A large squint SAR imaging method with precise squint angle estimation

Abstract

The present application belongs to synthetic aperture radar signal processing field, disclose a kind of oblique look angle accurate estimation's large oblique SAR imaging method: step 1, distance compression obtains enhanced image;Step 2: construct secondary distance curvature compensation quantity and transform into distance direction pixel offset;Interpolation is carried out to enhanced image and distance curvature compensation image is obtained;Step 3, using adaptive threshold, local density constraint filtering and connected domain screening process, obtain effective structure point;When the number of effective structure point meets threshold value, enter step 4, otherwise end;Step 4: calculate angle response function value, determine Hough detection angle;Step 5: calculate image domain straight line slope;When the current iteration number reaches preset value, enter step 6, otherwise return to step 2;Step 6: output result.The present application can effectively solve the problem that main direction estimation is unstable and low in precision caused by distance curvature under large oblique condition in existing Hough transform method.

Description

Technical Field

[0001] This invention belongs to the field of synthetic aperture radar signal processing technology, specifically relating to a large squint SAR imaging method for accurate estimation of the squint angle. Background Technology

[0002] High-quality imaging in Synthetic Aperture Radar (SAR) relies on accurate estimation of the Doppler center frequency. Under frontal or small oblique viewing angles, conventional frequency domain estimation methods can achieve high estimation accuracy. However, in large oblique viewing modes, severe integer multiples of Doppler ambiguity exist at the Doppler center. Antenna pattern distortion and strong interference further lead to serious deviations in the estimation results of frequency domain methods (such as the energy centroid method). Moreover, these spectrum-fitting-based methods can only obtain the Doppler center within the baseband and cannot directly solve for the Doppler ambiguity number, thus failing to meet the requirements of high-resolution, high-precision imaging.

[0003] The Hough transform method based on the echo envelope estimates the unambiguous Doppler center using the linear travel component of the target echo envelope, eliminating the need to solve for ambiguity numbers. However, under large squint conditions, the target envelope exhibits significant range curvature, deviating from the Hough transform's assumption that the target trajectory is approximately a straight line. Directly projecting the curved linear trajectory into the Hough parameter space leads to insufficient energy accumulation, broadening of the main peak, and even the appearance of spurious peaks, severely impacting the accuracy of main direction detection and squint angle estimation. Summary of the Invention

[0004] The purpose of this invention is to provide a large squint SAR imaging method with accurate squint angle estimation, so as to solve the problem that the existing Hough transform method has unstable principal direction estimation and low accuracy due to distance curvature under large squint conditions.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] A large squint SAR imaging method for accurate estimation of the squint angle includes the following steps: Step 1: Acquire 2D SAR echo data and perform range compression to obtain an amplitude image; perform contrast enhancement on the amplitude image to obtain an enhanced image; set the initial oblique angle estimation value. Step 2: Construct a secondary distance curvature compensation value from the oblique angle estimation value in the current iteration and convert it into a distance-to-pixel offset; Combine the distance-to-pixel offset with the enhanced image obtained in Step 1 and perform backward mapping linear interpolation to obtain the distance curvature compensation image in the current iteration. Step 3: Normalize the maximum value of the distance curvature compensation image in the current iteration, and then process it sequentially using adaptive thresholding, local density constraint filtering and connected component filtering to obtain effective structure points; when the number of effective structure points meets the preset threshold, proceed to step 4, otherwise the iteration ends. Step 4: Determine the search range of the Hough transform based on the estimated angle of view in the current iteration; within the search range of the Hough transform, obtain the angle response function value based on the magnitude of the effective structure point, and determine the Hough detection angle based on the angle response function value; Step 5: Calculate the current slope of the line in the image domain using the Hough detection angle; determine whether the current iteration number has reached the preset value. If yes, record the estimated angle of view in the current iteration and proceed to step 6; otherwise, update the estimated angle of view and return to step 2. Step 6: Output the mean of the oblique angle estimation values ​​and the current slope of the line in the image domain.

