Radar image reconstruction by discrete model fitting in a layered inhomogeneous medium

We have been studying radar image reconstruction based on the discrete model fitting (DMF) method to realize high performance of subsurface sensing. In the present paper, the DMF is extended to treat the inhomogeneity of the medium, and the performance of the algorithm is verified by computer simulation. We consider a situation in which several antennas are used both for transmission and reception, and then many time series data are obtained with all combination of transmitters and receivers. The medium is assumed to consist of layers with different permittivities, and point scatterers are embedded in it. Parameters to be estimated by the model fitting are positions and radar cross sections of targets, and permittivities and depths of layers. In the model fitting, nonlinear leastsquares improves the model parameters so that the estimated received data computed by ray tracing agree with the observed data. Since the nonlinear least squares is an iterative method, appropriate initial values for the model parameters are estimated from information on the delay time ofreceived echoes. To enhance the ability of detection of targets and layers, the combination of the initial guess and the model fitting is iterated as the number of assumed targets and layers is increased. The proposed method can treat multiple scattering and a large discontinuity of a medium, which the conventional methods based on Born approximation cannot treat. Moreover, this algorithm takes inhomogeneity of the medium into account, it can estimate a target location more precisely than the conventional aperture synthesis technique.