A method for LFM interference frequency estimation and interference removal in SAR image domain

Abstract

The present invention discloses a method for estimating and removing the LFM interference frequency modulation rate in the SAR image domain, and belongs to the technical field of SAR image processing. This method addresses the problem of interference to spaceborne SAR images caused by ground-based and shipborne radars. First, a frequency modulation rate fast estimator is proposed based on least squares estimation to obtain the frequency modulation rate parameters of the interference. A spectrum focusing algorithm is then used to remove the LFM interference focused on a narrow point response in the two-dimensional frequency domain through operations such as de-slope, Fourier transform, spectrum support generation, filtering, inverse transform, and phase compensation. The present invention can effectively remove LFM interference in SAR single-look complex (SLC) images, significantly filter out interference artifacts, improve image quality, and is robust to interference power levels with almost no loss of original image information.

Description

Technical Field

[0001] The present invention belongs to the technical field of synthetic aperture radar (SAR) image processing, and in particular relates to a method for estimating and removing LFM interference modulation frequency in SAR image domain. Background Art

[0002] In modern SAR imaging applications, radio frequency interference (RFI) is a common problem that seriously impacts image quality. In low-frequency bands, SAR images are often interfered with by various wireless signal sources, including AM / FM radio, television, and communication networks. In the C-band, signals transmitted by ground-based and shipborne radars are a major source of interference, and there is also mutual interference between spaceborne SARs. These interfering signals typically take the form of linear frequency modulation (LFM) pulses, which, compared to typical low-frequency interference, have broadband characteristics and are collectively referred to as broadband LFM interference.

[0003] When LFM interference is present in raw SAR image data and is not removed, bright radiometric artifacts are generated after image focusing. These artifacts not only negatively impact the visual quality of the image but, more seriously, can obscure or overwhelm useful information in the image, preventing the accurate acquisition of critical information such as surface reflectivity. This significantly reduces the effectiveness of SAR imagery in numerous applications, such as terrain mapping and surface monitoring. Summary of the Invention

[0004] The present invention aims to provide an efficient post-processing algorithm for SAR images, which can accurately remove LFM interference in images, thereby significantly improving the quality of SAR images, restoring useful information masked by interference, and enhancing the reliability and accuracy of SAR images in various applications.

[0005] The technical solution for achieving the purpose of the present invention is: a method for estimating and removing LFM interference modulation frequency in the SAR image domain, comprising:

[0006] Acquire single-view complex data of the SAR image, and divide the single-view complex data into a number of image blocks;

[0007] Estimating the frequency modulation of each image block to obtain the azimuth frequency modulation and range frequency modulation of the LFM interference in each data block;

[0008] The spectrum focusing algorithm is used to remove LFM interference, and several image blocks are synthesized to obtain the image after interference removal.

[0009] Preferably, the LFM interference signal in the SAR image is specifically:

[0010]

[0011] Where γ is a scalar related to the interference intensity level, rect(·) represents the rectangular function, and t a and t r Represent the azimuth and distance time respectively, ξ and ζ are the center time of the interference signal in the azimuth and distance directions respectively, T a and T r Represents the span of interference in azimuth and distance directions, K a and K r are the azimuth and distance modulation frequencies, respectively, c is the radar's carrier frequency;

[0012] The discrete form of the LFM interference signal is:

[0013]

[0014] Where, t a,m It is a discrete variable related to position and time, and is used to represent the position and time points in discrete sampling conditions.

[0015] Preferably, the discrete form of the SAR image is:

[0016] z[m,n]=γ n s[m,n]+v[m,n]

[0017] in, v[m,n] represents the image information and noise components without LFM interference; t r,n It is a discrete variable related to distance and time, used to represent the distance and time points in discrete sampling conditions.

[0018] Preferably, the specific method for estimating the frequency modulation rate for each image block and obtaining the azimuth frequency modulation rate and the range frequency modulation rate of the LFM interference in each data block is:

[0019] The azimuth frequency modulation rate is estimated using a least squares estimator, which is specifically:

[0020]

[0021] Where, Z′(f a ) is the Fourier transform of z′[m,n] in the azimuth direction; K a is the azimuth frequency modulation slope, f a is the azimuth frequency, N a It is related to the number of azimuth sampling points. f is a variable for spectrum analysis and optimized frequency estimation. K is a temporary variable that serves as an intermediate variable in the approximate calculation and optimization process.