[0007] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) It breaks through the traditional Hough transform's assumption that the target trajectory is approximately a straight line. By iteratively straightening the curved trajectory through distance curvature compensation, the Hough transform can still stably detect the main direction under large squint conditions, which significantly improves the reliability of the squint angle estimation and thus improves the accuracy.

[0008] (2) By adopting an effective structural point extraction strategy based on adaptive threshold screening and local density constraints, as well as an energy-weighted Hough voting mechanism, the interference of isolated noise points, weak clutter and local strong scattering structures on the main direction detection can be effectively suppressed, thereby improving the robustness of the estimation.

[0009] (3) The sliding window mean convergence criterion determines whether the iteration is stable by taking the average of the most recent instantaneous estimation results. It can effectively smooth the estimation fluctuations caused by noise, multi-target coupling and discrete sampling errors, avoid the interference of single detection outliers on the final result, and thus improve accuracy.

[0010] (4) The entire process only requires distance to the compressed amplitude image, without relying on precise initial values ​​and preset Doppler parameters, and is suitable for scenarios where the initial information is inaccurate or missing. Attached Figure Description

[0011] Figure 1 It is the convergence curve of the estimated angle during the iteration process; Figure 2 This is a simulation imaging result of a point target; Figure 3 This is the focusing imaging result of the method of the present invention on the measured data. Detailed Implementation

[0012] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments.

[0013] The large squint SAR imaging method for accurate estimation of the squint angle provided in this invention includes the following steps: Step 1: Acquire 2D SAR echo data and perform range compression to obtain an amplitude image. Then, enhance the contrast of the amplitude image to obtain an enhanced image. Set the initial oblique angle estimation value. The specific operations are as follows: Two-dimensional SAR echo data is acquired and range-compressed to obtain range-compressed two-dimensional SAR echo data. Based on two-dimensional SAR echo data Obtain amplitude image :

[0014] In the formula: and These are the distance and azimuth pixel coordinates, respectively, both of which are positive integers; Set the angle of view estimate in the first iteration. ,Pick The initial angle value may be provided by the flight platform's inertial navigation system, which is relatively coarse.

[0015] For amplitude image Amplitude normalization was performed using MATLAB's mat2gray function. The normalized image is then subjected to contrast enhancement using MATLAB's imadjust function within the specified interval to obtain the enhanced image.

[0016] Step 2, Distance Curvature Compensation: Construct a secondary distance curvature compensation value from the oblique angle estimation value in the current iteration and convert it into a distance-to-pixel offset; combine the distance-to-pixel offset with the enhanced image obtained in Step 1 and perform backward mapping linear interpolation to obtain the distance curvature compensated image in the current iteration.

[0017] Step 2 includes the following sub-steps: Step 21, using the estimated oblique angle value from the current iteration. Construct the secondary distance bending compensation amount, where The expression for the quadratic distance curvature compensation in the current iteration is as follows:

[0018] In the formula: —The secondary distance curvature compensation amount in the current iteration; —Azimuth slow time with the center of the synthetic aperture as the zero point; —Airborne platform flight speed, unit: m / s; —The estimated angle of view in the current iteration; —Scene center reference slant distance; Step 22: Convert the secondary distance curvature compensation amount in the current iteration into a distance-to-pixel offset amount:

[0019] In the formula: —The distance to the pixel offset in the current iteration; —Distance sampling rate; —Speed ​​of light; Step 23: Combining the distance-to-pixel offset, perform backward mapping linear interpolation on the enhanced image obtained in Step 1 to obtain the distance curvature compensation image in the current iteration:

[0020] In the formula: —Distance curvature compensation image in the current iteration.

