No products in the cart.
Automating Image Matching, Cataloging, and Analysis for Photo-Identification Research
Abstract: The expanding use of digital photography for marine mammal photo-identification has created a need for tools to analyze and manage growing image file archives. While database management systems have been commonly employed to manage text and numerical data generated by photo-identification research, their use for the analysis and management of associated image files has been limited. This paper describes a photo-identification database management system with embedded image analysis and management capabilities. Matching and cataloging are expedited using a multiple-attribute, non-metric catalog sorting algorithm. Algorithm efficiency at locating catalog matches for bottlenose dolphins was compared to the performance of a more traditional single-attribute, non-metric approach. Locating catalog matches under the multiple-attribute approach required at least 50% fewer comparisons for 90% of the 409 individuals tested. For 50% of the individuals, 80% fewer comparisons were required. System utility is further extended through embedded mapping components that allow researchers to visually inspect sighting locations following each survey and to examine sighting histories for specific individuals. In addition, a companion ArcGIS™ extension allows researchers to quickly explore and interact with the photo-identification data within a GIS environment. This system, while created for a bottlenose dolphin research application, can be adapted to accommodate photo-identification research on a variety of other species.
Key Words: PHOTO-IDENTIFICATION; NON-METRIC; DATABASE MANAGEMENT SYSTEM; DIGITAL IMAGES; BOTTLENOSE DOLPHIN; TURSIOPS TRUNCATUS
Document Type: Research article
Page Numbers: 374 – 384