[0022] Preferably, the specific method of removing LFM interference using the spectrum focusing algorithm is:

[0023] The original mixed data z(t a ,t r ) is used for deslope processing, and the specific formula is:

[0024]

[0025] For the signal z′(t a ,t r ) performs a two-dimensional Fourier transform and obtains:

[0026]

[0027] According to the set threshold T, the spectrum support Ω of LFM interference is generated in the two-dimensional frequency domain, which is defined as Ω={(f a ,f r )||Z(f a ,f r )|≥T};

[0028] For the spectrum components that fall within the range of the spectrum support Ω, that is, Z(f a ,f r ), set its value to 0, that is, Z(f a ,f r )←0;

[0029] The filtered spectrum Z(f a ,f r ) is inverse Fourier transformed to obtain

[0030] The inverse Fourier transformed signal Multiply by the conjugate of the de-ramping function, which is:

[0031]

[0032] Compared with the prior art, the present invention has the following significant advantages:

[0033] Significant interference removal: This method effectively removes LFM interference from SAR images, significantly reducing the presence of interference artifacts. Experimental verification using actual S-1 interferometric wide-band single-look complex images shows that the processed images clearly restore the reflectivity of ground objects previously obscured by interference, significantly improving image quality and providing a more accurate and reliable data foundation for subsequent image analysis and applications.

[0034] Robustness to Interference Power: The algorithm of this invention demonstrates excellent performance under interference conditions of varying strengths. It can effectively detect and remove both strong and weak interference, demonstrating strong adaptability and robustness to interference power levels. This characteristic makes the method widely applicable to SAR image processing in complex interference environments.

[0035] Complete Information Preservation: During the interference removal process, the algorithm of this invention loses virtually no information from the original image. Analysis of the interference map (the difference between the input and output images) reveals that, in addition to effectively removing interference artifacts, information such as ground objects, ocean reflectivity, and background noise is well preserved. This ensures the accuracy and integrity of SAR images after interference removal, fully realizing their value in various applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 Examples of LFM interference and focusing interferometry in SAR images.

[0037] Figure 2 This is a flowchart of the interference removal algorithm in SAR images.

[0038] Figure 3 This is experimental image example 1. The top row is the SAR image before applying the proposed algorithm. It can be seen that there are many bright radiation artifacts introduced by the interference signal; the second row is the output image after applying the proposed algorithm. The interference artifacts are significantly reduced, and the land reflections that were originally obscured in the land area are clearly presented; the bottom row is the interference map generated by subtracting the output SLC image from the input SLC image. There are interference artifacts of varying strengths, and there are almost no sea and ground reflectivity and background noise components, which proves the robustness of the algorithm to interference intensity.

[0039] Figure 4 This is experimental image example 2. The top row is the SAR image before applying the proposed algorithm. A large number of bright radiation artifacts caused by interference signals can be seen in the top area. The second row is the output image after applying the algorithm. The interference artifacts are effectively removed, indicating that the algorithm is effective in improving image quality. The bottom row is the interference map generated by subtracting the input and output SLC images. There are interference artifacts of varying strengths, reflecting the algorithm's adaptability to interference intensity and strongly demonstrating the algorithm's ability to preserve the original image information well when removing interference. DETAILED DESCRIPTION

[0040] A method for estimating and removing LFM interference frequency in SAR image domain, comprising:

[0041] 1. Signal model establishment

[0042] In order to accurately describe the characteristics of the LFM interference signal observed in the spaceborne SAR image, the present invention adopts the following approximate signal model: In this model, γ is a scalar related to the interference intensity level, and rect(·) represents a rectangular function, which is used to limit the effective range of the signal in the time domain. a and t r Represents azimuth and range time, respectively. These two parameters are key factors in describing the signal location and propagation characteristics in SAR images. ξ and ζ are the center times of the interference signal in azimuth and range directions, respectively. They determine the center position of the interference signal in the time domain. a and T r They represent the span of interference in the azimuth and distance directions, respectively, and describe the duration of the interference signal in the time domain. a and K r These two parameters are important characteristics of LFM interference signals, which determine the rate of change of signal frequency over time and are crucial for subsequent accurate estimation and interference removal. c It is the carrier frequency of the radar and the basic frequency parameter of the entire SAR system. a ,t r ) is the LFM interference signal.

