You are here

Next Generation Tracking Architectures for Urban Surveillance Areas

Description:

OBJECTIVE: Develop and implement next generation tracking architectures which exploit wide area motion imagery and leverage projected High Performance Computing capacities. DESCRIPTION: Wide area motion imagery (WAMI) systems, such as the Autonomous Real-time Ground Ubiquitous Surveillance - Imaging System (ARGUS-IS), produce tens of thousands of moving target indicator (MTI) detections from city- size urban areas (over 40 square kilometers) at video rates of greater than 12 Hz. On-board processing currently provides a limited number of operator-nominated real-time Stationary Video Windows (SVW) and Tracking Video Windows (TVW) for target monitoring and engagement. Off-board processing currently provides a much greater number of automated tracks for forensic analysis but only for a partial portion of the urban surveillance area (Reference 1). To improve mission effectiveness, the on-board number of real-time targets monitored and engaged needs to be significantly increased and the off-board forensic analysis portion of the urban surveillance area needs to be dramatically expanded. Thus, innovative next generation track processing architectures are needed to fully exploit the revolutionary advances being achieved by WAMI systems with regard to target detection and surveillance area capabilities. As the densities of the surveillance areas have increased (from rural, to suburban, and ultimately to urban), it has become especially challenging for current generation track processing architectures to correctly and efficiently associate individual MTI detections with vehicle and dismount targets. For example, Multiple Hypothesis Tracking (MHT) has been deployed to address dense environments by forming and maintaining hypotheses until sufficient confidence is achieved to declare an association. However, current MHT"s are limited by the density of the surveillance areas as the track processing requirements grow exponentially as a function of the total number of targets and their relative proximities. Other current track processing architectures such as Particle Filtering and Joint Probabilistic Data Association are also severely challenged by urban surveillance areas. A potentially rewarding avenue of investigation is the future relationship of target density and High Performance Computing (HPC) capacities with target density having reached an upper bound (i.e. only so many vehicles can be on a roadway; even in an urban traffic jam) whereas HPC processor developments and associated capacities are projected to be unbounded (i.e. multicore processors and distributed databases; for handling Big Data). While these future HPC capacities may extend current tracking architectures which are based on target trajectory modeling; more importantly, they can potentially enable next generation tracking architectures which may be based on higher level multiple-target activity modeling which focuses on correlations of appearance events at the pixel level identified through various learning methods. These pixel-level approaches can be sensitive to imagery noise, etc. and can be computationally demanding due to the large number of pixel-level events which need to be continuously and simultaneously monitored and detected (Reference 2). PHASE I: Develop a prototype next generation tracking architecture which exploit wide area motion imagery (WAMI) systems and leverage projected High Performance Computing (HPC) capacities for on-board and off-board processing of vehicle and dismount targets within urban surveillance areas. PHASE II: Determine feasibility of next generation tracking architectures and demonstrate representative implementations of on-board and off-board processing through utilization of a High Performance Computing Environment which includes interactive computing resources, enhanced data management capabilities, and increased data storage. PHASE III: Exploitation of WAMI over adversarial urban surveillance areas provides for timely target engagement and more comprehensive forensic analysis. Exploitation of WAMI over domestic urban surveillance areas provides for timely emergency response and more comprehensive disaster monitoring. REFERENCES: 1. Leininger, Brian et. al. of DARPA, Edwards, Jonathan et. al. of BAE Systems (2008). Autonomous Real-time Ground Ubiquitous Surveillance Imaging System (ARGUS-IS). Proc. of SPIE Vol. 6981. 2. Xiang, Tao, and Gong, Shaogang of Department of Computer Science, Queen Mary, University of London (2006). Beyond Tracking: Modeling Activity and Understanding Behavior. International Journal of Computer Vision 67(1), (pp. 2151). 3. HPCMP Enhanced User Environment Begins Production Operations. http://www.ccac.hpc.mil/news/index.html#CCAC-33.
US Flag An Official Website of the United States Government