ZOOMETRICS – Biometric Identification for Wildlife
Effective approaches for the conservation of wildlife require a sound knowledge of their population demography. Such information is obtained though mark-recapture studies that enable following the lives of individual animals. Matching body patterns to uniquely identify individual wildlife members, particularly those of endangered species, is commonly used in many conservation studies. Being cost-effective & non-invasive, this is often preferred over other methods such as tagging, marking, or amputation. Manual matching can be slow and error-prone especially when the population size is large. Computer-based image analysis techniques can provide not only a low cost solution but also make analysis of the patterns fast, accurate and repeatable.
The objective of the project is to develop a low-cost, portable but versatile system for reliable identification of individual wildlife members. Following a successful pilot study phase, the project is now directed at automatically identifying Great Crested Newts from their distinctive belly patterns. Ongoing research activities currently include:
- preparation of an expanded database of newt images
- automated segmentation of the Region of Interest
- evaluation of various distinguishing features and matching algorithms
- study of ageing effects
- incorporate image quality based system adaptation
- implementations in a mobile platform
The study will subsequently be extended to other wildlife data (e.g. Grass snakes, adders).
This project is in collaboration with Professor Richard Griffiths and his team at Durrell Institute of Conservation & Ecology (DICE) in the School of Anthropology and Conservation.