[0043] 2. Frequency modulation estimation method

[0044] (1) Discrete form definition: To facilitate mathematical calculations and parameter estimation, we first define the discrete forms of LFM interference and image samples. The discrete form of LFM interference is The discrete form of image samples is z[m,n]=γ n s[m,n]+v[m,n], where v[m,n] represents the image information without LFM interference and noise components.

[0045] (2) Derivation and approximation of the least squares estimator: Based on the above discrete form, the least squares estimator is derived However, directly computing this estimator requires two-dimensional space (K a ,ξ) to maximize the likelihood function, the computational cost is high and not suitable for practical applications. Therefore, an approximate method is used Get an approximate estimator in Z′(f a ) is the Fourier transform of z′[m,n] in the azimuth direction. Through this approximation, the complex two-dimensional search problem is simplified, so that the estimator can be implemented through grid search, thereby effectively estimating the modulation frequency K a For the distance modulation frequency Kr , a similar method can be used for estimation to ensure that the frequency modulation rate parameters of the LFM interference signal can be fully and accurately obtained, providing a key basis for subsequent interference removal.

[0046] 3. Use spectrum focusing algorithm to remove LFM interference, including

[0047] De-slope processing: First, the original data z(t a ,t r ) to perform a de-slope operation, i.e. The purpose of this step is to adjust the characteristics of the LFM interference signal in the time-frequency domain so that it can be better distinguished from other signal components in subsequent processing, laying the foundation for accurate spectrum analysis and interference removal.

[0048] Fourier transform: the signal z′(t a ,t r ) performs a two-dimensional Fourier transform and obtains Through Fourier transform, the signal is converted from the time domain to the frequency domain, which makes it possible to observe the spectral characteristics of the signal more intuitively and facilitates the subsequent identification and processing of the interference signal spectrum.

[0049] Generate the spectrum support of WLI in the two-dimensional frequency domain: According to the set threshold T, generate the spectrum support Ω of LFM interference in the two-dimensional frequency domain, which is defined as Ω = {(f a ,f r )||Z(f a ,f r )|≥T}. The selection of the threshold T depends on the characteristics of the image data and the actual situation of the interference. By setting the threshold reasonably, the range of the interference signal in the spectrum can be accurately defined, thus providing an accurate basis for subsequent filtering operations.

[0050] Spectral filtering: For the spectral components that fall within the range of the spectrum support Ω, that is, Z(f a ,f r ), set its value to 0, that is, Z(f a ,f r )←0. This step effectively filters out the interference signal spectrum and removes the main LFM interference components in the image.

[0051] Inverse Fourier transform: The filtered spectrum Z(f a ,f r ) is inverse Fourier transformed to obtain Convert the signal from the frequency domain back to the time domain and restore the representation of the signal in the time domain for further processing.

[0052] Phase compensation: Finally, the inverse Fourier transformed signal Multiply by the conjugate of the de-ramping function, which is This compensates for the phase change introduced during the de-slope process, ensuring the phase accuracy of the final processed image and thus improving image quality.

[0053] like Figure 1 As shown, the upper part is an image containing LFM interference, and the lower part is a schematic diagram of the spectrum after focusing LFM interference, which intuitively shows that LFM interference can be focused in the form of a narrow point response in the two-dimensional spectrum domain of the image.

[0054] like Figure 2 As shown, the entire process from data input to interference removal output is clearly presented.

[0055] Example

[0056] 1. Data Acquisition and Preparation

[0057] First, SAR single-look complex (SLC) data are obtained from a suitable data source to ensure the integrity and accuracy of the data for subsequent interference removal processing.

[0058] 2. Interference estimation and removal operations

[0059] The acquired input SLC is divided into several data blocks. The size and method of the division can be reasonably selected according to the actual situation of the image and computing resources to balance processing efficiency and accuracy.

[0060] For each data block, do the following:

[0061] (1) Using the above frequency modulation estimation method, accurately estimate the azimuth frequency modulation K of the LFM interference in the data block a and distance modulation frequency K r This step is the key to the entire interference removal process, and accurate frequency modulation estimation provides an important basis for subsequent spectrum processing.

