False target identification method and device based on low-rank sparse decomposition shadow extraction

Abstract

The invention discloses a false target identification method and device based on low-rank sparse decomposition shadow extraction, and belongs to the field of synthetic aperture radar false target identification. Aiming at pain points with inaccurate shadow extraction and poor recognition robustness caused by speckle noise, the method comprises the following steps of: firstly, dividing regions of interest for a deception jamming SAR (Synthetic Aperture Radar) image; then, the CA-CFAR is used for eliminating strong scattering points; converting the image into a low-rank-sparse matrix through translation reconstruction, so that the shadow is in a low-rank state and the noise is in a sparse state; a residual threshold is adopted to drive low-rank sparse decomposition adaptive separation, and sparseness priori is not needed; oTSU binaryzation and morphological processing are carried out to obtain a preliminary shadow; according to a target-shadow fixed geometric angle, clustering correction is carried out, false detection is eliminated, and a final shadow is obtained; and finally, completing identification according to the condition that no shadow is a false target. According to the method, shadow and noise decoupling is realized under the single-channel condition, and the extraction precision and the recognition robustness are remarkably improved.

Description

Technical Field

[0001] This invention belongs to the field of synthetic aperture radar (SAR) false target recognition, specifically relating to a false target recognition method and apparatus based on low-rank sparse decomposition shadow extraction. Background Technology

[0002] Synthetic Aperture Radar (SAR), as an advanced active microwave remote sensing technology, plays a crucial role in reconnaissance, environmental awareness, target detection, and identification. SAR jamming disrupts SAR imaging results through specific interference methods, thereby hindering the acquisition of effective intelligence. Unlike suppression jamming, which obscures protected targets in the imaging results, deception jamming can create highly realistic false targets, deliberately misleading SAR reconnaissance systems and seriously threatening the reliability of SAR information acquisition. To address this challenge, it is urgent to develop methods for identifying false targets in SAR deception jamming to improve the credibility of SAR image interpretation under jammed environments.

[0003] Existing methods for identifying deceptive targets in SAR systems fall into three categories: spatial diversity, time-frequency feature analysis, and image analysis. Spatial diversity methods identify deceptive targets by analyzing the differences in signal phase and imaging position between deceptive targets and real targets in a multi-channel / multi-static SAR system. However, these methods rely on specific observation conditions in multi-channel / multi-static systems, resulting in limitations such as high system complexity, restricted application scenarios, and inflexible equipment deployment. Time-frequency feature analysis methods perform time-domain windowing and frequency-domain filtering on single-channel echo signals to enhance the separability between real and deceptive targets, thus enabling deceptive target identification. However, in the vicinity of the jammer, the performance of these methods deteriorates sharply due to the high similarity in time-frequency characteristics between deceptive and real targets.

[0004] Image analysis methods directly analyze imaging results, avoiding reliance on complex multi-channel / multi-static SAR systems and effectively mitigating failures in the vicinity of jammers. They offer significant advantages such as simple system structure and wide applicability. Existing image analysis methods mainly include shape-based and shadow-based false target identification methods. Shape-based methods effectively identify false targets with specific morphological features, such as intermittent sampling forwarding jamming and micro-motion modulation jamming. Shadow-based methods rely on the lack of a physical basis for shadow generation in deception jamming, using shadow extraction methods to analyze the presence of shadows associated with the target, thereby achieving false target identification.

[0005] With the rapid development of SAR deception and jamming technology, the types of false targets generated by deception jamming are becoming increasingly diverse, and their morphological features and electromagnetic properties are becoming increasingly similar to those of real targets. However, shape-based false target identification methods rely excessively on specific false target shapes, resulting in insufficient generalization ability and difficulty in effectively coping with the continuous evolution of jamming techniques. Shadow-based false target identification methods have a generalization advantage in principle, extracting the accompanying shadows of targets based on image histogram information to identify shadowless false targets. However, due to the influence of speckle noise in SAR images, shadow extraction is quite difficult, significantly reducing the reliability of this method. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention proposes a method and apparatus for false target identification based on low-rank sparse decomposition shadow extraction. This method effectively suppresses the influence of speckle noise in SAR images, significantly improves the quality and robustness of shadow extraction, and achieves high-precision and high-reliability false target identification. Furthermore, this invention employs a low-rank sparse decomposition algorithm driven by residual thresholds. This algorithm can achieve decomposition without prior information on sparsity, effectively avoiding dependence on prior sparsity.

