Date of Graduation


Document Type


Degree Type



Statler College of Engineering and Mineral Resources


Lane Department of Computer Science and Electrical Engineering

Committee Chair

Hany H. Ammar

Committee Co-Chair

Donald A. Adjeroh


Identification of deceased individuals based on dental characteristics is receiving increased attention, especially with the large volume of victims encountered in mass disasters. In this work we consider three important problems in automated dental identification beyond the basic approach of tooth-to-tooth matching.;The first problem is on automatic classification of teeth into incisors, canines, premolars and molars as part of creating a data structure that guides tooth-to-tooth matching, thus avoiding illogical comparisons that inefficiently consume the limited computational resources and may also mislead the decision-making. We tackle this problem using principal component analysis and string matching techniques. We reconstruct the segmented teeth using the eigenvectors of the image subspaces of the four teeth classes, and then call the teeth classes that achieve least energy-discrepancy between the novel teeth and their approximations. We exploit teeth neighborhood rules in validating teeth-classes and hence assign each tooth a number corresponding to its location in a dental chart. Our approach achieves 82% teeth labeling accuracy based on a large test dataset of bitewing films.;Because dental radiographic films capture projections of distinct teeth; and often multiple views for each of the distinct teeth, in the second problem we look for a scheme that exploits teeth multiplicity to achieve more reliable match decisions when we compare the dental records of a subject and a candidate match. Hence, we propose a hierarchical fusion scheme that utilizes both aspects of teeth multiplicity for improving teeth-level (micro) and case-level (macro) decision-making. We achieve a genuine accept rate in excess of 85%.;In the third problem we study the performance limits of dental identification due to features capabilities. We consider two types of features used in dental identification, namely teeth contours and appearance features. We propose a methodology for determining the number of degrees of freedom possessed by a feature set, as a figure of merit, based on modeling joint distributions using copulas under less stringent assumptions on the dependence between feature dimensions. We also offer workable approximations of this approach.