mathematical model support for decorrelation.

[0060] Step S104: Based on the coherence matrix, perform a decorrelation projection transformation on the original complex scattering vector to obtain the target complex scattering vector corresponding to each pixel position.

[0061] In this embodiment, after obtaining the SAR image coherence matrix, it is necessary to further eliminate the statistical correlation between different images to improve the speckle suppression effect during subsequent multi-image fusion. Since multiple images acquired by a distributed InSAR system at the same time have similar scattering characteristics, their speckle noise often exhibits strong correlation. Directly performing multi-image averaging fusion is insufficient to effectively reduce speckle noise. Therefore, an orthogonal transformation method based on the eigenvalue decomposition of the coherence matrix can be used to eliminate the statistical correlation between images, thereby obtaining the target complex scattering vector corresponding to each pixel location.

[0062] Here, the coherence matrix at each pixel location can be decomposed into eigenvalues ​​to obtain a unitary matrix (eigenvector matrix) and an eigenvalue diagonal matrix. A decorrelation transformation matrix is ​​constructed as the conjugate transpose of the eigenvector matrix. The original complex scattering vector is multiplied by this transformation matrix to achieve linear orthogonal projection, yielding the target complex scattering vector. This transformation maps the original correlation channels to a statistically independent orthogonal basis space, making the covariance between the transformed components approach zero, effectively eliminating the linear coupling of speckle noise between channels.

[0063] Step S105: Based on the fusion weight of each received satellite, the target complex scattering vector corresponding to each pixel position is weighted and fused to obtain the target image after speckle suppression for the target area.

[0064] In this embodiment, after decorrelation processing, the statistical correlation between multiple SAR images is significantly reduced, and the speckle noise in each image data tends to be independent. Utilizing this characteristic, speckle noise can be further reduced and image quality improved through multi-image fusion.

[0065] In this embodiment, the baseline lengths of each receiving satellite and the virtual reference point (multi-satellite geometric center) can be calculated based on the spatial geometric relationship of the distributed satellite formation. A fusion weighting function (such as an inverse baseline length function) is then constructed to assign weights to the target complex scattering vectors of each image. The complex scattering vectors of each target are then summed using their respective weights to obtain the fused complex scattering value, which serves as the final value for that pixel in the output image.

[0066] For example, suppose the multi-baseline complex scattering vector after uncorrelatedness is: ,in, Indicates the first The complex scattering values ​​of each image after decorrelation processing. A pixel-wise weighted fusion is performed on each image, and the fusion result can be expressed as: ,in, The complex scattering value after fusion; For the first The fusion weight of each satellite; Let the number of satellites be [number], and satisfy the following condition: .

[0067] By rationally designing the fusion weights, redundant information in multiple SAR image observation data can be fully utilized, thereby effectively reducing speckle noise while maintaining the spatial resolution and structural texture of the images.

[0068] By using multi-image weighted fusion processing, the equivalent number of views can be significantly improved, the speckle noise level can be reduced, and the scattering structure information and spatial resolution of the images can be maintained.

[0069] This step utilizes the statistical independence between channels after decorrelation and improves the equivalent number of views through geometrically consistent weighted fusion, thereby achieving efficient suppression of speckle noise while preserving the spatial structure and scattering characteristics of the original image.

[0070] In summary, a distributed InSAR satellite approach can be used to acquire multiple images simultaneously. After high-precision registration of the images, a pixel-level coherence matrix can be constructed. Then, based on eigenvalue decomposition, a decorrelation projection transformation can be performed to eliminate noise correlation between channels. Combined with the method of weighted fusion of target complex scattering vectors using virtual reference point baselines, the statistical correlation of multi-channel speckle noise can be effectively reduced. This achieves the technical effect of improving the equivalent number of views without sacrificing spatial resolution and ground texture.

[0071] To improve the accuracy of registration for all images, the distributed InSAR satellite image speckle suppression method provided in Embodiment 1 of this application obtains the satellite orbit parameters and imaging geometric relationships of each receiving satellite; based on the satellite orbit parameters and imaging geometric relationships of each receiving satellite, a preset imaging model is used to perform geometric mapping on all images to obtain the initial pixel correspondence between each image and the target area; based on the initial pixel correspondence, pixel-level registration is performed on all images to obtain multiple images after initial registration; sub-pixel-level registration is performed on all images after initial registration to obtain multiple images after registration.

[0072] In this embodiment, a combination of coarse and fine registration is used to achieve high-precision alignment of multiple images. First, using satellite orbital parameters, imaging geometry, and a range-Doppler (RD) imaging model, geometric mapping calculations are performed on multiple SAR images to obtain the initial pixel correspondence between each image, achieving pixel-level coarse registration. This eliminates large-scale geometric shifts caused by differences in orbital position and imaging geometry. After coarse registration, a fine registration method based on image correlation is further employed to achieve sub-pixel-level alignment.

