Fast registration method for data-clustering video SAR

Abstract

The application discloses a data bunching type video SAR fast registration method, and belongs to the technical field of radar imaging. The method utilizes SAR imaging characteristics, and decomposes the affine transformation required by SAR image registration into rotation transformation and translation transformation. The method can well utilize prior information of an inertial navigation system and echo data, and improves the calculation efficiency of SAR image registration. In addition, the method only involves simple arithmetic operation and does not involve complex calculation operation such as a sliding window, so that the implementation manner of the application is simple.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of radar imaging, and particularly relates to a data-bunching type video SAR fast registration method. BACKGROUND

[0002] Video Synthetic Aperture Radar (Video SAR) is an advanced radar imaging technology, which aims to generate a sequence of continuous high-resolution radar images similar to video, and can capture continuous images in the time dimension, thereby realizing real-time monitoring and dynamic observation of targets. However, due to platform motion, imaging angle change and environmental factors, the images obtained continuously may have spatial inconsistency, which makes image registration play a key role in Video SAR. Moreover, high frame rate, high resolution, and real-time performance have become a major trend in the development of Video SAR, and the realization of high frame rate, high resolution, and real-time performance of Video SAR faces great technical challenges, including huge data processing amount and high real-time requirement. Therefore, it is of great significance to research a fast image registration method.

[0003] SAR image registration specifically refers to the process of geometrically correcting two or more images obtained under different conditions, so as to finally make the reference image and the image to be registered in the same coordinate system. The Video SAR system can obtain a series of SAR image sequences in time sequence, so a certain frame in the SAR image sequence can be selected as a reference frame, and the SAR image sequences of the remaining frames to be registered are unified to the spatial coordinate system of the reference frame. The existing SAR image registration methods are all first extracting the feature information of the reference image and the image to be registered, and then obtaining the transformation matrix according to the feature information. From the feature detection, the SAR image registration method can be divided into two categories: a registration method based on region information and a registration method based on feature matching. The registration method based on region information needs to perform intensive calculation in the entire image or selected region, usually involves image correlation measurement, image transformation and interpolation, global search and optimization, has the characteristics of high registration accuracy and high calculation cost, and is difficult to meet the processing time requirement of Video SAR. The registration method based on feature matching only needs to process the key regions in the image, which improves the operation speed while reducing the registration accuracy, especially when the radar viewing angle changes, the image resolution is low, and the texture information is blurred, the registration accuracy will be seriously reduced, and even mismatched. Therefore, in the application scenario of Video SAR, there is an urgent need for a SAR image registration method which meets the registration accuracy, and also has the characteristics of high processing speed and strong robustness. SUMMARY

[0004] The application aims to overcome the defects of the prior art, provide a SAR image registration method with high processing speed and strong robustness while meeting registration accuracy in a video SAR application scenario, and specifically provide a data-bundling-based video SAR fast registration method.

[0005] The technical problem of the application is solved in the following manner:

[0006] A data-bundling-based video SAR fast registration method comprises the following steps:

[0007] Step 1: convert the SAR imaging radar position coordinates (lon, lat, alt) in the WGS-84 coordinate system into point (X, Y, Z) in the geocentric geodetic rectangular coordinate system;

[0008] Step 2: select any frame of SAR original echo signal as a reference frame, and take the center position coordinates of the SAR imaging radar corresponding to the reference frame in the geocentric geodetic rectangular coordinate system as the reference frame imaging center; take the reference frame imaging center as the coordinate origin, and convert the SAR imaging radar position coordinates in the geocentric geodetic rectangular coordinate system and the position coordinates of the set scene center in the geocentric geodetic rectangular coordinate system to the north-east-geodetic coordinate system;

[0009] Step 3: generate a rotation matrix using the roll angle, pitch angle and yaw angle corresponding to the reference frame imaging center, and convert the coordinates of the set scene center and the SAR imaging radar in the north-east-geodetic coordinate system to the imaging coordinate system;

[0010] Step 4: the included angle θ between the line-of-sight direction of any position of the radar during movement and the line-of-sight direction of the reference frame imaging center is:

[0011]

[0012] wherein, is the displacement vector from the reference frame imaging center to the set scene center, is the displacement vector from the imaging center of any frame of the SAR imaging radar during the movement of the radar to the set scene center, and ||| represents a modulo operation.