[0007] The technical problem to be solved by this invention is achieved through the following technical solution:

[0008] A method for identifying false targets based on shadow extraction using low-rank sparse decomposition includes the following steps performed sequentially:

[0009] Step 1: Acquire SAR images collected under deception and interference conditions, and determine several regions of interest for false target identification;

[0010] Step 2: The region of interest is preprocessed using the constant false alarm rate (CA-CFAR) method to remove high-brightness target pixels;

[0011] Step 3: Generate a reconstruction matrix based on the preprocessing results. And will reconstruct the matrix Modeled as a low-rank matrix sparse matrix With error matrix A linear combination of, where Preserve the strong low-rank characteristics of the shaded region. Capture the sparsity characteristics of speckle noise;

[0012] Step 4, reconstruct the matrix Perform low-rank sparse decomposition to obtain the separated low-rank matrix. ;

[0013] Step 5, for the low-rank matrix OTSU adaptive binarization and morphological processing are performed sequentially to obtain a connected and complete preliminary shadow mask;

[0014] Step 6: Based on the fixed geometric angle relationship between the target and the shadow in SAR imaging, perform adaptive angle correction on the preliminary shadow mask, remove false shadows that deviate from the main cluster range, obtain the final shadow extraction result, and determine false targets based on whether the target in the region of interest lacks associated shadows.

[0015] A false target identification device based on low-rank sparse decomposition shadow extraction includes:

[0016] The region of interest identification module acquires SAR images collected under deception and interference conditions and identifies several regions of interest for false target identification.

[0017] The preprocessing module preprocesses the region of interest using the constant false alarm rate (CA-CFAR) method to remove high-brightness target pixels.

[0018] The reconstruction matrix generation module generates a reconstruction matrix based on the preprocessing results. And will reconstruct the matrix Modeled as a low-rank matrix sparse matrix With error matrix A linear combination of, where Preserve the strong low-rank characteristics of the shaded region. Capture the sparsity characteristics of speckle noise;

[0019] The low-rank matrix separation module performs matrix reconstruction. Perform low-rank sparse decomposition to obtain the separated low-rank matrix. ;

[0020] The shadow mask acquisition module is used for low-rank matrices. OTSU adaptive binarization and morphological processing are performed sequentially to obtain a connected and complete preliminary shadow mask;

[0021] The false target identification module adaptively corrects the initial shadow mask based on the fixed geometric angle relationship between the target and the shadow in SAR imaging, eliminates false shadows that deviate from the main cluster range, obtains the final shadow extraction result, and identifies false targets based on whether the target in the region of interest lacks associated shadows.

[0022] An electronic device includes: one or more processors; and a memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the method.

[0023] A computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, cause the processor to implement the method described thereon.

[0024] The present invention has the following beneficial effects:

[0025] 1. This invention proposes a false target identification method based on low-rank sparse decomposition shadow extraction. This method reconstructs the original SAR image by translation into a reconstruction matrix with a low-rank sparse structure. In this matrix space, the shadow region exhibits strong low-rank characteristics due to the structural correlation brought about by the translation, while speckle noise exhibits sparse distribution characteristics due to its statistical independence. This reconstruction transformation overcomes the limitation that the signal and speckle noise are difficult to separate in spatial distribution in the original image domain. Based on this, the low-rank sparse decomposition method is used to decouple shadow features from speckle noise, effectively solving the problems of inaccurate shadow extraction and poor recognition robustness caused by speckle noise, and significantly improving the accuracy of false target identification.

[0026] 2. This invention proposes a low-rank sparse decomposition algorithm based on residual thresholding. Addressing the difficulty in predicting the sparsity of speckle noise in the translation reconstruction matrix, this invention innovatively designs a residual thresholding mechanism to adaptively adjust the threshold parameter during the decomposition iteration process. This mechanism solves the problem of decomposition algorithm failure caused by the lack of prior sparsity information, thus eliminating the algorithm's dependence on prior sparsity information.

