Performance analysis of one dimensional radar imaging based on sparse and non-uniform frequency samplings

The performance of one dimensional radar imaging based on sparse and non-uniform frequency samplings is analyzed from the view of parameter estimation. Theoretical analysis indicates that using N sparse and non-uniform samplings instead of M uniform samplings only causes a performance deterioration of N/M, on the average, which is far better than the performance deterioration of N3/M3 if we reduce the data to N uniform samplings. This relationship is verified by numerical simulations. An algorithm to select a good set of samplings is also proposed to ensure a good performance among the possible random sampling sets. Results show that a good set tends to cover both ends of the frequency band well whereas a bad set tends to cover a shorter effective frequency band. The performance advantage of the selected good sampling set is testified using compact range data.