[0013] Step 5: correct the viewing angle of the SAR imaging radar in different azimuth directions to the azimuth viewing angle in the imaging coordinate system of the reference frame by means of two-dimensional interpolation, take the SAR two-dimensional image of each frame as the imaging scene data V to be interpolated, perform two-dimensional interpolation processing on the imaging scene data V, and obtain the interpolated imaging scene data, so that the spatial frequency distribution of the imaging scene data is (X a , Ya ) are converted to (X q ,Y q ), (X a ,Y a ) respectively represent the spatial frequency distribution of the imaging scene before interpolation, (X q ,Y q ) respectively represent the spatial frequency distribution of the imaging scene after arbitrary frame view correction (after rotation);

[0014] Step 6: The correlation of the SAR imaging echo data is used to correct the offset respectively in the range frequency domain and the azimuth time domain and the azimuth frequency domain and the range time domain;

[0015] When the range offset correction is performed, the SAR original echo signals of the reference frame and the frame to be corrected are subjected to Fourier transform in the range direction to obtain the SAR signal data of the reference frame and the frame to be corrected in the range frequency domain and the azimuth time domain; the range vector f ref of a certain azimuth direction in the SAR signal data of the reference frame in the range frequency domain and the azimuth time domain is conjugated and multiplied by the range vector f any of each azimuth direction in the SAR signal data of the frame to be corrected in the range frequency domain and the azimuth time domain, and then inverse Fourier transform is performed to obtain the correlation coefficient vector corr τ of the range offset, and the offset offset τ corresponding to the maximum value in the correlation coefficient vector is selected as the range offset correction amount;

[0016] Similarly, when the azimuth offset correction is performed, the SAR original echo signals of the reference frame and the frame to be corrected are subjected to Fourier transform in the azimuth direction to obtain the SAR signal data of the reference frame and the frame to be corrected in the azimuth frequency domain and the range time domain; the azimuth vector f tref of a certain range direction in the SAR signal data of the reference frame in the azimuth frequency domain and the range time domain is conjugated and multiplied by the azimuth vector f tany of each range direction in the SAR signal data of the frame to be corrected in the azimuth frequency domain and the range time domain, and then inverse Fourier transform is performed to obtain the correlation coefficient vector corr t of the azimuth offset, and the offset offset t corresponding to the maximum value in the correlation coefficient vector is selected as the azimuth offset correction amount;

[0017] Step 7: The imaging scene data after interpolation is circularly shifted by offset τ units of length in the range direction to complete the range offset correction; similarly, the imaging scene data after interpolation is circularly shifted by offset t units of length in the azimuth direction to complete the azimuth offset correction.

[0018] Further, in step 1, the conversion relationship between the WGS-84 coordinate system and the ECEF rectangular coordinate system is shown as follows:

[0019]

[0020] wherein, lon, lat and alt respectively represent the longitude, latitude and altitude of the SAR imaging radar; X, Y and Z respectively represent the x-axis, y-axis and z-axis coordinates of the SAR imaging radar in the ECEF coordinate system; N is the curvature radius of the reference ellipsoid, e is the eccentricity of the WGS-84 coordinate system, and a is the long radius of the reference ellipsoid.

[0021] Further, in step 2, the central position coordinates of the SAR imaging radar corresponding to the reference frame in the ECEF rectangular coordinate system are denoted as Let the position coordinates of the SAR imaging radar in the ECEF rectangular coordinate system be x ECEF ,y ECEF ,z ECEF , and let the position coordinates of the set scene center in the ECEF rectangular coordinate system be The position coordinates of the SAR imaging radar position coordinates in the NED coordinate system and the position coordinates of the set scene center in the NED coordinate system are respectively denoted as:

[0022]

[0023]

[0024] wherein, N x , E y and D z respectively represent the position coordinates of the SAR imaging radar position coordinates in the N-axis, E-axis and D-axis in the NED coordinate system; N a , E a and D a respectively represent the position coordinates of the set scene center in the N-axis, E-axis and D-axis in the NED coordinate system.

