Please use this identifier to cite or link to this item: http://10.1.7.192:80/jspui/handle/123456789/4775
Title: Implementation of RADAR Algorithm on GPU using CUDA
Authors: Pandya, Bhaumik R.
Keywords: EC 2012
Project Report
Project Report 2012
EC Project Report
EC (ES)
Embedded Systems
Embedded Systems 2012
12MEC
12MECE
12MECE31
Issue Date: 1-Jun-2014
Publisher: Institute of Technology
Series/Report no.: 12MECE31;
Abstract: The science of remote sensing ranges from aerial photography, digital scanners, wide band sensors and the recent advance of Synthetic Aperture Radar (SAR). The processing of initial SAR systems was based on the principles of Fourier Optics. Lenses provided a real time two-dimensional Fourier transform of the data. However the development of SAR processing algorithms made a complete paradigm shift from the optical processing. However with the increase in the demands for high resolution and larger swaths for processing, pressure on the existing systems available for SAR processing increases. The recent advances in General Purpose Graphical Processing Unit(GPGPU) technology and its programmability could suffice the increased processing requirements. Various modes of SAR are discussed and parallel Implementation of algorithms are carried out over the GPGPU operating in near real time and real time modes.In this thesis basically two SAR processing algorithms has been discussed which are Range Doppler and Chirp Scaling Algorithms. Range Doppler Algorithm(RDA) is one of the most widely used in the processing of SAR. This algorithm takes the advantage of the approximate separability of processing in range direction and azimuth direction. The process of Range Cell Migration Correction (RCMC) is done in range time and azimuth frequency domain or called the Range-Doppler domain which gives the algorithm its name.In Chirp Scaling Algorithm(CSA) all the basic operations are performed by multiplication of the RAW data with the filter functions.The Chirp Scaling Algorithm compensates the need of Interpolator in RDA by using scaling method and algorithm becomes Azimuth frequency dependent.In this way it overtakes disadvantages of Range Doppler algorithm.In this thesis two algorithms are implemented on NVIDIA GPU Platform and its simulation results has been discussed.The results shows that RDA takes less time due to its less complex manipulations,but not useful when data is squinted.
URI: http://hdl.handle.net/123456789/4775
Appears in Collections:Dissertation, EC (ES)

Files in This Item:
File Description SizeFormat 
12MECE31.pdf12MECE313.58 MBAdobe PDFThumbnail
View/Open


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.