Matching spot patterns for animal re-ID

Speaker: Dr Kasim Terzic (University of St Andrews)

Abstract

We present a photo-based re-identification framework for Arctic charr (Salvelinus alpinus) based on matching characteristic spot patterns. We apply machine learning to the problem of automatic spot extraction and apply two matching algorithms to re-ID specimens based on past observation. In our case study based on a cave system in Iceland, our system managed to successfully identify unmarked invididuals from photos alone, and re-identified individuals that had lost tags, resulting in a 4% increase in our estimate of survival rate.