[0025] S is the coordinate transformation matrix from the ECEF rectangular coordinate system to the NED coordinate system, and is denoted as:

[0026]

[0027] wherein, lon0, lat0 and alt0 respectively represent the longitude, latitude and altitude of the imaging center of the reference frame.

[0028] Further, in step 3, the rotation matrix T corresponding to the roll angle θ x , the pitch angle θ y and the yaw angle θ z is denoted asx , T y and T z respectively represent:

[0029]

[0030] The SAR imaging radar in the imaging coordinate system and the position coordinates (x, y, z) and (x a , y a , z a ) of the set scene center are respectively:

[0031]

[0032]

[0033] Further, in step 5, (X a , Y a ) and (X q , Y q ) have the following relationship:

[0034]

[0035] Wherein, c is the speed of light, f c is the center frequency of the radar working, f is the distance frequency, is the pitch angle of the current frame imaging center of the SAR imaging radar, and φ is the instantaneous azimuth angle of the phase center of the SAR imaging radar antenna.

[0036] The beneficial effects of the present application are:

[0037] The method of the present application realizes relatively simple implementation, reduces algorithm complexity, and improves calculation efficiency by splitting the affine transformation into rotation transformation and translation transformation for image correction according to the characteristics of radar imaging at different viewing angles. The method of the present application can complete efficient image registration for SAR images with unobvious features while ensuring the registration accuracy of the SAR images. BRIEF DESCRIPTION OF DRAWINGS

[0038] Figure 1 is a flowchart of the method of the present application;

[0039] Figure 2 is a SAR image in the reference frame in the method of the embodiment;

[0040] Figure 3 is the 55th frame SAR image before correction in the method of the embodiment;

[0041] Figure 4This refers to the 55th frame of the SAR image to be corrected after rotation transformation in the method described in the embodiment.

[0042] Figure 5 The image shown is the 55th frame of the SAR image to be corrected after rotation and translation transformations as described in the embodiment. Detailed Implementation

[0043] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0044] This embodiment provides a fast registration method for video SAR based on data clustering, and its flowchart is shown below. Figure 1 As shown, the input is the raw SAR echo signal, which is processed by imaging to obtain a SAR two-dimensional image. The specific implementation includes the following steps:

[0045] Step 1: Transform the WGS-84 coordinate system (ALL) to the Earth-centered, Earth-fixed Cartesian coordinate system (ECEF). The SAR imaging radar position coordinates (lon, lat, alt) in the ALL coordinate system are converted to points (X, Y, Z) in the ECEF coordinate system. The transformation relationship is as follows:

[0046]

[0047] Where lon, lat, and alt represent the longitude, latitude, and elevation of the SAR imaging radar, respectively; X, Y, and Z represent the x-axis, y-axis, and z-axis coordinates of the SAR imaging radar in the ECEF coordinate system, respectively; and the eccentricity of the ellipsoid in the WGS-84 coordinate system is e = 8.1819190842622. -2 The radius of curvature of the reference ellipsoid The major radius of the reference ellipsoid is a = 6378137m.

[0048] Step 2: Select any frame of the raw SAR echo signal as the reference frame, and determine the center position coordinates of the SAR imaging radar corresponding to the reference frame in the ECEF coordinate system. Using the reference frame imaging center as the origin, the SAR imaging radar position coordinates (x, y, y) of all frames in the ECEF coordinate system are respectively... ECEF ,y ECEF ,z ECEF And set the position coordinates of the scene center in the ECEF coordinate system. Transform to the NED coordinate system (Northeastern Geodetic Dimension):

[0049]

[0050] Where, N x E y and D zrepresents the position coordinates of the SAR imaging radar position coordinates in the North-East-Down coordinate system respectively; N a represents the position coordinates of the SAR imaging radar position coordinates in the North-East-Down coordinate system respectively; N a represents the position coordinates of the SAR imaging radar position coordinates in the North-East-Down coordinate system respectively; N a represents the position coordinates of the SAR imaging radar position coordinates in the North-East-Down coordinate system respectively; N represents the transformation matrix from the ECEF coordinate system to the NED coordinate system, and is expressed as:

[0051]

[0052] wherein, lon0, lat0 and alt0 represent the longitude, latitude and altitude of the reference frame imaging center respectively.