[0073] For example, to achieve precise spatial alignment of multiple images, the orbital parameters of each receiving satellite in the distributed InSAR satellite system at the imaging time are first obtained, including its three-dimensional geocentric coordinates, velocity vector, attitude angles (yaw, pitch, roll), and radar system parameters (center frequency, pulse repetition frequency, bandwidth). Simultaneously, the imaging geometry of each satellite is acquired, i.e., the geometric mapping model of the radar beam illuminating the Earth's surface, determined by the orbital parameters and antenna pointing. Based on these parameters, a preset range-Doppler imaging model is used to perform forward geometric mapping on each image, calculating the three-dimensional coordinates of the Earth's surface corresponding to each pixel. These coordinates are then back-projected onto the imaging coordinate system of the reference image, obtaining the initial pixel correspondence between each image and the reference image. Then, based on this initial correspondence, an affine transformation is performed on all images to achieve pixel-level coarse registration, keeping the maximum geometric offset between images within 2 pixels, resulting in multiple images after initial registration. Based on this, the cross-correlation function is calculated for a local window (e.g., 15×15 pixels) of the initially registered image. By traversing the subpixel-level offset (step size 0.1 pixels) and using multiple polynomial interpolations, the residual offset is accurately estimated, and the subpixel-level registration is completed, resulting in multiple registered images.

[0074] In this embodiment, by combining coarse registration and fine registration, the geometric offset between multiple SAR images can be effectively eliminated, achieving high-precision pixel alignment and providing a reliable data foundation for subsequent coherence matrix estimation and speckle suppression processing.

[0075] To improve the accuracy of subpixel-level registration of all images after initial registration, the distributed InSAR satellite image speckle suppression method provided in Embodiment 1 of this application determines a local window; selects the master image from all images and determines all remaining images as images to be registered; slides the local window, and for each image to be registered, calculates the correlation coefficient between the master image and the image to be registered under the local window; determines the residual offset of the position indicated by the maximum correlation coefficient, and performs subpixel-level displacement of the image to be registered based on the residual offset to obtain the target pixel correspondence between the image to be registered and the master image; based on the target pixel correspondence, registers each image to be registered with the master image to obtain multiple registered images.

[0076] In this embodiment, the cross-correlation function between the main image and the image to be registered can be calculated within a local window of the image. The residual offset is determined by finding the position with the largest correlation coefficient, and sub-pixel level displacement estimation is achieved by using an interpolation method, thereby obtaining an accurate pixel correspondence.

[0077] For example, to achieve subpixel-level alignment of multiple registered images, a fixed-size local window, such as 15×15 pixels, can be defined first to ensure that the window contains sufficient speckle texture features and does not cross significant object boundaries. Then, from all the images that have been pixel-level registered, one image is selected as the master image, and the remaining images are used as images to be registered. For each image to be registered, the local window is slid across the image space with the master image as the reference, and the complex cross-correlation coefficient between the master image and the image to be registered within the window is calculated pixel by pixel. The calculation method is to sum the conjugate products of the complex scattering values ​​of all pixels within the window and then normalize them to the [0,1] interval. All possible subpixel offsets (e.g., with a step size of 0.1 pixels) are traversed, and the offset corresponding to the maximum value of the correlation coefficient is determined, which is the residual offset at the center position of the window. Multiple polynomial interpolation steps are used to fit the peak values ​​of the correlation coefficients to obtain a residual offset with sub-pixel precision. Based on this, bilinear interpolation resampling is performed on the image to be registered to complete the sub-pixel displacement and generate the target pixel correspondence corresponding to the center of the window. This correspondence is then extended to the entire image to complete the sub-pixel registration between the image to be registered and the master image, ultimately obtaining multiple registered images.

[0078] In this embodiment, by combining local window sliding and cross-correlation analysis with subpixel interpolation, high-precision registration under the condition of no dependence on ground features is achieved, which effectively suppresses the influence of the inherent speckle noise of SAR images on the matching stability and ensures the spatial consistency of multiple images at the subpixel scale.

[0079] To improve the accuracy of constructing the coherence matrix, in the distributed InSAR satellite image speckle suppression method provided in Embodiment 1 of this application, the complex scattering value of the pixel position in each image is processed to obtain the original complex scattering vector corresponding to the pixel position; a desired operation window is constructed with the pixel position as the center; based on the original complex scattering vector, the desired operation window is statistically averaged to obtain the coherence matrix corresponding to the pixel position.