[0027] 3. This invention proposes a false target identification method based on low-rank sparse decomposition shadow extraction, which exhibits excellent performance in terms of false target identification accuracy and robustness. Specifically, this invention uses shadow features as the core criterion for false target identification, effectively decoupling shadow features from speckle noise in the translation reconstruction matrix. Simultaneously, a residual threshold-driven mechanism is introduced during the decomposition process, enabling adaptive adjustment of decomposition parameters and avoiding dependence on prior sparsity information. This invention effectively overcomes the challenge of limited shadow feature extraction accuracy under the influence of speckle noise in SAR images, thereby achieving high-precision and highly robust false target identification. Attached Figure Description

[0028] Figure 1 This is a flowchart illustrating a method for identifying false targets based on shadow extraction using low-rank sparse decomposition.

[0029] Figure 2 To reconstruct the matrix Generate a schematic diagram.

[0030] Figure 3 This is a schematic diagram showing the geometric angles between the preliminary shadow extraction results and their corresponding targets.

[0031] Figure 4 This is a schematic diagram of the deception interference simulation and region of interest based on measured data.

[0032] Figure 5 This is a schematic diagram comparing the performance of the method proposed in this invention with existing shadow extraction methods.

[0033] Figure 6 This is a schematic diagram showing the visualization results of the method proposed in this invention and existing shadow extraction methods. Detailed Implementation

[0034] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of the invention described below can be combined with each other as long as they do not conflict with each other. To achieve the above objectives, the invention adopts the following technical solutions.

[0035] Figure 1 This is a flowchart illustrating a false target identification method based on low-rank sparse decomposition shadow extraction provided in an embodiment of the present invention, as shown below. Figure 1 As shown, the method mainly includes the following steps:

[0036] Step 1: Acquire SAR images captured under deception and interference conditions, and determine several regions of interest (ROIs) for false target identification. The ROIs are local areas in the image that may contain real targets or deception / interference targets, determined through manual interpretation or automatic target detection algorithms.

[0037] Step 2: Preprocess the region of interest (ROI) using the Constant False Alarm Rate (CA-CFAR) method. To avoid high-brightness targets dominating the image grayscale distribution within the ROI and hindering accurate analysis of low-grayscale shadows, the target pixels within the ROI undergo CA-CFAR detection and removal preprocessing. Specifically, the pixel to be detected is used as the detection unit, and its surrounding pixels are... A reference window is formed by 10 pixels, and the pixel value of each reference unit is denoted as 1. ( In CA-CFAR, the average background power within the reference window. for:

[0038] (1)

[0039] Then set the detection threshold. :

[0040] (2)

[0041] Among them, threshold factor It is determined by a preset false alarm probability. Total length of reference window Confirmed. If the pixel value at the detection unit... If the pixel is a target pixel, its amplitude is set to zero to eliminate the interference of strong scattering points on subsequent shadow extraction; otherwise, it is determined to be a non-target pixel and its original amplitude is maintained.

[0042] Step 3: Reconstruction Matrix Generation and Low-Rank Sparse Modeling. A reconstruction matrix is ​​generated based on the preprocessing results. The reconstructed matrix is ​​constructed as a linear combination of three matrix components: The low-rank matrix The shadow regions with strong low-rank properties are preserved, and the sparse matrix is ​​preserved. Speckle noise in SAR images was captured. It is the error matrix of the low-rank sparse decomposition.

[0043] Step 4, reconstruct the matrix A low-rank sparse decomposition is performed. The reconstructed matrix is ​​processed using a low-rank sparse decomposition algorithm driven by a residual threshold. The separated low-rank matrix This achieves the goal of suppressing speckle noise. The core of this approach is to solve the following optimization problem:

[0044] (3)

[0045] in, Denotes the Frobenius norm. express Norm. Threshold. It is the sparsity regularization parameter. Yes Low-rank constraints, Represents low-rank components The upper limit of the rank constraint.

[0046] Step 5: Preliminary shadow extraction based on OTSU and morphological processing. The OTSU algorithm is used to process the low-rank matrix. Adaptive binarization is performed to initially obtain the shadow mask. Then, morphological processing is used to optimize the initial mask, resulting in a preliminary shadow extraction result with complete connectivity.