[0021] Step 3, Effective Structural Point Extraction: The distance curvature compensation image in the current iteration is normalized to its maximum value, and then adaptive thresholds are applied sequentially. The process involves local density constraint filtering and connected component screening to obtain effective structure points. When the number of effective structure points meets the preset threshold, the process proceeds to step 4; otherwise, the iteration ends.

[0022] Step 3 includes the following sub-steps: Step 31, compensating for distance curvature in the current iteration. After normalizing the maximum value, we get:

[0023] In the formula: —The normalized distance curvature compensation image in the current iteration; —A very small positive number, used to avoid the denominator being equal to 0. In this invention, it is taken as... ; Normalized distance curvature compensation image in the current iteration All pixel values ​​are sorted in ascending order, and the quantile function in MATLAB is used to extract the values. The adaptive threshold is calculated from the 80% sample quantile of the pixel value amplitude. ,Right now Greater than or equal to 80% of the pixel values; furthermore, to prevent adaptive thresholding when scene echo is extremely weak or extremely strong. Values ​​too close to 0 or 1 result in a binary mask that is almost entirely white or black, making it impossible to extract effective structural points. This invention addresses this issue by... Limited to Within the range; Step 32, adopt an adaptive threshold. Normalized distance curvature compensation image in the current iteration Binary segmentation is performed to obtain the initial mask:

[0024] Then the initial mask Local density-constrained filtering is performed as follows: For the initial mask... For each pixel in the mask, count the number of pixels with a value of 1 in its surrounding 9×9 neighborhood window. If this number is less than 18% of the total number of pixels in the window, the point is considered isolated noise and is set to 0; otherwise, it is kept as 1. Then, use the bwareaopen function in MATLAB to remove regions with a connected component area of ​​less than 50 pixels to obtain the final mask. Pixels with a value of 1 in the final mask are defined as valid structure points.

[0025] Step 33: When the number of valid structural points is less than 10, the distance curvature compensation image in the current iteration is considered to lack sufficient structural information, and the current iteration is considered invalid and the iteration terminates.

[0026] Step 4, Energy-weighted Hough Direction Detection: Determine the search range of the Hough transform based on the estimated angle of view in the current iteration; within the search range of the Hough transform, calculate the angle response function value based on the amplitude of the effective structure points, and determine the Hough detection angle based on the angle response function value.

[0027] Step 4 includes the following sub-steps: Step 41, let the effective structural point be the first... The pixel coordinates of the points are The amplitude is The Hough function in MATLAB is used to detect the slope of lines in the image domain, and the search range of the Hough transform is determined based on the estimated angle of view in the current iteration. The specific steps are as follows: From the geometric relationship between the oblique angle and the slope of the line in the image domain, we can know that:

[0028] In the formula: —Slope of a straight line in the image domain; —Airborne platform flight speed, unit: m / s; —Distance sampling rate, in Hz; —Speed ​​of light; —Pulse repetition frequency, unit: Hz; In the Hough parameter space, the slope of the line in the image domain With Hough detection angle satisfy:

[0029] This determines the search range of the Hough transform:

[0030] In the formula: —Angle search step size; Step 42, within the search range of the Hough transform, the first... i Candidates Project each effective structure point onto the Hough parameter space. The axis is used to obtain the projection value:

[0031] In the formula: —No. Projecting each effective structural point to The projection value of the axis, , The number of valid structural points; where i Represents the sequential index of candidate angles within the search range of the Hough transform; —No. The distance and azimuth pixel coordinates of each effective structural point; Step 43, the first i The projection values ​​of all valid structural points corresponding to each candidate angle Rounding and shifting yields the discrete element index for each valid structural point:

[0032]

[0033]

[0034] In the formula: —Rounding function; —for the first One effective structural point, The rounding function; —Among all effective structural points, the smallest .

