Classification, identification, and modeling of unexploded ordnance in realistic environments
| Autor: | Zhang, Beijia, 1980 |
|---|---|
| Jazyk: | angličtina |
| Rok vydání: | 2008 |
| Předmět: | |
| Druh dokumentu: | Diplomová práce |
| Popis: | Thesis (Ph. D.)--Massachusetts Institute of Technology, Dept. of Electrical Engineering and Computer Science, 2008. Includes bibliographical references (p. 205-218). Recovery of buried unexploded ordnance (UXO) is very slow and expensive due to the high false alarm rate created by clutter. Electromagnetic induction (EMI) has been shown to be a promising technique for UXO detection and discrimination. This thesis uses the EMI response of buried targets to identify or classify them. To perform such discrimination, accurate forward models of buried UXO are needed. This thesis provides a survey of existing target models: the dipole model, the spheroid model, and the fundamental mode model. Then the implementation of a new model, the spheroidal mode model, is described and validated against measurements of a UXO. Furthermore, an in-depth study of the effects of permeable soil, modeled as a permeable half space, is presented. This study concludes that the discontinuity created by the air to permeable soil interface produces minimal effect in the response of a buried object. The change is limited to a magnitude shift of the real portion of the EMI response and can be reproduced by superposition of a permeable half space response on the response of the same object in frees pace. Accurate soil modeling also allows one to invert for soil permeability values from measured data if such data are in known units. However, the EMI sensor used in this study provides measurements in consistent but unknown units. Furthermore, the instrument is from a third party and is proprietary. Therefore, this thesis describes the development of a non-invasive method to model and calibrate non-adaptive instruments so that all measurements can be converted into units consistent with modeled data. This conversion factor is shown to be a constant value across various conditions, thus demonstrating its validity. (cont.) Given that now a more complete model of the measurable response of a buried UXO is implemented, this study proceeds to demonstrate that EMI responses from UXO and clutter objects can be used to identify the objects through the application of Differential Evolution (DE), a type of Genetic Algorithm. DE is used to optimize the parameters of the UXO fundamental mode model to produce a match between the modeled response and the measured response of an unknown object. When this optimization procedure is applied across a library of models for possible UXO, the correct identity of the unknown object can be ascertained because the corresponding library member will produce the closest match. Furthermore, responses from clutter objects are shown to produce very poor matches to library objects, thus providing a method to discriminate UXO from clutter. These optimization experiments are conducted on measurements of UXO in air, UXO in air but obscured by clutter fragments, buried UXO, and buried UXO obscured by clutter fragments. It is shown that the optimization procedure is successful for shallow buried objects obscured by light clutter contributing to roughly 20 dB SNR, but is limited in applicability towards very deeply buried UXO or those in dense clutter environments. The DE algorithm implemented in this study is parallelized and the optimization results are computed with a multi-processor supercomputer. Thus, the computational requirement of DE is a considerable drawback, and the method cannot be used for real time, on-site inversion of measured UXO data. To address this concern, a different approach to inversion is also implemented in this study. Rather than identifying particular UXO, one may do a discrimination between general UXO and general clutter items. Previous work has shown that the expansion coefficients of EMI responses in the spheroidal coordinate system can uniquely characterize the corresponding targets. (cont.) Therefore, these coefficients readily lend themselves for use as features by which objects can be classified as likely to be UXO or unlikely to be UXO. To do such classification, the relationship between these coefficients and the physical properties of UXO and clutter, such as differences in size or body-of-revolution properties or material heterogeneity properties, must be found. This thesis shows that such relationships are complex and require the use of the automated pattern recognition capability of machine learning. Two machine learning algorithms, Support Vector Machines and Neural Networks, are used to identify whether objects are likely to be UXO. Furthermore, the effects of small diffuse clutter fragments and uncertainty about the target position are investigated. This discrimination procedure is applied on both synthetic data from models and measurements of UXO and clutter. It is found that good discrimination is possible for up to 20 dB SNR. But the discrimination is sensitive to inaccurate estimations of a target's depth. It is found that the accuracy must be within a 10 cm deviation of an object's true depth. The general conclusion forwarded by this work is that while increasingly accurate discrimination capabilities can be produced through more detailed forward modeling and application of robust optimization and learning algorithms, the presence of noise and clutter is still of great concern. Minimization or filtering of such noise is necessary before field deployable discrimination techniques can be realized. by Beijia Zhang. Ph.D. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|