Compressive sensing for radar target detection

Çağlıyan, Firuze
Compressive sampling, also known as compressive sensing and sparse recovery, is a new type of sampling theory, which predicts that sparse signals and images can be reconstructed from far less amount of data than what was traditionally considered necessary (i.e. Nyquist/Shannon sampling theory). The theory has many applications such as design of new imaging systems, cameras, sensor networks and analog to digital converters. Several algorithms have been proposed for the measurement and recovery process of the theory. The theory uses only a small amount of measurements which are linear, nonadaptive and suitably designed. The reconstruction process is nonlinear and simply depends on searching for the sparsest vector that is coherent with the measurements. The compressive sensing theory and its key points are explained in detail. In this thesis, compressive sensing (CS) is used to reconstruct the target scene of a radar. The target scene is discretized so that a total of N possible target locations exist. The number of targets K is assumed to be small (i.e., K<<N) meaning that the target scene is sparsely populated. A theoretical lower bound on the number of measurements M depending on the sparsity K and the total number of data N is presented based on the results in the literature. The target scene reconstruction results for different noise levels are compared. Three different compressive sensing reconstruction methods are described and their performances are compared. The compressive sensing radar target detection and the classical radar detection performance difference is investigated. The change in probability of detection due to SNR variation under constant false alarm rate (FAR) is analyzed. Finally, the effect of Doppler to the compressive sensing radar target detection is analyzed. When the number of measurements is limited, i.e., M<N, there is an SNR loss in detection performance. The CS method roughly attains the performance of classical detection when received SNR is boosted by M/N either with higher power at the transmitter or lower noise figure at the receiver.
Citation Formats
F. Çağlıyan, “Compressive sensing for radar target detection,” M.S. - Master of Science, Middle East Technical University, 2014.