Meeting Abstract

6.5  Saturday, Jan. 4 09:00  High Performance Tracker for Subsurface Locomotive Studies SERRANO, MM*; SHARPE, SS; KUCKUK, R; GOLDMAN, DI; VELA, PA; Georgia Institute of Technology; Georgia Institute of Technology; Georgia Institute of Technology; Georgia Institute of Technology; Georgia Institute of Technology mserrano6@gatech.edu

Studying the sub-surface locomotive strategies of burrowing animals through x-ray imaging presents several challenges to automated imaging-based tracking systems. These include a low signal-to-noise ratio, spatially variable imaging statistics arising from heterogeneity in the granular media, and imaging sensor noise. Markers are typically used to compensate for the first issue, however the benefits of markers are often nullified through the latter two issues. Additionally, as it is desirable to impede the subjects as little as possible, the markers are purposefully set flush to the subject’s tissue; thus the marker will vanish when aligned normal to the imaging device. Unlike existing methods that require unique correspondences between markers in subsequent images, the method described here can handle false positive marker detections arising from the noise inherent to x-ray imaging. A Bayesian approach to trajectory estimation using linear models with Gaussian noise leads to a combined Kuhn-Munkres algorithm and Kalman filtering formulation. Coupled with a b-spline model for the subject, the system is able to handle image artifacts and unobservable markers while still capturing a wide range of motions. To validate our system’s robustness, a simulated sandfish (Scincus scincus) is tracked in an environment where the noise levels and the relative contrast are varied. We find that the tracker is able to perform under the majority of these conditions and also present challenging cases. Additional performance evaluations are made on pre-annotated videos of subsurface undulating sandfish and shovel-nose snakes (Chionactis occipitalis) and of the slow-intermittent burying ocellated skinks (Chalcides ocellatus) in granular media.