[0062] (2) According to the algorithm flow based on spectrum focusing, de-slope processing, Fourier transform, spectrum support generation, spectrum filtering, inverse Fourier transform and phase compensation are performed in sequence. In the spectrum support generation step, the threshold is reasonably selected according to the statistical characteristics of the image data and the manifestation of the interference to ensure accurate identification and removal of the interference spectrum.

[0063] The processed data blocks are merged according to the structure of the original image to obtain the final output image after removing interference.

[0064] 3. Result Verification and Analysis

[0065] Compare the images before and after processing intuitively, observe the removal of interference artifacts, and check whether the information such as ground objects originally obscured by interference is clearly visible, so as to preliminarily evaluate the interference removal effect of the algorithm.

[0066] An interference map is generated by calculating the difference between the input and output images. The distribution of pixel values ​​other than interference artifacts in the interference map is analyzed in detail to determine the degree of preservation of original image information (such as ground object reflectance, ocean reflectance, and background noise). This fully verifies the algorithm's ability to preserve original image information and the effectiveness of interference removal.

[0067] In actual applications, various algorithm parameters (such as thresholds, data block sizes, etc.) can be flexibly adjusted and optimized based on factors such as the specific characteristics of the SAR image data, application requirements, and computing environment to achieve optimal interference removal. Furthermore, with the continuous development of SAR technology and the increasing diversification of application scenarios, the algorithm of the present invention also has the potential for further improvement and expansion to adapt to more complex interference situations and higher-quality image processing requirements. For example, with the development of radar technology, new types of interference or more complex interference patterns may appear in the future. The algorithm framework of the present invention can be expanded by introducing new signal processing techniques or improving frequency modulation estimation methods to ensure effective interference removal in an ever-changing technological environment, improve SAR image quality, and provide strong support for applications in related fields.

Claims

1. A method for estimating and removing LFM interference frequency in SAR image domain, characterized in that: include: Obtain the single-view complex data of the SAR image and divide the single-view complex data into several image blocks. The LFM interference signal in the SAR image is specifically: Where γ is a scalar related to the interference intensity level, rect(·) represents the rectangular function, and t a and t r Represent the azimuth and distance time respectively, ξ and ζ are the center time of the interference signal in the azimuth and distance directions respectively, T a and T r Represents the span of interference in azimuth and distance directions, K a and K r are the azimuth and distance modulation frequencies, respectively, c is the radar's carrier frequency; The discrete form of the LFM interference signal is: Where, t a,m It is a discrete variable related to azimuth time, used to represent the azimuth time point in the case of discrete sampling; The discrete form of SAR image is: z[m,n]=γ n s[m,n]+v[m,n] in, v[m,n] represents the image information and noise components without LFM interference; t r,n It is a discrete variable related to distance and time, used to represent the distance and time points in the case of discrete sampling; The frequency modulation rate is estimated for each image block to obtain the azimuth frequency modulation rate and range frequency modulation rate of the LFM interference in each data block. The specific method is as follows: The azimuth frequency modulation rate is estimated using a least squares estimator, which is specifically: Where, Z′(f a ) is the Fourier transform of z′[m,n] in the azimuth direction; K a is the azimuth frequency modulation slope, f a is the azimuth frequency, N a Related to the number of azimuth sampling points, f is a variable for spectrum analysis and optimized frequency estimation, and K is a temporary variable that serves as an intermediate variable in the approximate calculation and optimization process; The same least squares estimator is used to estimate the range frequency modulation rate; The spectral focusing algorithm is used to remove LFM interference, and several image blocks are synthesized to obtain the image after interference removal. The specific method is as follows: The original mixed data z(t a ,t r ) is used for deslope processing, and the specific formula is: For the signal z′(t a ,t r ) performs a two-dimensional Fourier transform and obtains: According to the set threshold T, the spectrum support Ω of LFM interference is generated in the two-dimensional frequency domain, which is defined as Ω={(f a ,f r )||Z(f a ,f r )|≥T}; For the spectrum components that fall within the range of the spectrum support Ω, that is, Z(f a ,f r ), set its value to 0, that is, Z(f a ,f r )←0; The filtered spectrum Z(f a ,f r ) is inverse Fourier transformed to obtain The inverse Fourier transformed signal Multiply by the conjugate of the de-ramping function, which is:

Images