GPU-accelerated synthetic aperture radar backprojection in CUDA

Fasih, Ahmed; Hartley, Timothy D. R.

doi:10.1109/radar.2010.5494395

Cited by 37 publications

(32 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The interpolation step for the 1D NER-NUFFT has been deeply described in [11], while that for the FFT + interpolation approach has been implemented by the same CUDA kernel, exploiting different device functions corresponding to the different considered interpolators. It should be noticed that, although nearest and linear interpolators exploit the texture memory [30,45], this solution has not been considered here to avoid further losses of accuracy [48].…”

Section: F P(: K) Stands For the Values Of S(f τ K ) Q Is The Projmentioning

confidence: 99%

“…Graphics Processing Units (GPUs), which provide platforms for parallel computing with very competitive "flops per dollar" ratios [23][24][25][26][27][28], represent an excellent opportunity in this framework. The use of GPUs is spreading over the SAR community [10,11,[29][30][31][32][33][34] whose first attempts in this context date even back to the times when extensions of the ANSI C simplifying the programming of graphic cards (as CUDA -Compute Unified Device Architecture -or OpenCL) [35] were not yet available [36].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Fast Gpu-Based Interpolation for Sar Backprojection

Capozzoli¹,

Curcio²,

Liseno³

2013

PIER

View full text Add to dashboard Cite

Abstract-We introduce and discuss a parallel SAR backprojection algorithm using a Non-Uniform FFT (NUFFT) routine implemented on a GPU in CUDA language.The details of a convenient GPU implementation of the NUFFT-based SAR backprojection algorithm, amenable to further generalizations to a multi-GPU architecture, are also given.The performance of the approach is analyzed in terms of accuracy and computational speed by comparisons to a "standard", parallel version of the backprojection algorithm exploiting FFT + interpolation instead of the NUFFT. Different interpolators have been considered for the latter processing scheme. The NUFFT-based backprojection has proven significantly more accurate than all the compared approach, with a computing time of the same order. An analysis of the computational burden of all the different steps involved in both the considered approaches (i.e., standard and NUFFT backprojections) has been also reported.Experimental results against the Air Force Research Laboratory (AFRL) airborne data delivered under the "challenge problem for SAR-based Ground Moving Target Identification (GMTI) in urban environments" and collected over circular flight paths are also shown.

show abstract

Section: F P(: K) Stands For the Values Of S(f τ K ) Q Is The Projmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Fast Gpu-Based Interpolation for Sar Backprojection

Capozzoli¹,

Curcio²,

Liseno³

2013

PIER

View full text Add to dashboard Cite

show abstract

“…The processing time is generally an order of magnitude greater than Fourier based methods; however, the framework of backprojection allow it to be easily run on massively parallel processors like the NVIDIA Graphics Cards. There have been other developments of GPU based backprojection implementations such as [5]. Our implementation was developed independently from [5] and is optimized for accuracy and efficiency.…”

Section: Backprojection Processingmentioning

confidence: 99%

“…There have been other developments of GPU based backprojection implementations such as [5]. Our implementation was developed independently from [5] and is optimized for accuracy and efficiency.…”

Section: Backprojection Processingmentioning

confidence: 99%

Improved processing of the casie SAR data

Stringham

Long

2011

2011 IEEE International Geoscience and Remote Sensing Symposium

View full text Add to dashboard Cite

In the summer 2009 NASA Characterization of Arctic Sea Ice Experiment (CASIE09), the microASAR, a small LFM-CW SAR, was operated on the NASA Sierra unmanned aerial system (UAS). An overview of the microASAR and its role in CASIE09 are described in [1,2]. While the limitations in the motion measurements stored with the microASAR data during the CASIE09 mission originally precluded full motion compensation, motion data collected for other CASIE sensors can be employed to improve the SAR image focus and calibration. This paper describes the methodology developed to time-align this motion data and applies this data along with other algorithm improvements to the processing of the microASAR data. This paper also presents a concise description of the backprojection image processing used.

show abstract

“…The trend continues to signal processing where graphics processing units (GPUs) with hundreds of processing cores are now widely used for applications other than graphics processing. With the release of the NVIDIA CUDA (Compute Unified Device Architecture) software interface in 2007, the GPU has become more openly exploited in signal processing [2]. No longer solely used for a specific graphical application, the general purpose GPU (GPGPU) is made up of hundreds of individual processing cores that typically share memory resources making it favorable for many parallel systems.…”

Section: Introductionmentioning

confidence: 99%

Near real-time simultaneous range and velocity processing in a random noise radar

Thorson

Akers

2012

2012 IEEE Radar Conference

View full text Add to dashboard Cite

Abstract-The Air Force Institute of Technology (AFIT) has developed an ultrawideband (UWB) random noise radar (RNR) that operates in the UHF band. The large fractional bandwidth and lack of phase coherency within the AFIT RNR necessitate an alternative to classical Doppler processing for velocity estimation. A time-domain approach to simultaneous range and velocity processing has been implemented in the AFIT RNR but results in significant memory requirements and impractical processing times. This paper presents a method to reduce the twodimensional (2D) processing time by over an order of magnitude by segmenting the signal processing algorithm in order to parallelize on a graphics processing unit (GPU).

show abstract

GPU-accelerated synthetic aperture radar backprojection in CUDA

Cited by 37 publications

References 8 publications

Fast Gpu-Based Interpolation for Sar Backprojection

Fast Gpu-Based Interpolation for Sar Backprojection

Improved processing of the casie SAR data

Near real-time simultaneous range and velocity processing in a random noise radar

Contact Info

Product

Resources

About