[0053] Step 3: using the roll angle θ x , pitch angle θ y and yaw angle θ z recorded by the IMU corresponding to the reference frame imaging center to generate a rotation matrix, and convert the coordinates of the set scene center and the SAR imaging radar in the North-East-Down coordinate system to the imaging coordinate system, and is expressed as:

[0054]

[0055] wherein, (x, y, z) and (x a , y a , z a ) are the position coordinates of the SAR imaging radar and the set scene center in the imaging coordinate system respectively; the rotation matrix corresponding to the roll angle, the pitch angle and the yaw angle is as follows:

[0056]

[0057] At this point, the position coordinates of the set scene center and the SAR imaging radar have been converted to the imaging coordinate system with the reference frame imaging center as the origin.

[0058] Step 4: according to the included angle formula of two vectors, the included angle θ between the line of sight direction of the radar at any position in the motion process and the line of sight direction of the reference frame imaging center can be calculated:

[0059]

[0060] wherein, the displacement vector of the reference frame imaging center to the set scene center is the displacement vector of the imaging center (x c , y c ) of any frame of the SAR imaging radar in the motion process of the radar to the set scene center is || || represents the modulus.

[0061] Step 5: The beam domain distribution of SAR imaging radar in different azimuth directions can be regarded as the beam distribution of the reference coordinate system (the imaging coordinate system of the reference frame) after rotation, so the view angle of SAR imaging radar in different azimuth directions can be corrected to the azimuth view angle in the imaging coordinate system of the reference frame by two-dimensional interpolation.

[0062] The two-dimensional interpolation method in the embodiment is implemented by using the two-dimensional interpolation function interp2(X a ,Y a ,V,X q ,Y q ) of matlab, and the SAR two-dimensional image of each frame is taken as the imaging scene data V to be interpolated. The imaging scene data V is subjected to two-dimensional interpolation processing to obtain the interpolated imaging scene data, so that the spatial frequency distribution of the imaging scene data is converted from (X a ,Y a ) to (X q ,Y q ), (X a ,Y a ) respectively represent the imaging scene spatial frequency distribution before interpolation, and (X q ,Y q ) respectively represent the imaging scene spatial frequency distribution after view angle correction (rotation), and there is the following relationship:

[0063]

[0064] The spatial frequency distribution of the center of the imaging scene before interpolation is:

[0065]

[0066] where c is the speed of light, f c is the center frequency of the radar operation, f is the range frequency, is the pitch angle of the imaging center of the current frame of the SAR imaging radar, and φ is the instantaneous azimuth angle of the phase center of the antenna of the SAR imaging radar.

[0067] Step 6: Because the beam center cannot always focus on the set scene center during the movement of the radar, the position offset in the range direction and the azimuth direction occurs.

[0068] The correlation of the focused SAR imaging echo data is used to correct the offset in the range frequency-azimuth time domain and the azimuth frequency-range time domain.

[0069] In the distance offset correction, the SAR original echo signals of the reference frame and the frame to be corrected are subjected to distance Fourier transform to obtain SAR signal data of the reference frame and the frame to be corrected in distance frequency domain and azimuth time domain; the distance vector f ref of a certain azimuth direction in the SAR signal data of the reference frame in distance frequency domain and azimuth time domain is conjugate multiplied with the distance vector f any of a certain azimuth direction in the SAR signal data of the frame to be corrected in distance frequency domain and azimuth time domain, and then inverse Fourier transform is performed to obtain the correlation coefficient vector corr τ of the distance offset; the offset offset τ corresponding to the maximum value in the correlation coefficient vector is selected as the distance offset correction amount.

[0070] Similarly, in the azimuth offset correction, the SAR original echo signals of the reference frame and the frame to be corrected are subjected to azimuth Fourier transform to obtain SAR signal data of the reference frame and the frame to be corrected in azimuth frequency domain and distance time domain; the azimuth vector f tref of a certain distance direction in the SAR signal data of the reference frame in azimuth frequency domain and distance time domain is conjugate multiplied with the azimuth vector f tany of a certain distance direction in the SAR signal data of the frame to be corrected in azimuth frequency domain and distance time domain, and then inverse Fourier transform is performed to obtain the correlation coefficient vector corr t of the azimuth offset; the offset offset t corresponding to the maximum value in the correlation coefficient vector is selected as the azimuth offset correction amount.