[0080] In this embodiment, for each pixel location in the target region after registration of multiple images, the complex scattering values ​​in all N received satellite channels are extracted to form an N-dimensional complex vector, i.e., the original complex scattering vector, where N is the number of received satellites, and each element is the complex echo value of that channel at that pixel, containing amplitude and phase information. Then, a rectangular region of shape M×M is constructed as the expected calculation window centered on the pixel location, where M is an odd number (e.g., 7, 9) to ensure that the center of the window is strictly aligned with the target pixel. Within this window, the original complex scattering vectors of all covered pixels are statistically averaged element-wise, i.e., the average of the complex scattering values ​​of each channel is calculated, and then the outer product is calculated based on the averaged N-dimensional vector, i.e., the average vector is multiplied by its conjugate transpose to obtain an N×N complex matrix, which is the coherence matrix corresponding to the pixel location. This process suppresses single-pixel noise variance through local spatial averaging, improving the statistical robustness of coherence matrix estimation.

[0081] For example, suppose a distributed InSAR system acquires data in the same observation. For each of the complex SAR images, the complex scattering values ​​at a certain pixel location are as follows: This can then form complex scattering vectors from multiple SAR images: The coherence matrix at this pixel location is defined as: ,in, for Complex coherence matrix, Indicates the statistical expectation operation, superscript This represents the conjugate transpose. The diagonal elements of the matrix represent the power information of each image, while the off-diagonal elements represent the correlation between different images.

[0082] In practical calculations, to improve the stability of coherence matrix estimation, statistical averaging can be performed within a local window centered on the current pixel to obtain the estimated coherence matrix. The size of the local window can be set according to the image resolution and noise level; for example, using... , Or use a larger sliding window for estimation.

[0083] The coherence matrix obtained in the above manner can effectively characterize the statistical correlation characteristics between multiple SAR images and provide a basis for subsequent coherence matrix decorrelation processing.

[0084] In this embodiment, a robust estimation of the coherence matrix under finite sample conditions is achieved through a spatial neighborhood statistical averaging mechanism. This solves the problem of excessive variance in single-pixel complex scattering values ​​caused by multiplicative noise, and provides an input basis that conforms to statistical assumptions for subsequent discorrelation and weighted fusion steps, thus ensuring the stability and effectiveness of the entire speckle suppression process.

[0085] To improve the accuracy of the decorrelation projection transformation of the original complex scattering vector, the distributed InSAR satellite image speckle suppression method provided in Embodiment 1 of this application performs eigenvalue decomposition on the coherence matrix to obtain a unitary matrix and an initial diagonal matrix, wherein each element in the initial diagonal matrix corresponds to an original eigenvalue; the initial diagonal matrix is ​​regularized to obtain the target diagonal matrix; a transformation matrix is ​​constructed based on the target diagonal matrix and the unitary matrix; and the original complex scattering vector is spatially projected based on the transformation matrix to obtain the target complex scattering vector.

[0086] In this embodiment of the application, eigenvalue decomposition is performed on the coherence matrix corresponding to each pixel position. Assume the coherence matrix is... Its dimensions are , Let be the number of images. Since the coherence matrix is ​​a positive semi-definite Hermitian matrix, we perform eigenvalue decomposition on this coherence matrix: ,in, The unitary matrix, composed of eigenvectors, describes the statistical orthogonal basis directions of each satellite image. ; Let be a diagonal matrix composed of eigenvalues, where This represents the original scattering vector at the th... Power projections over orthogonal components; superscript This indicates the conjugate transpose.

[0087] The initial diagonal matrix Λ is regularized to obtain the target diagonal matrix composed of the regularized eigenvalues.

[0088] eigenvector matrix This describes the main correlation structure of the original data. Using this matrix, a linear transformation matrix can be constructed to perform an uncorrelation transformation on the original complex scattering vectors. The linear transformation matrix is ​​defined as follows. for: Then, based on the transformation matrix For the original complex scattering vector Perform spatial projection to obtain the uncorrelated target complex scattering vector. .

[0089] For example, suppose the original complex scattering vector is: The original complex scattering vector is then transformed using the aforementioned transformation matrix. Spatial projection is performed to obtain the uncorrelated scattering vector. : ,in, This is the scattering vector after uncorrelated analysis.

[0090] The physical meaning of uncorrelated transformations lies in mapping the original coupled observation space to mutually orthogonal feature spaces. Mathematically, the transformed vector... covariance matrix satisfy:

[0091] ;

[0092] because It is a diagonal identity matrix, which means the transformed scattering vector In a statistical sense, the covariance between the components is zero, which means that the linear correlation between different images is completely eliminated.

[0093] Through linear transformation, multiple statistically correlated SAR images can be mapped to a new feature space, making the different components more independent and thus reducing the correlation between speckle noise in each channel. The decorcorrelated multiple SAR image data are statistically closer to independent random variables, so the statistical averaging characteristics can be more effectively utilized in the subsequent multi-image fusion process to achieve speckle noise suppression.

