UAV-borne Video SAR Imaging And GPU Implementation

As a newly proposed concept,video SAR has the advantages of high resolution and high frame rate imaging characteristics compared to traditional SAR.The high frame rate feature allows continuous multi-frame imaging of the ground scene,while the high resolution feature allows more accurate target information to be obtained.The imaging frame rate of video SAR is directly proportional to the operating frequency of the radar,and the imaging frame rate of video SAR can be increased by increasing the operating frequency or subaperture overlap rate.Because video SAR platforms generally use small platforms such as drones,and the system carrier frequency is high,their imaging processing is particularly sensitive to platform positioning errors,and is more susceptible to atmospheric disturbances,resulting in greater motion errors,which brings about motion compensation.A big challenge.In view of the large amount of calculation caused by the point-by-point traversal and interpolation operations of the current video SAR imaging algorithm,this paper proposes a video SAR imaging method based on Dechirp reception.This method uses the RD algorithm combined with motion compensation to directly construct the existing The function does not require complicated operations such as interpolation,the calculation amount is small,and the processing efficiency is high.For this method,a video SAR signal model is first established,data preprocessing is performed,Fourier transform and variable substitution are performed on the echo signal received by Dechirp,so as to obtain the same data form as the traditional direct acquisition signal,which is convenient for imaging algorithms The use of;secondly,a sub-aperture-based MD method is used to estimate the residual motion error of the full aperture;finally,two image geometric correction methods based on azimuth scaling and Chirp-Z transform are used to achieve adaptive geometric correction of video SAR images.Finally,imaging is carried out in a unified coordinate system.Since the accuracy of the navigation system cannot meet the high-precision motion compensation requirements of the airborne video SAR,and the high-precision motion error estimation of the traditional PGA algorithm depends on the quality of the distinctive points in the imaging scene,for special scenes such as grass and desert,PGA estimation Performance is greatly reduced.In response to this problem,this paper adopts a sub-aperturebased MD method,and uses the MD method to estimate the phase error for each sub-aperture.Finally,the phase errors of each sub-aperture are stitched to obtain the full-aperture error and realize the full-aperture residual.The estimation of motion error improves the robustness of motion error estimation.The high frame rate characteristics of video SAR require that continuous multi-frame SAR images can be formed in a short period of time during imaging processing.As the resolution requirements increase,the amount of data during imaging processing will become very large,which is very important for real-time imaging.Put forward higher requirements.Traditional CPU imaging platforms based on serial processing methods are difficult to guarantee realtime imaging,while heterogeneous parallel computing platforms based on CPU+GPU can accelerate the implementation of imaging algorithms,reduce data processing time,improve processing efficiency,and meet real-time video SAR imaging processing.In order to meet the requirements,this paper proposes a parallel estimation method of modulation frequency combining MD method and contrast method,which improves the speed and accuracy of modulation frequency estimation,and improves the efficiency of video SAR imaging algorithm as a whole.