Main | Browse | Search | Author Links | Manage ETD List | Review ETDs | Catalog ETDs | Help

Type of Document	Master's Thesis
Author	Gentile, James Edward
Author's Email Address	jgentile@nd.edu
URN	etd-10082008-203258
Title	Evaluations and Innovations in Face Detection
Degree	Master of Science in Computer Science and Engineering
Department	Computer Science and Engineering
Advisory Committee	Advisor Name Title Dr. Greg Madey Committee Member Dr. Kevin Bowyer Committee Member Dr. Patrick Flynn Committee Member Dr. Scott Emrich Committee Member
Keywords	pattern recognition learning face detection profile
Date of Defense	2008-09-22
Availability	restricted
Abstract Face detection is the first step in an automated system that identifies individuals based on their face’s appearance. Detection is not trivial because the face is a non-rigid object that varies in position, rotation and size. The face’s appearance can be altered by variations in expression, skin tone and lighting. This thesis evaluates popular detectors on imagery with known conditions. It shows that the Bayesian classifier is preferred under structured lighting conditions and the cascade classifier is preferred under unstructured lighting conditions. Finally, the lessons learned from detecting in unstructured light are applied to detect drivers as they pass a security checkpoint. A drive-by detection system in unique because the driver’s side window and outdoor conditions pose possible problems. A multi-biometric detector that uses ears and profile faces to detect drivers is shown to work better than a simple, single-site detector.
Files	Filename       Size       Approximate Download Time (Hours:Minutes:Seconds)     28.8 Modem   56K Modem   ISDN (64 Kb)   ISDN (128 Kb)   Higher-speed Access  GentileJ102008.pdf 1.59 Mb 00:07:22 00:03:47 00:03:19 00:01:39 00:00:08 indicates that a file or directory is accessible from the campus network only.