[0094] In this embodiment, a decorrelation transformation framework for distributed InSAR multi-channel data is constructed through eigenvalue decomposition and adaptive regularization mechanism. This achieves physical separation of noise correlation while maintaining the integrity of the signal structure, providing input data that satisfies the statistical independence assumption for subsequent weighted fusion based on independent components, and improving the reliability and effectiveness of speckle suppression.

[0095] To improve the accuracy of regularization processing of the initial diagonal matrix, the distributed InSAR satellite image speckle suppression method provided in Embodiment 1 of this application determines the coherence coefficient, the scaling constant, and the minimum regularization factor; based on the coherence coefficient, the scaling constant, and the minimum regularization factor, a target regularization factor is determined; the largest eigenvalue in the initial diagonal matrix is ​​determined, and a reference value is determined based on the target regularization factor and the largest eigenvalue; for each element in the initial diagonal matrix, the original eigenvalue corresponding to the element is compared with the reference value, and the larger value between the original eigenvalue and the reference value is taken as the target eigenvalue of the element; based on the target eigenvalue corresponding to each element, the initial diagonal matrix is ​​adjusted to obtain the target diagonal matrix.

[0096] In this embodiment of the application, in order to keep the inversion process stable in the noise-only dimension, a method based on local coherence coefficient is introduced. With proportionality constant Adaptive regularization factor :

[0097] ;

[0098] in, It is a pre-defined minimum regularization factor.

[0099] This logarithmic mapping ensures that the regularization factor increases significantly as the coherence coefficient decreases, thus providing stronger regularization constraints in the low coherence region. The regularized eigenvalues ​​are obtained by the following formula:

[0100] ;

[0101] in, It is the largest eigenvalue in the initial diagonal matrix. The reference value is determined based on the target regularization factor and the largest eigenvalue. It is the original feature value corresponding to the nth element. It is the target feature value.

[0102] Here, the coherence coefficient refers to the ratio of the trace of the coherence matrix to the largest eigenvalue. Its range is [1, N], where N is the number of channels. The closer the value is to 1, the more concentrated the signal energy is in a single dominant direction, representing a high coherence region. The larger the value, the more dispersed the energy is, representing a low coherence or noise-dominated region. It is the input basis for adaptive regularization.

[0103] The proportionality constant is a preset scalar parameter used to control the sensitivity of the regularization factor to changes in the coherence coefficient. Its function is to adjust the regularization intensity to match the system noise level and avoid excessive enhancement of constraints in the high signal-to-noise ratio region.

[0104] The minimum regularization factor is a set lower threshold to ensure that the target feature value is not affected by regularization in the high coherence (low noise) region, thus preserving the original signal energy and preventing excessive smoothing of detailed information.

[0105] In this embodiment, by using an adaptive regularization mechanism based on coherence coefficients and a reference threshold constraint, the dual objectives of enhancing eigenvalues ​​in low-coherence regions and preserving the original signal in high-coherence regions are achieved. This effectively suppresses the non-physical attenuation of eigenvalues ​​in noise-dominated regions and improves the numerical stability and physical consistency of the regularization process.

[0106] To improve the accuracy of calculating the fusion weights for each receiving satellite, in the distributed InSAR satellite image speckle suppression method provided in Embodiment 1 of this application, before weighted fusion of the target complex scattering vectors corresponding to each pixel location based on the fusion weights of each receiving satellite to obtain the speckle-suppressed target image for the target region, the spatial position vector of each receiving satellite in the distributed InSAR satellite is determined; based on the spatial position vectors of each receiving satellite and the number of receiving satellites, the spatial geometric center position of the distributed InSAR satellite is determined; for each receiving satellite, the spatial baseline length of the receiving satellite is calculated based on the spatial geometric center position and the spatial position vector; and based on the spatial baseline length of each receiving satellite, the fusion weight of each receiving satellite is calculated.

[0107] In this embodiment, the image fusion weights can be constructed based on the spatial geometric relationships of distributed InSAR formation satellites. This can be achieved by first determining the spatial position vectors of each observed satellite. Calculate the spatial geometric center position of the formation: ,in, This represents the geometric center position of the multi-satellite formation, which can be considered a virtual reference point, where N is the number of receiving satellites. Then, the spatial baseline length between each observed satellite and this virtual reference point is calculated: A fusion weight function is constructed based on the obtained baseline length, giving greater weight to observations with smaller baselines and higher geometric consistency during the fusion process. For example, the fusion weight can be expressed as: ,in, Let be the spatial baseline length of the k-th satellite. The baseline weighting function can be designed according to the system's geometric characteristics, such as an inverse proportional function or an exponential decay function, so that images with smaller baselines and higher geometric consistency can make a greater contribution during the fusion process.

[0108] In this embodiment, by calculating the baseline length based on the satellite formation geometry and allocating inverse weights, the physical modeling of the observation geometric consistency during the fusion process is realized, enabling the geometrically optimal channel to play a dominant role in speckle suppression and suppressing the noise contribution of low-quality channels.