[0047] Step 6: Obtain the final shadow extraction result and false target identification result through adaptive angle correction. Based on the fixed geometric angle relationship between the target and its shadow in SAR imaging, the preliminary shadow extraction result is optimized. Specifically, firstly, the geometric angle between each preliminary shadow extraction result and its corresponding target is statistically analyzed. Secondly, cluster analysis is performed on the statistically obtained angle values. Shadow regions corresponding to discrete angle values ​​that significantly deviate from the main cluster range are identified as false detections and removed, thus obtaining the final shadow extraction result. Based on the final shadow extraction result, false target identification is achieved by determining whether the target in the region of interest lacks its associated shadow.

[0048] The reconstruction matrix provided in the embodiments of the present invention Generate a schematic diagram as follows Figure 2 As shown, step 3 can be achieved through the following steps:

[0049] Step 3.1, perform preprocessing on the constant false alarm rate (CFAR) detection results. Second translation, generating A collection of images with similar frame shapes and grayscale distributions , and These represent the width and height of the preprocessed result, respectively.

[0050] Step 3.2, for each image in the image set Vectorize the data, arrange them sequentially, and concatenate them to obtain the reconstructed matrix. .

[0051] Step 3.3, the reconstructed matrix is ​​constructed as a linear combination of three matrix components: The low-rank matrix The shadow regions with strong low-rank properties are preserved, and the sparse matrix is ​​preserved. Speckle noise in SAR images was captured. It is the error matrix of the low-rank sparse decomposition.

[0052] In this embodiment of the invention, the reconstruction matrix... Perform low-rank sparse decomposition to obtain the separated low-rank matrices. The optimization problem described in step 4 above can be solved through the following steps:

[0053] Step 4.1, Initialize the low-rank matrix sparse matrix .

[0054] Step 4.2: Solve the optimization problem using an alternating iterative optimization method. Specifically, in the... During rounds of iteration:

[0055] Fixed sparse matrix The optimization problem of equation (3) can be rewritten as:

[0056] (4)

[0057] In the optimization subproblem (4), according to Optimal low-rank approximation update of low-rank matrix .

[0058] Fixed update The optimization problem of equation (3) can be rewritten as:

[0059] (5)

[0060] in, Let be the residual matrix.

[0061] based on Absolute median difference generates threshold parameter , Representing the residual matrix The One element, yes The median of all elements in the set. This represents the set of numerical values ​​consisting of the calculation results corresponding to all elements in the matrix. The optimal solution to the optimization subproblem (5) is:

[0062] (6)

[0063] Among them, residual threshold According to residuals The statistical properties are used to dynamically separate speckle noise with abnormal amplitude into a sparse matrix. .

[0064] Step 4.3, when the decomposition error ,satisfy When the iteration converges, the final separation result is obtained. and .in This is a preset minimum value, set to 0.001 in some implementation examples.

[0065] In this embodiment of the invention, after obtaining the preliminary shadow extraction result based on OTSU and morphological processing, step 6 above can be implemented through the following steps:

[0066] Step 6.1: Calculate the geometric angle between each preliminary shadow extraction result and its corresponding target. A schematic diagram of the geometric angle between the preliminary shadow extraction result and its corresponding target is shown below. Figure 3 As shown, through vectors (From the centroid of the target) Centroid of the initial shadow extraction results The angle between the x-axis and the reference direction To quantify the geometric relationship between the target and its shadow:

[0067] (7)

[0068] Step 6.2: Perform cluster analysis on the angle statistics results, identify the shadow areas corresponding to discrete angle values ​​that significantly deviate from the main cluster range as false detections and remove them, thereby obtaining the final shadow extraction results.

[0069] Step 6.3: Based on the final shadow extraction result, false target identification is achieved by determining whether the target in the region of interest lacks a shadow associated with it.

[0070] The effectiveness of this invention will be further explained below with reference to simulation experiments. The simulation experiment software platform of this invention is: convolutional modulation deception interference simulation of the acquired measured SAR image in Matlab R2021a. The simulation experiment provided by this invention is based on the analysis of the deception interference simulation image, and the simulation parameters are shown in Table 1. Figure 4 (a) and (b) are the deception interference simulation images based on measured data, and the regions of interest for false target identification determined by the automatic target detection algorithm (all regions of interest are numbered, of which number 4 and 12 are deception interference false targets generated by simulation).

