Seabed characterization by inversion of parametric sonar data: selection of the cost function

Abstract

A class of possible cost functions to be used in model-based estimation of bottom parameters from highfrequency backscatter signals is introduced. The cost is based on the properttes of the signal in the wavelettransformed domain. It is shown, through a simulative study, that the cost can be selectively tuned to be sensitive only to subsets of the parameters to be estimated. This in turn suggests that a "divide-and-conquer" inversion strategy can be applied at least in the case of normal incidence bottom returns from parametric sonar data.