[0109] Figure 3 This is an optional SAR image data decorrelation comparison map based on the coherence matrix according to an embodiment of this application, such as... Figure 3 As shown, taking two images transmitted and received simultaneously as an example, the horizontal axis represents the amplitude of the first image SLC (Single Look Complex)-1, and the vertical axis represents the amplitude of the second image SLC-2. Figure 3 (a) in the figure shows the amplitude distribution of the two original images. Figure 3(b) in the figure is the amplitude distribution of the two images after decorrelation.

[0110] Figure 4 This is an optional image comparison before and after speckle suppression according to an embodiment of this application, such as... Figure 4 As shown, (a) is the original image, (b) is the image after direct statistical averaging, and (c) is the image after speckle suppression of distributed InSAR image based on coherence matrix feature solution correlation in this embodiment.

[0111] Table 1 shows the equivalent number of views for small uniform scattering region slices in SAR images before and after speckle suppression (taking two images from a single transmission and dual reception as an example):

[0112] Table 1

[0113]

[0114] Figure 5 This is a schematic diagram of an optional distributed InSAR image speckle suppression process based on coherence matrix feature solution correlation according to an embodiment of this application, as shown below. Figure 5 As shown, the process includes the following steps: For the acquired distributed InSAR satellite images-1, distributed InSAR satellite image-2, ..., distributed InSAR satellite image-N, sub-pixel-level high-precision registration can be performed, which is carried out in two stages: coarse registration (based on orbital signals and RD models) and fine registration (based on image coherence). Then, pixel-by-pixel coherence matrix estimation is performed, and a multi-channel complex scattering vector is constructed at each pixel location. The coherence matrix is ​​calculated based on the scattering vector. After that, the original data is uncorrelated based on the coherence matrix to eliminate the statistical correlation between images from different channels. Then, multi-image weighted fusion is performed, and the weighted fusion is based on the spatial geometric relationship of the distributed InSAR formation satellites to obtain images with suppressed speckle noise.

[0115] In this application, taking advantage of the capability of distributed InSAR satellites to simultaneously acquire multiple complex SAR images of the same area under single-orbit conditions, a speckle noise suppression method based on synchronously observed data from multiple SAR images is proposed. This method fully utilizes the statistical redundancy information between multiple images to achieve speckle suppression. Addressing the issue of strong correlation between speckle noise in different channels of distributed InSAR satellite images and the difficulty of effectively reducing noise through direct statistical averaging, a decorrelation processing method based on coherence matrix eigenvalue decomposition is proposed. This method reduces the statistical correlation between images through linear orthogonal transformation, thereby improving the speckle suppression capability during multi-image fusion. To address the issues of susceptibility to noise interference and poor stability in low-coherence regions, an adaptive regularization method based on local coherence coefficients is proposed. This method uses logarithmic mapping to adaptively adjust the regularization factor according to coherence, automatically enhancing constraints in low-coherence regions and preserving detailed information in high-coherence regions, thus improving the robustness of the results from a fundamental mechanism. Finally, to address the difficulty in reasonably determining weights during multi-SAR image fusion, a baseline-weighted fusion method based on virtual reference points is proposed. By constructing a virtual reference point determined by the spatial geometric relationship of a multi-satellite formation and calculating the virtual baseline length between each satellite and the virtual reference point, a fusion weight function is constructed based on the baseline length, thereby achieving adaptive weighted fusion of multiple SAR images. This allows observation channels with higher geometric consistency to make a greater contribution during the fusion process, further improving the speckle suppression effect while maintaining image structure information.

[0116] The following is a detailed description with reference to another embodiment.

[0117] Example 2

[0118] The distributed InSAR satellite image speckle suppression device provided in this embodiment includes multiple implementation units, each of which corresponds to a specific implementation step in Embodiment 1 above.

[0119] Figure 6 This is a schematic diagram of an optional distributed InSAR satellite image speckle suppression device according to an embodiment of this application, as shown below. Figure 6 As shown, the speckle suppression device may include: an acquisition unit 60, a registration unit 61, a construction unit 62, a transformation unit 63, and a fusion unit 64.

[0120] The acquisition unit 60 is used to acquire multiple images of the target area. The images are acquired by distributed InSAR satellites, which include multiple receiving satellites. Each receiving satellite is used to capture one image, and each pixel position in the image corresponds to a complex scattering value.

[0121] Registration unit 61 is used to register all images to obtain multiple registered images;

[0122] The construction unit 62 is used to determine the position of each pixel in the registered multiple images, construct the original complex scattering vector corresponding to the pixel position based on the complex scattering value corresponding to the pixel position, and construct the coherence matrix corresponding to the pixel position based on the original complex scattering vector.

