Adnan Sahin

Near Field Forward Scattering, and Object-Based Localization Algorithms for Subsurface Objects

August 20, 1998
11:00 AM
206 Egan

Abstract

Non-invasive detection and localization of objects in the near field of a receiver array have been of interest to many researchers in recent years. Some of the most promising application areas for this technology include landmine remediation, where relatively small metallic or plastic objects are located a few centimeters from the sensors, and hazardous waste remediation, where relatively large metallic objects (eg. steel metal drums) are located on the order of meters from the sensor array. In this thesis, we consider a form of this problem where a plane wave illuminates the region of interest, which is assumed to be a homogeneous, possibly lossy medium containing one or more targets located in the near field of an array of receivers.

In the first part of the thesis we deal with the forward problem in order to obtain an efficient, flexible, and stable algorithm. The forward problem refers to calculation of the scattered fields given the scatterer geometry and properties, and the incident field. The forward solver should lend itself to repeated uses while keeping the computational complexity within practical limits. Specifically, we are interested in development and verification of a recursive algorithm capable of computing scattered fields from multiple dielectric and/or metallic objects in the near field of the array. These conditions are typical of mine detection problems for which the scattered field is observed in the near field, and a mixture of metallic and dielectric objects may be present in the same medium. For this purpose, we present an alternative tessellation scheme and a modification to Chew's well-known recursive T-matrix algorithm.

In the second part, we deal with the inverse problem and introduce algorithms that can detect and localize subsurface objects for near-field measurement geometries. The inherent array structure of the problem suggests that the high resolution array processing techniques quite popular in signal processing community would be well suited for the subsurface detection problem. Based on such a technique, called the Multiple Signal Classification (MUSIC) algorithm, we present two subsurface detection algorithms: computationally simple, but approximate, subarray processing, and computationally intensive, but accurate, matched field processing. Finally, we derive the cra performance bounds for the multiple object detection scenario where the observations are made in the near field. Analytical bounds of estimated object coordinates are then validated by running Monte-Carlo experiments for the MUSIC-based estimator.

Thesis Committee:
Prof. E.L. Miller (advisor)
Prof. D.H. Brooks
Prof. C. Rappaport