Microwave Imaging From Sparse Measurements for Near-Field Synthetic Aperture Radar

This paper reports the experimental studies for four image reconstruction methods from sparse measurement using wideband microwave synthetic aperture radar systems. The four methods include two denoising methods using zero filling (ZF) and nonuniform fast Fourier transform (NUFFT), and two compressed sensing (CS) methods using the orthogonal matching pursuit and the conjugate gradient algorithms. The specimens under test (SUTs) consist of a tray of small rocks with different densities with/without one piece wrapped in an aluminum foil. The raw measurements of the SUTs are randomly undersampled in the spatial domain, and the images are reconstructed from the measurements of 10%-60% sparse-sampling rates. The results show that the CS method achieves good image quality with as low as 30% sparse-sampling rate, while ZF and NUFFT require 50% to obtain acceptable quality. An enhanced Otsu's method is also proposed to detect the foiled rock from sparse reconstructions, which improves detection performance for the sparse-sampling rate of 5%-15%. The reduction of spatial measurement leads to reduced cost or reduced measurement time.