[0123] Transformation unit 63 is used to perform a discorrelation projection transformation on the original complex scattering vector based on the coherence matrix to obtain the target complex scattering vector corresponding to each pixel position;

[0124] The fusion unit 64 is used to perform weighted fusion of the target complex scattering vector corresponding to each pixel position based on the fusion weight of each received satellite, so as to obtain the target image after speckle suppression for the target region.

[0125] The aforementioned speckle suppression device can acquire multiple images simultaneously using distributed InSAR satellites. After high-precision registration of the images, a pixel-level coherence matrix is ​​constructed. Based on eigenvalue decomposition, a decorrelation projection transformation is performed to eliminate noise correlation between channels. Combined with the method of weighted fusion of target complex scattering vectors using virtual reference point baselines, the device effectively reduces the statistical correlation of multi-channel speckle noise, thereby achieving the technical effect of increasing the equivalent number of views without sacrificing spatial resolution and ground texture.

[0126] Optionally, the registration unit includes: a first acquisition module for acquiring the satellite orbit parameters and imaging geometric relationships of each receiving satellite; a first mapping module for performing geometric mapping on all images using a preset imaging model based on the satellite orbit parameters and imaging geometric relationships of each receiving satellite, to obtain the initial pixel correspondence between each image and the target area; a first registration module for performing pixel-level registration on all images based on the initial pixel correspondence, to obtain multiple images after initial registration; and a second registration module for performing sub-pixel-level registration on all images after initial registration, to obtain multiple images after registration.

[0127] Optionally, the second registration module includes: a first determining submodule for determining a local window; a first selecting submodule for selecting a master image from all images and determining all remaining images as images to be registered; a first calculating submodule for sliding the local window and calculating the correlation coefficient between the master image and the image to be registered under the local window for each image to be registered; a second determining submodule for determining the residual offset at the position indicated by the maximum correlation coefficient and performing subpixel-level displacement on the image to be registered based on the residual offset to obtain the target pixel correspondence between the image to be registered and the master image; and a first registration submodule for registering each image to be registered with the master image based on the target pixel correspondence to obtain multiple registered images.

[0128] Optionally, the construction unit includes: a first processing module, used to process the complex scattering value of the pixel position in each image to obtain the original complex scattering vector corresponding to the pixel position; a first construction module, used to construct a desired operation window centered on the pixel position; and a first averaging module, used to perform statistical averaging on the desired operation window based on the original complex scattering vector to obtain the coherence matrix corresponding to the pixel position.

[0129] Optionally, the transformation unit includes: a first decomposition module for performing eigenvalue decomposition on the coherence matrix to obtain a unitary matrix and an initial diagonal matrix, wherein each element in the initial diagonal matrix corresponds to an original eigenvalue; a second processing module for performing regularization processing on the initial diagonal matrix to obtain a target diagonal matrix; a second construction module for constructing a transformation matrix based on the target diagonal matrix and the unitary matrix; and a first projection module for performing spatial projection on the original complex scattering vector based on the transformation matrix to obtain a target complex scattering vector.

[0130] Optionally, the second processing module includes: a third determining submodule for determining the coherence coefficient, the scaling constant, and the minimum regularization factor; a fourth determining submodule for determining the target regularization factor based on the coherence coefficient, the scaling constant, and the minimum regularization factor; a fifth determining submodule for determining the largest eigenvalue in the initial diagonal matrix and determining a reference value based on the target regularization factor and the largest eigenvalue; a first comparison submodule for comparing the original eigenvalue corresponding to each element in the initial diagonal matrix with the reference value, and taking the larger value between the original eigenvalue and the reference value as the target eigenvalue of the element; and a first adjusting submodule for adjusting the initial diagonal matrix based on the target eigenvalue corresponding to each element to obtain the target diagonal matrix.

[0131] Optionally, the speckle suppression device further includes: a first determining module, used to determine the spatial position vector of each receiving satellite in the distributed InSAR satellites before performing weighted fusion of the target complex scattering vector corresponding to each pixel position based on the fusion weight of each receiving satellite to obtain a speckle-suppressed target image for the target region; a second determining module, used to determine the spatial geometric center position of the distributed InSAR satellites based on the spatial position vector of each receiving satellite and the number of receiving satellites; a first calculation module, used to calculate the spatial baseline length of each receiving satellite based on the spatial geometric center position and the spatial position vector; and a second calculation module, used to calculate the fusion weight of each receiving satellite based on the spatial baseline length of each receiving satellite.

[0132] The aforementioned speckle suppression device may also include a processor and a memory. The aforementioned acquisition unit 60, registration unit 61, construction unit 62, transformation unit 63, fusion unit 64, etc., are all stored in the memory as program units, and the processor executes the aforementioned program units stored in the memory to realize the corresponding functions.

