IMPROVED ENUMERATION OF SCATTERERS USING MULTIFREQUENCY DATA FUSION IN MDL FOR MICROWAVE IMAGING APPLICATIONS

This paper presents a modified version of minimum description length (MDL) method, referred as multifrequency MDL (FMDL), for scatterers enumeration before using the multiple signal classification (MUSIC) algorithm in microwave imaging applications. The inclusion of data from multiple frequencies should make an attempt to reduce the error in number estimation due to noise and multiple scattering. Data fusion in multiple frequencies is performed based on two schemes called averaging and maximization rules. The solution for MDL criterion which is a minimum for one frequency is not likely to be the solution for other frequencies, so by averaging the MDL criterion over the total frequencies or by maximization of the solution for each frequency, we can achieve the correct source number. Furthermore, a whitening step before applying FMDL method is employed to compensate the aspect limitations of the measured data due to the limited number of antennas. The superiority of the proposed FMDL approach with respect to the other competing methods is confirmed by both the numerical examples and the Institut Fresnel experimental dataset. The results indicate that the FMDL yields more accurate estimate of the targets number specially for the cases of low SNR values and very closely spaced scatterers.