SAR (Synthetic Aperture Radar) image domain LFM (Linear Frequency Modulation) interference modulation rate estimation and interference removal method

Abstract

The invention discloses an SAR image domain LFM interference frequency modulation rate estimation and interference removal method, and belongs to the technical field of SAR image processing. According to the method, aiming at the problem that a satellite-borne SAR image is interfered by ground and ship-borne radars and the like, a frequency modulation rate fast estimator is provided based on least square estimation to obtain an interfered frequency modulation rate parameter; and removing the LFM interference focused as the narrow point response in the two-dimensional frequency domain through operations such as slope removal, Fourier transform, spectrum support generation, filtering, inverse transformation and phase compensation by using a spectrum focusing algorithm. According to the method, the LFM interference in the SAR single-view complex (SLC) image can be effectively removed, interference artifacts are remarkably filtered out, the image quality is improved, the method is robust to the interference power level, and the original image information is hardly lost.

Description

Technical Field

[0001] The 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 the application of modern SAR imaging technology, radio frequency interference is a common problem that seriously affects image quality. In the low-frequency band, SAR images are often interfered by various wireless signal sources such as AM / FM broadcasting, television, and communication networks. In the C-band, the signals emitted by ground and shipborne radars become one of the main sources of SAR image interference, and there is also mutual interference between spaceborne SARs. These interference signals are usually in the form of linear frequency modulation pulses. Compared with ordinary low-frequency band interference, they have broadband characteristics and are collectively referred to as broadband LFM interference.

[0003] When there is unremoved LFM interference in the original SAR image data, bright radiation artifacts will be generated after the image focusing process. These artifacts will not only have a negative impact on the visual effect of the image, but more seriously, they may cover up or submerge the useful information in the image, making it impossible to accurately obtain key information such as the reflectivity of the ground object, thereby greatly reducing the effectiveness of SAR images in many application fields such as terrain mapping and ground object monitoring. Summary of the invention

[0004] The present invention aims to provide an efficient post-processing algorithm for SAR images, which can realize the accurate removal of 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 fields of application.

[0005] The technical solution to achieve the purpose of the present invention is: a method for estimating and removing LFM interference modulation frequency in 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 distance frequency modulation of the LFM interference in each data block;

[0008] The spectral 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 Respectively represent 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] In the formula, t a,m It is a discrete variable related to position and time, used to represent the position and time points under 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 situations.

[0018] Preferably, the specific method of 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 least squares estimator is used to estimate the azimuth frequency modulation rate. The least squares estimator is specifically:

[0020]

[0021] In the formula, Z′(f a ) is the Fourier transform of z′[m,n] in the azimuth direction; K a is the azimuth chirp 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 optimization of estimated frequency, and 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 ) is subjected to a two-dimensional Fourier transform to obtain:

[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 spectral components falling within the range of the spectral 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 effect: The present invention can effectively remove LFM interference in SAR images and significantly reduce the presence of interference artifacts. Through experimental verification of actual S-1 interferometric wide-band single-view complex images, the reflectivity information of ground objects that was originally obscured by interference can be clearly restored in the processed images, and the image quality is significantly improved, providing a more accurate and reliable data basis for subsequent image analysis and application.

[0034] Strong robustness to interference power: The algorithm of the present invention shows good performance under interference conditions of different strengths. Whether it is strong interference or weak interference, it can be effectively detected and removed, and has strong adaptability and robustness to the interference power level. This feature makes the method widely applicable in SAR image processing under complex interference environments.

