Imaging of stepped frequency continuous wave GPR data using the Yule-Walker parametric method

Ground penetrating radar for humanitarian demining has a large bandwidth. It should discriminate small objects against a complex background. The stepped frequency continuous wave radar of IRCTR has a synthesized bandwidth of 4.5 GHz. Classical methods for generating an image of the subsurface include synthetic aperture processing for cross-range resolution and inverse Fourier transform to generate the range profile. This paper presents a novel scheme, based on Yule-Walker auto regression. This method is widely known as an estimator of the power spectral density based on time domain signals. But here it is interpreted as an estimator of the range profile based on frequency domain signals. The result is a significantly improved detectability and range-resolution if the signal is not too weak. The paper presents some results on the effect of the order of the Yule Walker model and a comparison with the MUSIC method