Parallel implementation of an algorithm for high resolution range profiling using a stepped frequency radar

Mukherjee, Soma

Parallel implementation of an algorithm for high resolution range profiling using a stepped frequency radar

Thesis / Dissertation

2004

Abstract

This thesis describes the implementation and analysis of a frequency domain approach for reconstructing wide-bandwidth high resoultion range profiles using stepped-frequency waveforms. In order to minimize the overall processing time, a parallel algorithm was developed and tested in a homogeneous cluster of computers. In Ultra-wideband synthetic aperture radar (SAR) systems, stepped-frequency waveforms are preferred for their ability to achieve high range resolution without putting burden to severe instantaneous bandwidth and sampling rate requirements. In a stepped-frequency system, the wide bandwidth is reconstructed by transmitting a group of narrow-bandwidth pulses, which are then combined to obtain the wide bandwidth. Several approaches to stepped-frequency processing exist, namely an IFFT method [9], a time domain method [12] and more recently a frequency domain method [23]. The IFFT method is unsuitable for SAR processing because it produces multiple "ghost" targets in high resolution profiles and the time domain method is computationally inefficient. The inefficiency of these two methods led to the development of the fast, computationally efficient frequency domain method which does not have those previously mentioned drawbacks. In the frequency-domain method of reconstructing wide bandwith pulses, the narrow-band pulses are Fourier transformed and placed next to each other in the frequency domain with or without (splicing) any spectral overlap. In this method, however, a compensation filter is applied to the reconstructed spectrum to compensate for the amplitude ripples that generate paired echos in the impulse response. In order to demonstrate the application of a stepped-frequency algorithm, ii reconstructions were performed using real and artificially generated data sets. With real data, the splicing has been proved to be more successful in achieving high resolution range profiles as the spectral overlap can sometimes cause distortion in phase. Real-time SAR processing is both computationally intensive and time consuming. The evolution of low cost, desktop machines at the commercial market together with the availability of 'Open Source' (OS) software has made the distributed parallel computing a viable solution for intensive SAR processing. A parallel version of stepped-frequency algorithm was created to decompose the task into multiple tasks by taking advantage of the inherent parallel nature of SAR data. In this model, the stepped-frequency processing algorithm adopts the master-worker programing paradigm where the worker process performs the same task on diferent sets of data. The parallel virtual machine (PVM) was used as a messaging 'middleW81'e' of the parallel system. After having successfully implemented the parallel algorithm in a 5 node cluster some timing tests were performed. From the performance analysis it can be inferred that though the parallel system is highly scalable it suffers from high communication overhead. In order to reduce the communication and disk I/O opearation, the previously developed algorithm was modified and some timing analysis was done.

Keywords

Electrical Engineering

Reference:

Collections

Masters

Full item page