[0133] The aforementioned processor contains a kernel, which retrieves the corresponding program units from memory. One or more kernels can be configured. By adjusting kernel parameters, the complex scattering vectors of the target at each pixel location are weighted and fused based on the fusion weights of each received satellite, resulting in a target image with speckle suppression for the target region.

[0134] The aforementioned memory may include non-permanent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0135] This application also provides a computer program product, which, when executed on a data processing device, is suitable for executing a program that initializes the following steps: acquiring multiple images of a target region, registering all images to obtain registered multiple images, determining the position of each pixel in the registered multiple images, constructing an original complex scattering vector corresponding to the pixel position based on the complex scattering value corresponding to the pixel position, constructing a coherence matrix corresponding to the pixel position based on the original complex scattering vector, performing a decorrelation projection transformation on the original complex scattering vector based on the coherence matrix to obtain a target complex scattering vector corresponding to each pixel position, and performing weighted fusion on the target complex scattering vector corresponding to each pixel position based on the fusion weight of each receiving satellite to obtain a target image with speckle suppression for the target region.

[0136] According to another aspect of the embodiments of this application, a computer program product is also provided, including a non-volatile computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the distributed InSAR satellite image speckle suppression method described above.

[0137] According to another aspect of the embodiments of this application, an electronic device is also provided, including one or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by one or more processors, the one or more processors cause the one or more processors to implement the above-described distributed InSAR satellite image speckle suppression method.

[0138] Figure 7This is a hardware structure block diagram of an electronic device (or mobile device) for a distributed InSAR satellite image speckle suppression method according to an embodiment of this application. Figure 7 As shown, an electronic device may include one or more processors (e.g., Figure 7 The processors 702a, 702b, ..., 702n, etc., may include, but are not limited to, processing devices such as microprocessors (MCUs) or programmable logic devices (FPGAs), and a memory 704 for storing data. In addition, it may include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the I / O interface), a network interface, a keyboard, a power supply, and / or a camera. Those skilled in the art will understand that... Figure 7 The structure shown is for illustrative purposes only and does not limit the structure of the electronic device described above. For example, the electronic device may also include components that are more... Figure 7 The more or fewer components shown, or having the same Figure 7 The different configurations shown.

[0139] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0140] The embodiments or examples disclosed herein are not exhaustive, but merely illustrative of some embodiments or examples, and are not intended to limit the scope of protection of this disclosure. Unless otherwise specified, each step in a particular embodiment or example can be implemented as an independent embodiment, and the steps can be arbitrarily combined. For example, a solution after removing some steps in a particular embodiment or example can also be implemented as an independent embodiment, and the order of the steps in a particular embodiment or example can be arbitrarily interchanged. Furthermore, optional methods or examples in a particular embodiment or example can be arbitrarily combined; moreover, embodiments or examples can be arbitrarily combined. For example, some or all steps of different embodiments or examples can be arbitrarily combined, and a particular embodiment or example can be arbitrarily combined with optional methods or examples of other embodiments or examples.

[0141] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0142] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0143] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0144] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0145] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0146] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A distributed InSAR satellite image speckle suppression method, characterized in that, include: Multiple images of a target area are acquired, wherein the images are acquired synchronously by distributed InSAR satellites under single-orbit observation conditions. The distributed InSAR satellites include multiple receiving satellites, each of which is used to capture one image. Each pixel position in the image corresponds to a complex scattering value. All the images are registered to obtain multiple registered images; Determine the position of each pixel in the registered images, construct the original complex scattering vector corresponding to the pixel position based on the complex scattering value corresponding to the pixel position, and construct the coherence matrix corresponding to the pixel position based on the original complex scattering vector; Based on the coherence matrix, the original complex scattering vector is subjected to a decorrelation projection transformation to obtain the target complex scattering vector corresponding to each pixel position; Based on the fusion weight of each of the received satellites, the target complex scattering vector corresponding to each pixel position is weighted and fused to obtain a target image with speckle suppression for the target region; The images are registered using a combination of coarse and fine registration. The fine registration process includes: A local window is determined; a master image is selected from all the images, and all remaining images are determined as images to be registered; the local window is slid, and for each image to be registered, the correlation coefficient between the master image and the image to be registered is calculated under the local window; the residual offset of the position indicated by the maximum correlation coefficient is determined, and the image to be registered is displaced at the sub-pixel level based on the residual offset to obtain the target pixel correspondence between the image to be registered and the master image; based on the target pixel correspondence, each image to be registered is registered with the master image to obtain multiple registered images.

2. The method for suppressing speckle according to claim 1, characterized in that, The step of registering all the images to obtain a plurality of registered images includes: Obtain the satellite orbital parameters and imaging geometry of each of the receiving satellites; Based on the satellite orbit parameters and imaging geometry of each receiving satellite, a preset imaging model is used to perform geometric mapping on all the images to obtain the initial pixel correspondence between each image and the target region; Based on the initial pixel correspondence, pixel-level registration is performed on all the images to obtain multiple images after initial registration; Subpixel-level registration is performed on all the images after the initial registration to obtain multiple registered images.