[0071] corr τ and corr t The expression is:

[0072] corr t =ifft(f tref ·conj(f tany ))

[0073] corr τ =ifft(f τref ·conj(f τany ))

[0074] Where, ifft represents inverse Fourier transform, and conj represents taking conjugate.

[0075] Step 7: The imaging scene data after interpolation is circularly shifted in the distance direction by offset τ unit length, so that the distance offset correction is completed; similarly, the imaging scene data after interpolation is circularly shifted in the azimuth direction by offset t unit length, so that the azimuth offset correction is completed.

[0076] The method described in the embodiment can obtain the final corrected image through rotation transformation (two-dimensional interpolation) and translation transformation (data migration).

[0077] In the simulation example described in the embodiment, Figure 2 The SAR image of the reference frame is given, and the coordinate system of the image frame will be used as the reference coordinate system for subsequent SAR image registration under different viewing angles.

[0078] Figure 3 The 55th uncorrected SAR image is given, and it can be seen that due to the influence of different viewing angles and beam jitter, the to-be-corrected frame image needs to be rotated and translated compared with the reference frame image.

[0079] Figure 4 The 55th to-be-corrected SAR image after rotation transformation is given, and it can be seen that the viewing angle of the 55th frame has been calibrated to the reference frame viewing angle after rotation transformation.

[0080] Figure 5 The 55th to-be-corrected SAR image after rotation and translation transformation is given, and all steps of the SAR registration method described in the embodiment have been completed. By comparing the reference frame image, it can be seen that the to-be-registered image has been calibrated to the same viewing angle, the same coordinate system, and the same image window is selected for display, which can verify the effectiveness of the method described in the embodiment.

[0081] In summary, from the results of the processing, the method provided by the present application has practical value.

Claims

1. A data-clustering-based video SAR fast registration method, characterized in that, The method comprises the following steps: Step 1: converting the SAR imaging radar position coordinates (lon, lat, alt) in the WGS-84 coordinate system into point (X, Y, Z) in the geocentric and fixed rectangular coordinate system; Step 2: selecting the SAR original echo signal of any frame as a reference frame, taking the center position coordinates of the SAR imaging radar corresponding to the reference frame in the geocentric and fixed rectangular coordinate system as the imaging center of the reference frame; taking the imaging center of the reference frame as the coordinate origin, respectively converting the SAR imaging radar position coordinates in the geocentric and fixed rectangular coordinate system and the position coordinates of the set scene center in the geocentric and fixed rectangular coordinate system into the North-East-Geodetic coordinate system; Step 3: generating a rotation matrix by using the roll angle, pitch angle and yaw angle corresponding to the imaging center of the reference frame, and converting the coordinates of the set scene center and the SAR imaging radar in the North-East-Geodetic coordinate system into the imaging coordinate system; Step 4: the included angle θ between the line-of-sight direction of the radar at any position in the motion process and the line-of-sight direction of the imaging center of the reference frame is: wherein, is a displacement vector of the imaging center of the radar to the center of the scene, is a displacement vector of the imaging center of the SAR imaging radar to the center of the scene in the process of radar motion, and ||| represents a modulo operation. Step 5: Correct the SAR imaging radar's viewing angles in different azimuth directions to the azimuth viewing angles in the imaging coordinate system of the reference frame using two-dimensional interpolation. The SAR two-dimensional images of each frame are used as the imaging scene data V to be interpolated. Perform two-dimensional interpolation on the imaging scene data V to obtain the interpolated imaging scene data, thus changing the spatial frequency distribution of the imaging scene data from (X... a ,Y a ) convert to (X q ,Y q ), (X a ,Y a (X) represents the spatial frequency distribution of the imaged scene before interpolation. q ,Y q ) represent the spatial frequency distribution of the imaging scene after viewpoint correction for any frame; Step 6: using the correlation of the spotlight SAR imaging echo data to respectively correct the offset in the range frequency domain, azimuth time domain and azimuth frequency domain, range time domain; In the distance offset correction, the SAR original echo signals of the reference frame and the frame to be corrected are subjected to distance Fourier transform to obtain SAR signal data of the reference frame and the frame to be corrected in distance frequency domain azimuth time domain; the distance vector f ref of each azimuth direction in the SAR signal data of the reference frame in distance frequency domain azimuth time domain is multiplied by the distance vector f any of each azimuth direction in the SAR signal data of the frame to be corrected in distance frequency domain azimuth time domain in conjugate manner, and then inverse Fourier transform is performed to obtain a correlation coefficient vector corr τ of the distance offset; and the offset offset τ corresponding to the maximum value in the correlation coefficient vector is selected as the distance offset correction amount. Similarly, when the azimuth offset correction is performed, the SAR original echo signals of the reference frame and the frame to be corrected are subjected to Fourier transform in the azimuth direction to obtain SAR signal data of the reference frame and the frame to be corrected in the azimuth frequency domain distance-time domain; the azimuth vectors f tref of each range direction in the SAR signal data of the reference frame in the azimuth frequency domain distance-time domain are conjugate multiplied with the azimuth vectors f tany of each range direction in the SAR signal data of the frame to be corrected in the azimuth frequency domain distance-time domain, and then inverse Fourier transform is performed to obtain the correlation coefficient vector corr t of the azimuth offset; the offset offset t corresponding to the maximum value in the correlation coefficient vector is selected as the azimuth offset correction amount. Step 7: cyclically shift the interpolated imaging scene data in the range direction by offset τ units of length, to complete the range direction offset correction; similarly, cyclically shift the interpolated imaging scene data in the azimuth direction by offset t units of length, to complete the azimuth direction offset correction.