[0071] To quantitatively evaluate the accuracy of shadow extraction and false target identification, this invention employs two metrics: Intersection Over Union (IoU) and Critical Success Index (CSI), defined as follows:

[0072] (8)

[0073] (9)

[0074] in, By calculating the true shadow area With shadow extraction results The degree of overlap quantifies the accuracy of shadow extraction. It's important to note that when both the actual shadow area and the extracted result are empty sets (i.e.,...), the accuracy is affected. ),definition The value is 1, thus accurately reflecting the method's analysis of such unshaded samples. middle , and These represent the number of correctly identified false targets, the number of real targets misclassified as false targets, and the number of false targets not identified, respectively. This metric effectively measures the performance of false target identification by penalizing false positives and false negatives.

[0075] Another aspect of the present invention provides a false target identification device based on low-rank sparse decomposition shadow extraction, comprising:

[0076] The region of interest identification module acquires SAR images collected under deception and interference conditions and identifies several regions of interest for false target identification.

[0077] The preprocessing module preprocesses the region of interest using the constant false alarm rate (CA-CFAR) method to remove high-brightness target pixels.

[0078] The reconstruction matrix generation module generates a reconstruction matrix based on the preprocessing results. And will reconstruct the matrix Modeled as a low-rank matrix sparse matrix With error matrix A linear combination of, where Preserve the strong low-rank characteristics of the shaded region. Capture the sparsity characteristics of speckle noise;

[0079] The low-rank matrix separation module performs matrix reconstruction. Perform low-rank sparse decomposition to obtain the separated low-rank matrix. ;

[0080] The shadow mask acquisition module is used for low-rank matrices. OTSU adaptive binarization and morphological processing are performed sequentially to obtain a connected and complete preliminary shadow mask;

[0081] The false target identification module adaptively corrects the initial shadow mask based on the fixed geometric angle relationship between the target and the shadow in SAR imaging, eliminates false shadows that deviate from the main cluster range, obtains the final shadow extraction result, and identifies false targets based on whether the target in the region of interest lacks associated shadows.

[0082] Another aspect of the present invention provides an electronic device, comprising: one or more processors; and a memory for storing one or more programs, wherein, when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the method.

[0083] Another aspect of the present invention provides a computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, cause the processor to implement the method described thereon.

[0084] The shadow extraction performance and false target detection performance based on different shadow extraction methods are as follows: Figure 5 And as shown in Table 2. Figure 6(a)-(g) show the visualization results of shadow extraction in the region of interest using different shadow extraction methods. Among them, the method based on geometric information refers to the method in the published literature "S. Yan, Y. Fu, R. Yu, C. Luo, W. Zhang, and W. Yang, “High-precision moving target shadow detection algorithm for ViSAR based on information geometry,” IEEE J. Sel. Topics Appl. Earth Observ. Remote Sens., vol. 18, pp. 12 728-12 739, 2025, doi: 10.1109 / JSTARS.2025.3568810." OTSU-SLIC refers to the method in the published paper “Z. Wu, H. Xie, T. Gao, Y. Zhang, and H. Liu, “Moving target shadow detection method based on improved ViBe in VideoSAR images,” IEEE J. Sel. Topics Appl. Earth Observ. Remote Sens., vol. 17, pp. 14 575-14 587, 2024, doi: 10.1109 / JSTARS.2024.3443080.” The change detection-based method refers to the method in the published paper “H. Zhang, S. Quan, S. Xing, J. Wang, Y. Li, and P. Wang, “Shadow-based false target identification for SAR images,” Remote Sens., vol. 15, no. 21, 2023, Art. no.5259, doi: 10.3390 / rs15215259.”The dual-threshold OTSU refers to the method described in the published paper “F. Gao, J. You, J. Wang, J. Sun, E. Yang, and H. Zhou, “A novel target detection method for SAR images based on shadow proposal and saliency analysis,” Neurocomputing, vol. 267, pp. 220-231, 2017, doi: 10.1016 / j.neucom.2017.06.004.”

[0085] from Figure 5 As shown in Table 2, the information geometry-based method and OTSU-SLIC suffer from low shadow extraction accuracy due to speckle noise interference and fail to identify false targets, resulting in a CSI of 0. The change detection-based method and the dual-threshold OTSU method exhibit poor shadow extraction robustness, leading to misclassification of some real targets as false targets, with CSIs of only 40% and 66.7%, respectively. In contrast, the shadow extraction algorithm provided in this invention can accurately extract shadows in all regions of interest, demonstrating higher accuracy and robustness, and successfully identifying all false targets in the deceptive interference simulation image based on measured data. These results verify that the proposed method can effectively overcome the problem of decreased shadow extraction accuracy caused by speckle noise in SAR images, thereby achieving high-precision and highly robust false target identification.

