Increases in the number of cameras deployed, frame rate, and detector array sizes have led to a dramatic increase in the volume of motion imagery data that is collected. Without a corresponding increase in analytical manpower, much of the data is not analyzed to full potential. This creates a need for fast, automated, and robust methods for detecting signals of interest. Current approaches fall into two categories: detect-before-track (DBT), which are fast but often poor at detecting dim targets, and track-before-detect (TBD) methods which can offer better performance but are typically much slower. This research seeks to contribute to the near real time detection of low SNR, unresolved moving targets through an extension of earlier work on higher order moments anomaly detection, a method that exploits both spatial and temporal information but is still computationally efficient and massively parallelizable. It was found that intelligent selection of parameters can improve probability of detection by as much as 25% compared to earlier work with higherorder moments. The present method can reduce detection thresholds by 40% compared to the Reed-Xiaoli anomaly detector for low SNR targets (for a given probability of detection and false alarm).

Date Published: 17 May 2016
PDF: 12 pages
Proc. SPIE 9828, Airborne Intelligence, Surveillance, Reconnaissance (ISR) Systems and Applications XIII, 98280K (17 May 2016); doi: 10.1117/12.2224544

Published in SPIE Proceedings Vol. 9828:
Airborne Intelligence, Surveillance, Reconnaissance (ISR) Systems and Applications XIII
Daniel J. Henry; Gregory J. Gosian; Davis A. Lange; Dale Linne von Berg; Thomas J. Walls; Darrell L. Young, Editor(s)