Near-realtime Forensic Analysis Capabilities for Moving Target Indicator (MTI) Data

ABSTRACT: ObjectVideo proposes a three tiered system for analyzing activities in WAMI MTI data. The proposed system captures distinct notions of behavior, from event frequency to activity sequences and specific site inference, with independent subsystems. Each level implements an appropriate statistical model with known state-of-the-art performance and fast fitting and inference algorithms. The models build upon detectors, UI tools, and data management components built during the implementation of the ObjectVideo ARGUS-IS ground processing station. This existing software suite in tandem with the decoupled layer design for activity modeling mitigate the risk, inherent to a closed-loop systems spanning all three notions of activity, while yielding a design that promises short time horizon integration into an existing WAMI processing station. BENEFIT: Semi-automated and automatic WAMI forensic analysis tools promise to unleash the power wide area motion imagery (WAMI). The proposed system addresses the critical issues for real-time WAMI processing through: a. The detection of hot-spots in spatiotemporal WAMI data to facilitate the near real-time surveillance of large areas with limited human resources. b. The determination of which individuals in are responsible for the creation of incipient anomalous event clusters. c. The means with which to follow, describe, and analyze the long-term behavior of those individuals. The ObjectVideo Team believes that these advancements will have a high impact in the defense and private sector WAMI and multi-sensor surveillance industries.