[0035] Step 44, for the i-th candidate angle The projected energy of a discrete element is obtained by summing the magnitudes of valid structure points with the same discrete element index. ,in Index the discrete elements of these effective structural points; then use a sliding window of length 3 to project the energy. Summing yields ;

[0036] Projected energy outside the index range is recorded as 0; Construct the first [structure] based on the summation result. i Candidates Angular response function:

[0037] Step 45: Traverse all candidate angles, calculate the angle response function value for each candidate angle, and select the maximum value. Step 46: Perform a three-point parabolic fitting on the candidate angle corresponding to the maximum value and its two adjacent candidate angles to obtain the Hough detection angle with sub-pixel accuracy. The specific steps are as follows: Let the candidate angle corresponding to the maximum value and the angular response values ​​of the two adjacent candidate angles be arranged in ascending order as follows: Their corresponding candidate angles are as follows: ; Calculate the sub-pixel correction amount as Hoff detection angle ; In the formula: This is the step size for angle search.

[0038] Step 5: Slant view inversion and update: Calculate the current slope of the line in the image domain using the Hough detection angle; determine whether the current iteration number n has reached the preset sliding window length. If yes, record the estimated value of the slant view in the current iteration and proceed to step 6; otherwise, update the estimated value of the slant view and return to step 2.

[0039] Step 5 includes the following sub-steps: Step 51, based on the Hough detection angle The slope of the current image domain line is calculated. :

[0040] Step 52: Determine whether the current iteration count n has reached the preset sliding window length. (Pick If the angle of view is positive, record the estimated angle of view in the current iteration and proceed to step 6; otherwise, construct the estimated angle of view for the next iteration. And use it as the oblique angle estimate in the current iteration, then return to step 2;

[0041] In the formula: —The estimated angle of view in the next iteration; —Pulse repetition frequency, unit: Hz; Preferably, the input of the arcsine function in the above formula is restricted to... Inside.

[0042] Step 6: Sliding window mean stabilization decision: Output the mean of the oblique angle estimate and the slope of the current image domain line.

[0043] Step 6 includes the following sub-steps: Step 61, calculate the nearest The mean of the oblique angle estimates obtained in the next iteration:

[0044] In the formula: —The mean of the oblique angle estimates in the current iteration; —No. The estimated oblique angle in the next iteration, where Number the number of the most recent W times; Step 62, determine ( If the condition is met, the angle estimation sequence is considered to have entered a stable state. At this point, the mean of the oblique angle estimation values ​​obtained in step 61 is output. and the current slope of the line in the image domain Otherwise, return to step 2.

[0045] The feasibility of the method of this invention is based on the following mechanism: the range curvature in the echo envelope of a large-angle SAR target is determined by the second-order term of the slant-range history, and its curvature decreases monotonically as the estimated slant-angle view approaches the true value. Therefore, by using the current angle estimate to compensate for the curvature in reverse, the envelope trajectory can gradually approach a straight line; and the closer the trajectory is to a straight line, the more stable the Hough transform is in detecting the principal direction. This closed-loop process of compensation, straightening, detection, and updating has self-correcting properties: even if there is a large deviation in the initial angle, the compensation amount in the first few iterations is not accurate, but it can still partially reduce the curvature, so that the next round of Hough detection obtains a result closer to the true direction, thereby gradually converging to the correct value.

[0046] Example 1: Simulation Verification of Point Targets Simulation data was generated using X-band airborne large-slant-look SAR, with a carrier frequency of 10 GHz, a bandwidth of 100 MHz+, and a platform speed of 150. The oblique distance from the center of the scene is 3000. The true value of the oblique angle is 50.0. . Set to 0 (deviation 50) ), sliding window length Stability threshold The method of this invention is used to estimate the oblique angle.

[0047] The estimation results are as follows Figure 1 As shown, the algorithm reaches the sliding window mean stability condition before the 10th iteration, and the final estimate is 50.038. The estimation error is only 0.038. Substituting this oblique angle into the two-dimensional wavenumber domain Imaging algorithm, to obtain Figure 2 The focused image shown shows that all five point targets in the scene are well focused, and the imaging positions are consistent with the theoretical coordinates.

