Three-Dimensional Subterranean Target Identification by Use of Optimization Techniques

Abstract

The identification of a subterranean metallic ore from scattering experiments, conducted on the surface of the ground or in a bore hole, is a classic geophysical problem. In general this problem is not well-posed. However, a priori information about the shape of the target provides enough regularization to make the problem numerically stable. The problem is solved by minimizing the mean-square error between an eleven parameter model, based on the null field approach, and the data. The optimization is done with a Newton technique in which a singular value decomposition of the model Jacobian is employed. The algorithm is very stable to noise and makes good reconstructions from feasible starting guesses, for realistically noise contaminated data. (Less)