2. The data-clustering-based video SAR fast registration method according to claim 1, wherein, In step 1, the conversion relationship between the WGS-84 coordinate system and the geocentric and fixed rectangular coordinate system is as follows: Wherein, lon, lat and alt represent the longitude, latitude and altitude of the SAR imaging radar respectively; X, Y and Z represent the x-axis, y-axis and z-axis coordinates of the SAR imaging radar in the ECEF coordinate system respectively; N is the curvature radius of the reference ellipsoid, e is the eccentricity of the ellipsoid in the WGS-84 coordinate system, and a is the long radius of the reference ellipsoid.

3. The data-clustering-based video SAR fast registration method according to claim 2, wherein, In step 2, the center position coordinates of the SAR imaging radar corresponding to the reference frame in the Earth-Centered Earth-Fixed rectangular coordinate system are denoted as SAR imaging radar position coordinates in the Earth-Centered Earth-Fixed rectangular coordinate system are denoted as (x ECEF ,y ECEF ,z ECEF ), and the position coordinates of the set scene center in the Earth-Centered Earth-Fixed rectangular coordinate system are denoted as The position coordinates of the SAR imaging radar position coordinates in the North-East-Geodetic coordinate system and the position coordinates of the set scene center in the North-East-Geodetic coordinate system are respectively denoted as: wherein N x , E y and D z represent the position coordinates of the SAR imaging radar position coordinate in the N-axis, E-axis and D-axis of the North-East-Geodetic coordinate system respectively; N a , E a and D a represent the position coordinates of the set scene center in the N-axis, E-axis and D-axis of the North-East-Geodetic coordinate system respectively; S is the coordinate transformation matrix from the geocentric and fixed rectangular coordinate system to the North-East-Geodetic coordinate system, and is expressed as: Wherein, lon0, lat0 and alt0 respectively represent the longitude, latitude and elevation of the imaging center of the reference frame.

4. The data-clustering-based video SAR fast registration method according to claim 3, wherein, In step 3, the roll angle θ x , the pitch angle θ y and the yaw angle θ z corresponding to the rotation matrix T x , T y and T z are respectively expressed as: The SAR imaging radar in the imaging coordinate system and the position coordinates (x, y, z) and (x a ,y a ,z a ) of the set scene center are respectively:

5. The data-clustering-based video SAR fast registration method according to claim 1, wherein, In step 5, (X a ,Y a ) and (X q ,Y q ) have the following relationship: where c is the speed of light, f c is the center frequency of the radar, f is the range frequency, is the elevation angle of the center of the current frame imaging of the SAR imaging radar, and φ is the instantaneous azimuth angle of the phase center of the antenna of the SAR imaging radar.

Images