GPU-accelerated SAR backprojection in JACKET for MATLAB

For airship-mounted ultra-wideband Synthetic Aperture Radar (SAR), backprojection (BP) processing technique is more suitable for azimuth focusing than the algorithms in one or two-dimensional frequency domain. However, the high computational cost limits its application in real world. To conveniently...

Full description

Saved in:
Bibliographic Details
Main Authors: Fulai, Liang, Xiaojiang, Qu, Yanghuan, Li, Qian, Song, Hanhua, Zhang
Format: Conference Proceeding
Language:English
Subjects:
Acceleration
Azimuth
GPU
Graphics processing unit
JACKET for MATLAB
MATLAB
parallel computation
Radar imaging
Synthetic aperture radar
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:For airship-mounted ultra-wideband Synthetic Aperture Radar (SAR), backprojection (BP) processing technique is more suitable for azimuth focusing than the algorithms in one or two-dimensional frequency domain. However, the high computational cost limits its application in real world. To conveniently accelerate BP in Graphic Processing Unit (GPU) the authors present a new implementation manner of BP which we called as Grid-BP. There is only one level FOR-loop in this algorithm. The operation in FOR-loop includes matrix operation and interpolation which is easy for CUDA to cope with. Furthermore, the authors attempt to realizing Grid-BP in JCAKET for MATLAB to avoiding complicated CUDA programme. The result shows that the Grid-BP algorithm accelerated by JACKET on Tesla C1060 GPU attains an acceleration ratio of 40-50.