TRIDENT SPECTRE is an annual venue to test and evaluate emerging technologies hosted jointly by members of the
United States Department of Defense and the Intelligence Community. The event focuses on projects involving technical
collections, Geospatial Intelligence, Analysis, Human Intelligence, and communications. It offers the DoD and IC a
unique opportunity to test new ideas and concepts in a secure environment with users, operators, technicians, engineers,
scientists, and cleared industry partners collaboratively.
Wide area motion imagery (WAMI) sensors increasingly are being used for persistent surveillance of large urban areas.
One of the potential uses for such surveillance is the discovery of geo-spatial networks, which are sets of locations
linked by repeated traffic flow over an extended period of time. In this work we present a simple method of deriving
geo-spatial network links automatically from ambiguous track segments or tracklets. The method avoids making explicit
tracklet linking decisions and relies on temporal aggregation to identify the persistent origin-destination location pairs.
We present experimental network discovery results using simulated high density track data for a downtown urban
Smart Sensor Networks are becoming important target detection and tracking tools. The challenging problems in such networks include the sensor fusion, data management and communication schemes. This work discusses techniques used to distribute sensor management and multi-target tracking responsibilities across an ad hoc, self-healing cluster of sensor nodes. Although miniaturized computing resources possess the ability to host complex tracking and data fusion algorithms, there still exist inherent bandwidth constraints on the RF channel. Therefore, special attention is placed on the reduction of node-to-node communications within the cluster by minimizing unsolicited messaging, and distributing the sensor fusion and tracking tasks onto local portions of the network. Several challenging problems are addressed in this work including track initialization and conflict resolution, track ownership handling, and communication control optimization. Emphasis is also placed on increasing the overall robustness of the sensor cluster through independent decision capabilities on all sensor nodes. Track initiation is performed using collaborative sensing within a neighborhood of sensor nodes, allowing each node to independently determine if initial track ownership should be assumed. This autonomous track initiation prevents the formation of duplicate tracks while eliminating the need for a central “management” node to assign tracking responsibilities. Track update is performed as an ownership node requests sensor reports from neighboring nodes based on track error covariance and the neighboring nodes geo-positional location. Track ownership is periodically recomputed using propagated track states to determine which sensing node provides the desired coverage characteristics. High fidelity multi-target simulation results are presented, indicating the distribution of sensor management and tracking capabilities to not only reduce communication bandwidth consumption, but to also simplify multi-target tracking within the cluster.
This work investigates a distributed approach for fusion and sensor management in unattended sensor networks. The distributed approach not only improves robustness to node failure, but also reduces network communications load significantly over that of the more traditional non-managed centralized processing approach. Monte Carlo simulations show that bandwidth reductions of factors of two to three over that of traditional architectures are achievable, depending on such factors as radio communications range and node availability.
Distributed sensor networks will play a key role in the network centric warfighting environments of the future. We envision a ubiquitous sensing `fabric,' comprising sensors distributed over the terrain and carried on manned and unmanned, terrestrial and airborne vehicles. As a complex `system of systems,' this fabric will need to adapt and self-organize to perform a variety of higher-level tasks such as surveillance and target acquisition. The topology and availability of the sensors will be constantly changing, as will the needs of users as dictated by evolving missions and operational environments. In this work, focusing on the task of target tracking, we address approaches for locating and organizing sensing and processing resources and present algorithms for suitably fusing the observations obtained from a varied and changing set of sensors. Run-time discovery and access of new sensing resources are obtained through the use of Java Jini, treating sensing resources as `services' and viewing higher-level processes such as tracking as clients. Algorithms for fusing generic sensor observations for target tracking are based on the extended Kalman filter, while detection and track initiation are based on a new likelihood projection technique. We present results from an implementation of these concepts in a real- time sensor testbed and discuss lessons learned.
A new synthetic ladar image generation tool has been developed for use in ATR algorithm development and evaluation. The image generation tool, called SYLVER, was designed originally for air-to-ground targeting scenarios and simulates the effects of arbitrary platform and target motion as well as sensor scanning. Particular attention has been given to realistic rendering in the vicinity of range discontinuities, since edges are important to many recognition algorithms. This paper presents an overview of the image generation approach and details the mathematical models employed. Example images are shown and characteristics of the simulated imagery are compared with those of field collected data.