Center for Embedded Network Sensing

We present a scalable end-to-end system for vision-based monitoring of a biological phenomenon. Our system enables automated analysis of thousands of images, where manual processing would be infeasible. We automate the analysis of raw imaging data using statistics that are tailored to the task of interest, the study of avian behavior during nesting cycles. The system uses simple image statistics (features) as the low-level representation to be fed to generic classifiers and final inferences exploit the temporal and spatial consistencies. Our testbed achieves bird detection accuracy of 82%, and egg counting accuracy of 84%, allowing inference of avian nesting stage with accuracy within a day. Our results demonstrate the challenges and potential of using imagers as biological sensors. An exploration of system performance under varying image resolution and frame rate suggest that an in situ adaptive vision system is technically feasible.