[0086] Table 1. Simulation parameters for deception and interference

[0087]

[0088] Table 2. False target identification results of the proposed method and existing shadow extraction methods.

[0089]

Claims

1. A method for identifying false targets based on shadow extraction using low-rank sparse decomposition, characterized in that, The following steps are performed sequentially: Step 1: Acquire SAR images collected under deception and interference conditions, and determine several regions of interest for false target identification; Step 2: The region of interest is preprocessed using the constant false alarm rate (CA-CFAR) method to remove high-brightness target pixels; Step 3: Generate a reconstruction matrix based on the preprocessing results. And will reconstruct the matrix Modeled as a low-rank matrix sparse matrix With error matrix A linear combination of, where Preserve the strong low-rank characteristics of the shaded region. Capture the sparsity characteristics of speckle noise; Step 4, reconstruct the matrix Perform low-rank sparse decomposition to obtain the separated low-rank matrix. ; Step 5, for the low-rank matrix OTSU adaptive binarization and morphological processing are performed sequentially to obtain a connected and complete preliminary shadow mask; Step 6: Based on the fixed geometric angle relationship between the target and the shadow in SAR imaging, perform adaptive angle correction on the preliminary shadow mask, remove false shadows that deviate from the main cluster range, obtain the final shadow extraction result, and determine false targets based on whether the target in the region of interest lacks associated shadows.

2. The method according to claim 1, characterized in that, In step 2, the CA-CFAR method sets a detection threshold. The background average power The threshold factor is calculated from the average pixel value within the reference window. Based on the preset false alarm probability With reference window length Sure.

3. The method according to claim 1, characterized in that, Reconstructing the matrix in step 3 The generation method is as follows: preprocessing the results... The image set is obtained by secondary translation. The images are vectorized and then sequentially concatenated into a matrix. , and These represent the width and height of the preprocessed result, respectively.

4. The method according to claim 1, characterized in that, Residual threshold in step 4 Press at each iteration Adaptive update, where Representing the residual matrix The One element, yes The median of all elements in the set. It represents the set of numerical values ​​consisting of the calculation results corresponding to all elements in the matrix.

5. The method according to claim 1, characterized in that, In step 6, through vectors From the centroid of the target Centroid of the initial shadow extraction results Angle with the reference direction x-axis To quantify the geometric relationship between the target and its shadow: 。 6. The method according to claim 1, characterized in that, The iterative convergence condition in step 4 is the decomposition error. satisfy At that time, among them Take 0.

001.

7. The method according to any one of claims 1 to 6, characterized in that, The method is implemented on single-channel SAR images.

8. A false target recognition device based on low-rank sparse decomposition shadow extraction, characterized in that, include: The region of interest identification module acquires SAR images collected under deception and interference conditions and identifies several regions of interest for false target identification. The preprocessing module preprocesses the region of interest using the constant false alarm rate (CA-CFAR) method to remove high-brightness target pixels. The reconstruction matrix generation module generates a reconstruction matrix based on the preprocessing results. And will reconstruct the matrix Modeled as a low-rank matrix sparse matrix With error matrix A linear combination of, where Preserve the strong low-rank characteristics of the shaded region. Capture the sparsity characteristics of speckle noise; The low-rank matrix separation module performs matrix reconstruction. Perform low-rank sparse decomposition to obtain the separated low-rank matrix. ; The shadow mask acquisition module is used for low-rank matrices. OTSU adaptive binarization and morphological processing are performed sequentially to obtain a connected and complete preliminary shadow mask; The false target identification module adaptively corrects the initial shadow mask based on the fixed geometric angle relationship between the target and the shadow in SAR imaging, eliminates false shadows that deviate from the main cluster range, obtains the final shadow extraction result, and identifies false targets based on whether the target in the region of interest lacks associated shadows.

9. An electronic device, characterized in that, include: One or more processors; A memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the method of any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It stores executable instructions that, when executed by a processor, cause the processor to perform the method described in any one of claims 1 to 7.

Images