Effects of Movement on Biometric Facial Recognition in Body-Worn Cameras
Abstract
This study examined how three different manipulations of a single policing stance affected the quality scores and matching performance in a biometric facial recognition system; it was conducted in three phases. In the first phase, the researcher collected qualitative survey data from active, sworn law enforcement officers in 15 states. In the second phase, the researcher collected quantitative data using a single facial recognition subject and a static body-worn camera mounted to an adjustable tripod. In the third phase, the researcher collected quantitative data from bodyworn camera-equipped law enforcement officers who filmed a stationary target as they executed a series of specified movements from the interview stance. The second phase tested two different body-worn cameras: one that is popular among law enforcement agencies in the United States, the Axon Body 2; and one that is inexpensive and available to the general public via a popular internet commerce website. The third phase tested only the Axon Body 2. Results of the study showed that matching results are poor in a biometric system where the test body-worn camera was the sensor, with error rates as high as 100% when the body-worn camera wearer was in motion. The general conclusion of this study is that a body-worn camera is not a suitable sensor for a biometric facial recognition system at this time, though advances in camera technology and biometric systems may close the gap in the future.
Degree
Ph.D.
Advisors
Elliott, Purdue University.
Subject Area
Law enforcement|Artificial intelligence|Computer science
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.