ICCS07/267

Lisa Premerlani, Department of Computer Science, Boston University

Margrit Betke, Department of Computer Science, Boston University

Stan Sclaroff, Department of Computer Science, Boston University

Thomas Kunz, Department of Biology, Boston University

Abstract

Brazilian free-tailed bats (Tadarida brasiliensis) form some of the

largest known aggregations of mammals. In south-central Texas, these

bats disperse nightly from their roosts to feed on aerial insects,

including those from agricultural crops. One of the least understood

aspects of the biology of these bats is the behavior of individuals

when they emerge from their roosts and forage for food. We have

developed algorithms for analyzing infrared videos of bats in

flight. We compared two tracking approaches, an alpha-beta and a

Kalman filter, and analyzed the performance of two data-association

approaches, a greedy method and the Hungarian Method. Our analysis

revealed similar results for both filters and more accurate results

for the Hungarian Method. We applied the algorithms to video data of

bats that foraged over a small pond. We computed stereoscopic

reconstructions of bat flight paths from videos that were recorded

simultaneously from two uncalibrated infrared cameras placed

approximately 14 meters apart. To locate the three-dimensional

coordinates of each bat along its flight trajectory, we calibrated the

two cameras spatially using Horn's relative orientation algorithm and

then performed a triangulation. From the trajectories, we calculated

velocities and changes in flight direction of the bats. Our analysis

indicates that the minimum, maximum, and average speeds of these bats

were 1.53, 20.24, and 8.42 m/s, respectively. The minimum, maximum,

and average changes in flight direction in 1/60 seconds were 0.2,

96.1, and 6.4 degrees, respectively. Our results provide the first

stereoscopic analysis of the foraging behavior of free-ranging bats

based on thermal imaging technology.