Interactive Vision Tools to Index and Search Biological Image Databases for Natural Resource Conservation

University Of Massachusetts Amherst, Amherst MA

Investigators

Abstract

The University of Massachusetts-Amherst is awarded a grant to develop tools for employing digital photography to identify and monitor movements of individual animals in the environment. The project will be carried out in collaboration with the Massachusetts Institute of Technology. The tools developed in this project will make it possible for biologists to query image data bases on a scale that would otherwise be impossible using manual approaches. This project will develop standards for image acquisition and management, tools for rapid segmentation, lighting normalization, and recognition algorithms for individual animal identification. The proposed methodology focuses on recognition of surface markings and targets an "assisted" rather than fully automated identification process. The methodology is motivated by the design and use of generic visual features that can be adapted to many recognition tasks using statistical learning and user provided relevance feedback. The method will be developed and evaluated using three species of amphibians but is expected to scale to other organisms. The ability to identify individual animals is essential to many research endeavors in biology. In field studies, for example, biologists rely on multiple observations over time and space to quantify population sizes, demographic rates, habitat utilization, movement rates and distances. Addressing these types of issues is often critical not only to basic ecological and evolutionary studies, but also to inform conservation planning or natural resource management decisions, in many cases involving rare and endangered species. The emergence of cheap, high quality digital photography makes it possible to document the movements of large numbers of individual animals in species in which traditional methods are not practical.