System and algorithms for an autonomous observatory assisting the search for the Ivory-Billed Woodpecker

Ornithologists, conservationists, and millions of birdwatchers consider the Ivory-Billed Woodpecker (IBWO) the ‘Holy Grail of Birds.’ There have been hundreds of reports of sightings of this magnificent bird but no conclusive photo has been recorded since 1948. Under our broader research effort to develop networked autonomous “observatories” for natural environments, we have been working with the Cornell University researchers who have led a massive search initiated by eyewitness reports and low resolution video captured in the Cache River National Wildlife Refuge of Arkansas. In this paper we describe the two-camera autonomous observatory system we designed and installed in Arkansas that continuously scans the sky, recording high resolution video of candidate birds. We report on hardware, software, and algorithms based on 15 months of field experiments. Our image processing algorithm combines size filtering, nonparametric motion filtering, and temporal difference filtering to detect flying birds. Initial results suggest that we have met our four design goals: sensitivity, data reduction, accuracy, and robustness. Video segments with several bird species have been conclusively identified, video data has been consistently reduced by 99.9953%, the system is able to capture birds flying at a speed of up to 60km per hour with low false negative rates, and the system has held up to harsh field conditions. For latest updates and samples of video, please see http://www.c-o-n-e.org/acone/.