[0048] Example 2: Processing of Measured Airborne SAR Data Using measured data from an airborne X-band SAR system, the range axis is 4096 points and the azimuth axis is 16384 points. The Doppler center was estimated using the method of this invention, the traditional Hough transform method, and the energy centroid method, respectively. Focusing was then performed using the same motion compensation and PPFA imaging procedure.

[0049] The comparison results of global image quality indicators are shown in Table 1. The method of this invention focuses after 26 iterations, and the residual Doppler center is... 0.0998Hz, the absolute value is much smaller than that of the traditional Hough method. The image sharpness is improved by approximately 6.73% compared to the traditional Hough method, and the peak-to-average power ratio is increased to 65.6479 dB.

[0050] Table 1 Comparison Results of Global Image Quality Indicators

[0051] Table 2 shows the comparison results of the average focusing quality index of local strong scattering points. The method of this invention outperforms the two comparative methods in all four indices: 2D PSLR, range PSLR, range ISLR, and azimuth PSLR. For 2D ISLR and azimuth ISLR, the differences between the three methods are small and close to the optimal results.

[0052] Table 2 Comparison of Average Focusing Quality Indicators for Locally Strong Scattering Points

[0053] The focused imaging results of the method of the present invention are as follows: Figure 3 As shown, the main features, edge contours, and texture information in the scene are clearly identifiable.

Claims

1. A large squint SAR imaging method for accurate estimation of the squint angle, characterized in that, Includes the following steps: Step 1: Acquire 2D SAR echo data and perform range compression to obtain an amplitude image; perform contrast enhancement on the amplitude image to obtain an enhanced image; set the initial oblique angle estimation value. Step 2: Construct a secondary distance curvature compensation value from the oblique angle estimation value in the current iteration and convert it into a distance-to-pixel offset; Combine the distance-to-pixel offset with the enhanced image obtained in Step 1 and perform backward mapping linear interpolation to obtain the distance curvature compensation image in the current iteration. Step 3: Normalize the maximum value of the distance curvature compensation image in the current iteration, and then process it sequentially using adaptive thresholding, local density constraint filtering and connected component filtering to obtain effective structure points; when the number of effective structure points meets the preset threshold, proceed to step 4, otherwise the iteration ends. Step 4: Determine the search range of the Hough transform based on the estimated angle of view in the current iteration; within the search range of the Hough transform, obtain the angle response function value based on the magnitude of the effective structure point, and determine the Hough detection angle based on the angle response function value; Step 5: Calculate the current slope of the line in the image domain using the Hough detection angle; determine whether the current iteration number has reached the preset value. If yes, record the estimated angle of view in the current iteration and proceed to step 6; otherwise, update the estimated angle of view and return to step 2. Step 6: Output the mean of the oblique angle estimation values ​​and the current slope of the line in the image domain.

2. The large squint SAR imaging method for accurate estimation of the squint angle as described in claim 1, characterized in that, Step 4 includes the following sub-steps: Step 41: Determine the search range of the Hough transform based on the estimated oblique angle value in the current iteration; Step 42, for the Hough transform within the search range of the th i Each candidate angle is projected onto the Hough parameter space to obtain the projection value; Step 43, the first i The projection values ​​of all valid structural points corresponding to each candidate angle are rounded and shifted to obtain the discrete unit index of each valid structural point. Step 44, for the first i For each candidate angle, the magnitudes of valid structure points with the same discrete element index are summed to obtain the projected energy of that discrete element. Then, a sliding window is used to sum the projected energy, and the summation result is used to construct the first... i Angle response function for each candidate angle; Step 45: Traverse all candidate angles, calculate the angle response function value for each candidate angle, and select the maximum value. Step 46: Perform a three-point parabolic fitting on the candidate angle corresponding to the maximum value and its two adjacent candidate angles to obtain the Hough detection angle with sub-pixel accuracy.

Images