[0035] Information is preserved intact: In the process of removing interference, the algorithm of the present invention hardly loses the information of the original image. Through the analysis of the interference map (i.e., the difference image between the input image and the output image), it is found that in addition to the effective removal of interference artifacts, information such as ground objects, ocean reflectivity, and background noise in the image is well preserved. This ensures the accuracy and integrity of the SAR image after removing interference, so that the value of the image in various application fields can be fully reflected. 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 the experimental image example 1. The top 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 reflection that was originally obscured in the land area is clearly presented; the bottom is the interference map generated by subtracting the output SLC image from the input SLC image, in which 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 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 is the interference map generated by subtracting the input and output SLC images. There are interference artifacts of varying strengths, which reflects the algorithm's adaptability to interference intensity and strongly proves 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 modulation 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, which are the key factors to describe the signal position and propagation characteristics in SAR images. ξ and ζ are the center time of the interference signal in azimuth and range directions, respectively, which 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 removal of interference. c 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: In order to facilitate mathematical calculations and parameter estimation, the discrete forms of LFM interference and image samples are first defined. 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 the 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 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 and 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 ) is subjected to a two-dimensional Fourier transform, and we obtain 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 threshold T depends on the characteristics of image data and the actual situation of interference. By setting the threshold reasonably, the range of 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 spectral 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 can compensate for the phase change introduced during the de-slope process, ensure the phase accuracy of the final processed image, and thus improve the image quality.

[0053] like Figure 1 As shown, the top is an image containing LFM interference, and the bottom 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] Firstly, 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 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 step of generating spectrum support, the threshold is reasonably selected according to the statistical characteristics of the image data and the manifestation of 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 the interference is removed.

[0064] 3. Result verification and analysis

[0065] Make an intuitive comparison between the images before and after processing, observe the removal of interference artifacts, and check whether the information such as ground objects originally covered by interference is clearly visible, so as to preliminarily evaluate the interference removal effect of the algorithm.

[0066] The interference map is generated by calculating the difference between the input image and the output image. The distribution of other pixel values ​​in the interference map except for the interference artifacts is analyzed in detail to determine the degree of retention of the original image information (such as ground reflectivity, ocean reflectivity, and background noise, etc.), thereby fully verifying the algorithm's ability to protect the original image information and the effectiveness of interference removal.

[0067] In the actual application process, various parameters in the algorithm (such as thresholds, data block sizes, etc.) can be flexibly adjusted and optimized according to specific SAR image data characteristics, application requirements, and computing environment factors to achieve the best interference removal effect. At the same time, 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 technologies or improving frequency modulation estimation methods to ensure that interference can still be effectively removed in a constantly changing technical 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: Acquire single-view complex data of the SAR image, and divide the single-view complex data into a number of image blocks; Estimating the frequency modulation of each image block to obtain the azimuth frequency modulation and distance frequency modulation of the LFM interference in each data block; The spectral focusing algorithm is used to remove LFM interference, and several image blocks are synthesized to obtain the image after interference removal.

2. The SAR image domain LFM interference modulation frequency estimation and interference removal method according to claim 1 is characterized in that: 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 Respectively represent 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 ta,m is a discrete variable related to azimuth time, which is used to represent the azimuth time point in the case of discrete sampling.

3. The method for estimating and removing LFM interference modulation frequency in SAR image domain according to claim 2, characterized in that: 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 discrete sampling situations.

4. The method for estimating and removing LFM interference modulation frequency in SAR image domain according to claim 3, characterized in that: The specific method of estimating the frequency modulation of each image block and obtaining the azimuth frequency modulation and distance frequency modulation of the LFM interference in each data block is as follows: The least squares estimator is used to estimate the azimuth frequency modulation rate. The least squares estimator is specifically: In the formula, Z′(f a ) is the Fourier transform of z′[m,n] in the azimuth direction; K a is the azimuth chirp 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 optimization of estimated frequency, and K is a temporary variable that serves as an intermediate variable in the approximate calculation and optimization process.

5. The method for estimating and removing LFM interference modulation frequency in SAR image domain according to claim 4, characterized in that: The specific method of removing LFM interference using the spectrum focusing algorithm is: 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 ) is subjected to a two-dimensional Fourier transform to obtain: 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 spectral components falling within the range of the spectral 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