Study For Airborne Synthetic Aperture Radar(SAR) Processing Under Unstable Trajectory

Synthetic aperture radar(SAR)plays an important role in remote sensing field due to its all-time and all-day compared to optical camera.In recent years,airborne SAR has been widely used in fields like rescue,relief,target detection and weapon guidance benefit from the flexible platform.The topic of the thesis is motion compensation(MOCO)and image mosaic in airborne SAR processing.An improved data-based correction for spatially-variant motion errors,signal-based correction for spatially-variant motion errors,and two robust stitching approaches were proposed for unstable air-platform with unexpected trajectory,which may support the system or processor design.The main study and innovations of our works can be summarized as follows:1.Data-based airborne SAR spatially-variant motion errors compensation: An improved algorithm was proposed based on range-Doppler expansion to correction the spatially-variant motion errors effectively.At first,based on the solution of inertial navigation system(INS)measurement data,a novel airborne SAR geometry is developed in navigation coordinates system,so that the real motion status of platform can be described accurately and directly.Secondly,the range-Doppler expansion is derived to eliminate the cross-coupled echo phase errors caused by motion errors,so that chirp-z transform(CZT)can be performed to correct the spatially-variant errors.Thirdly,the computational load and residual phase errors are discussed.Simulation and real-data processing results demonstrate the performance of our algorithm.In our approach,the cross-coupled of phase errors is concerned instead of neglected by the classic two-step MOCO especially with significant motion errors.The scenario variant of motion errors along-and cross-track are both corrected,therefore the depth of filed is increased.Meanwhile,the computational burden of our approach is far less than fast factorized back-projection algorithm(FFBPA),which means it is more suitable for real-time processing.2.Signal-based airborne SAR spatially-variant motion errors compensation: Subject to the payload and cost of drones,it is impossible to employ a high-precision INS on unmanned aerial vehicle(UAV)SAR to implement MOCO.Therefore,an improved spatially-variant motion errors compensation based on weighted map-drift autofocus(MDA)is proposed.First of all,the depth of field impacted by measurement error and scenario variant are analyzed by means of the classical UAV SAR imaging geometry with motion errors.Then,the weighted MDA chirp rate estimation is developed to get accurate non-linear phase errors(NPE)with overlapped sub-aperture(OSA)strategy.A series of NPE equations corresponding to motion errors are constructed from each range block,and random sample consensus(RANSAC)iteration is designed to ensure the accuracy and robustness of solution.Thus,the spatially-variant motion errors can be decoupled by improved stolt interpolation and further corrected by CZT.The proposed algorithm performance can be demonstrated by simulation and real-data processing results.Our approach can be used for UAV SAR imaging with large measurement errors of INS.Compared to the traditional two-step MOCO based on MDA,the cross-coupled spatially-variant motion errors are corrected effectively without help of accurate motion data-based compensation.In addition,the performance of MDA is enhanced in our approach so that it can be expanded to lots of applications.3.Non-ideal airborne SAR image robust mosaic: In order to increase the performance of image mosaic for non-ideal airborne SAR,the thesis proposes two improved approaches based on cross-correlation filter and feature-extracted,respectively.First of all,non-ideal airborne SAR image mosaic geometries of strip-map and spotlight mode are constructed,respectively.Different with the traditional geometries,the effect of non-ideal condition is considered during modeling,and the geometric relationship between two SAR images is analyzed through image transformation.Then:1)An improved approach based on cross-correlation filter is proposed to implement strip-map SAR image stitching.Firstly,gray histogram equalization is performed as preprocessing to enhance SAR image.Then,select a group of features to construct the united-features description for each pixel on one image,which is mainly through the gray scale of finite surrounded area.Thirdly,find certain extremum points of united-features function and select the corresponding area as template to match with another image.Based on the developed non-ideal geometry of strip-map airborne SAR,the transformation matrix can be solved,thus two images can be registered and further stitched.Different with traditional simple cross-correlation just along azimuth direction,both offsets along and cross track caused by motion errors can be estimated,which indicates that the accuracy and robustness of the proposed algorithm is increased.2)An improved approach based on features-extracted is proposed to realize spotlight SAR image stitching.Firstly,two spotlight images are detected based on speeded-up robust features(SURF).By means of SURF,features and its description vectors are extracted.Secondly,these features extracted from two images are matched based on the principle that the Euclidean distance between description vectors reaches minimum.Then,the homography matrix is solved with the help of least square method so that the two spotlight images can be corrected and stitched.In our approach,SURF is adopted to describe the complicated geometric relationship between two spotlight images in actual,including scale,rotation,and even projection.Compared to traditional mosaic based on affine transformation,homography is applied to correct the projection distortion between different imaging planes varying with squint angle,which means that the proposed approach is more suitable for non-ideal airborne SAR working in spotlight mode.Lastly,real SAR images stitching experiments demonstrates the performance of the proposed methods.