3. The method for suppressing speckle according to claim 1, characterized in that, The steps of constructing an original complex scattering vector corresponding to the pixel location based on the complex scattering value corresponding to the pixel location, and constructing a coherence matrix corresponding to the pixel location based on the original complex scattering vector, include: The complex scattering value of the pixel location in each of the images is processed to obtain the original complex scattering vector corresponding to the pixel location; Construct the desired computation window centered on the pixel position; Based on the original complex scattering vector, the expected operation window is statistically averaged to obtain the coherence matrix corresponding to the pixel position.

4. The method for suppressing speckle according to claim 1, characterized in that, The step of performing a decorrelation projection transformation on the original complex scattering vector based on the coherence matrix to obtain the target complex scattering vector corresponding to each pixel position includes: The coherence matrix is ​​decomposed into eigenvalues ​​to obtain a unitary matrix and an initial diagonal matrix, wherein each element in the initial diagonal matrix corresponds to an original eigenvalue; The initial diagonal matrix is ​​regularized to obtain the target diagonal matrix; Based on the target diagonal matrix and the unitary matrix, construct the transformation matrix; Based on the transformation matrix, the original complex scattering vector is spatially projected to obtain the target complex scattering vector.

5. The method for suppressing speckle according to claim 4, characterized in that, The step of regularizing the initial diagonal matrix to obtain the target diagonal matrix includes: Determine the coherence coefficient, the proportionality constant, and the minimum regularization factor; The target regularization factor is determined based on the coherence coefficient, the proportionality constant, and the minimum regularization factor. Determine the largest eigenvalue in the initial diagonal matrix, and determine a reference value based on the target regularization factor and the largest eigenvalue; For each element in the initial diagonal matrix, the original eigenvalue corresponding to the element is compared with the reference value, and the larger of the original eigenvalue and the reference value is taken as the target eigenvalue of the element. Based on the target feature value corresponding to each element, the initial diagonal matrix is ​​adjusted to obtain the target diagonal matrix.

6. The method for suppressing speckle according to claim 1, characterized in that, Before weighted fusing the target complex scattering vector corresponding to each pixel location based on the fusion weight of each received satellite to obtain the speckle-suppressed target image for the target region, the method further includes: Determine the spatial location vector of each of the receiving satellites in the distributed InSAR satellites; Based on the spatial position vector of each of the receiving satellites and the number of receiving satellites, the spatial geometric center position of the distributed InSAR satellites is determined; For each of the receiving satellites, the spatial baseline length of the receiving satellite is calculated based on the spatial geometric center position and the spatial position vector; The fusion weight of each receiving satellite is calculated based on the spatial baseline length of each receiving satellite.

7. A distributed InSAR satellite image speckle suppression device, characterized in that, include: An acquisition unit is used to acquire multiple images of a target area, wherein the images are acquired synchronously by distributed InSAR satellites under single-orbit observation conditions. The distributed InSAR satellites include: multiple receiving satellites, each of which is used to capture one image, and each pixel position in the image corresponds to a complex scattering value. A registration unit is used to register all the images to obtain multiple registered images; A construction unit is used to determine the position of each pixel in the registered multiple images, construct the original complex scattering vector corresponding to the pixel position based on the complex scattering value corresponding to the pixel position, and construct the coherence matrix corresponding to the pixel position based on the original complex scattering vector. The transformation unit is used to perform a decorrelation projection transformation on the original complex scattering vector based on the coherence matrix to obtain the target complex scattering vector corresponding to each pixel position; The fusion unit is used to perform weighted fusion of the target complex scattering vector corresponding to each pixel position based on the fusion weight of each of the received satellites, so as to obtain a target image after speckle suppression for the target region; The images are registered using a combination of coarse and fine registration. The fine registration process includes: A local window is determined; a master image is selected from all the images, and all remaining images are determined as images to be registered; the local window is slid, and for each image to be registered, the correlation coefficient between the master image and the image to be registered is calculated under the local window; the residual offset of the position indicated by the maximum correlation coefficient is determined, and the image to be registered is displaced at the sub-pixel level based on the residual offset to obtain the target pixel correspondence between the image to be registered and the master image; based on the target pixel correspondence, each image to be registered is registered with the master image to obtain multiple registered images.

8. A computer program product, characterized in that, The method includes a non-volatile computer-readable storage medium storing a computer program that, when executed by a processor, implements the distributed InSAR satellite image speckle suppression method according to any one of claims 1 to 6.

9. An electronic device, characterized in that, It includes one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the distributed InSAR satellite image speckle suppression method according to any one of claims 1 to 6.

Images