Share Email Print

Proceedings Paper

Qualitative performance of a local track repair algorithm for video tracking on small UAVs
Author(s): Stephen DelMarco
Format Member Price Non-Member Price
PDF $17.00 $21.00

Paper Abstract

A persistent problem with real-time video tracking on small UAVs (SUAV) is that tracks can break due to camera motion, target occlusions, frame-to-frame mis-registrations, signal-to-noise ratio issues, and other causes. Repairing video tracks, by appropriately connecting broken tracks, is essential for high quality track maintenance. Previously, we developed a video track repair algorithm (VTR) to repair short track breaks. We investigated performance on tracking data from a video tracker operating on video data acquired from an SUAV. The repair approach accumulates evidence across frames using multihypothesis sequential probability ratio tests (MHSPRT). The MHSPRT framework propagates posterior probabilities associated with each track repair hypothesis to make track connections. In this paper we perform several numerical experiments using simulated tracking data to map out qualitative behavior of the VTR over a wider set of operating conditions. We examine effects of measurement noise level, track state-space separation, number of evidence accumulation frames, and stopping probability threshold on repair performance. We investigate the effect of these factors on posterior probability propagation in the MHSPRT. We indicate potential algorithm enhancements resulting from conclusions drawn from experimental results. We demonstrate how a multi-frame evidence accumulation approach can provide superior performance to a single-frame maximum likelihood approach. We demonstrate that with fewer frames MHSPRT performance compares favorably with maximum a posteriori performance, despite few analytical results on MHSPRT optimality. First we provide an overview of the track repair algorithm. Next we describe the numerical experiments, present results, and interpret results to infer performance behavior.

Paper Details

Date Published: 16 April 2008
PDF: 12 pages
Proc. SPIE 6963, Unattended Ground, Sea, and Air Sensor Technologies and Applications X, 69630N (16 April 2008); doi: 10.1117/12.771429
Show Author Affiliations
Stephen DelMarco, BAE Systems (United States)

Published in SPIE Proceedings Vol. 6963:
Unattended Ground, Sea, and Air Sensor Technologies and Applications X
Edward M. Carapezza, Editor(s)

© SPIE. Terms of Use
Back to Top