Challenge

While additional sensors provide means toward gaining a deeper understanding of hostile activities, new methods for sifting through the data haystack must be developed to identify the relevant information and to present it to the warfighters and analysts. Simultaneously tracking thousands of people and vehicles in dense urban traffic is not possible for an analyst to do alone. While many sensors and systems may be available, a key challenge is to optimize resource management of sensors, communication bandwidth, and airborne processing. The idea is to guide the right sensor to the right place at the right time and to present only the relevant data to the end-user.

Benefits of SIG’s Approach

With SIG’s exploitation software, analysts and warfighters can process more sensor data in less time, while maintaining full-scene awareness of high-priority events and real-time cueing to ambiguous activity requiring human interpretation. SIG’s exploitation solution enables accurate real-time detection and tracking of thousands of vehicles and people through challenging lighting conditions, dense traffic regions, and kinematic dynamics. These challenges cause excessive false-tracks or result in missed or inaccurate tracks using conventional exploitation approaches. SIG’s multi-model probabilistic framework is robust to these environmental challenges and enables visual display of track confidence to the warfighter, accurate integration of other sensor data and information sources, and serves as a foundation for behavior modeling and threat detection.

Products/Solution

SIG’s Processing, Exploitation, and Dissemination (PED) software suite can be deployed on real-time SWaP (size, weight, and power) constrained platforms. The software has been developed using a multi-threaded parallel architecture to meet processing requirements for both large format wide area motion imagery (WAMI) and for full motion video systems.

Technology Summary

SIG has developed a unique probabilistic framework that sifts the relevant data from the overwhelming volume of raw pixels and sensor data. These data products include stabilized and geo-registered imagery, vehicle and dismount tracks, traffic and behavior models, and analyst-aided-cueing for improved accuracy. The mathematics of probability provides a critical mission need by formally quantifying the confidence of estimated target tracks and other exploitation products. This confidence can be displayed visually to help the warfighter make better decisions and can also be used to optimally task other sensors to resolve current tactical ambiguities.

SIG’s foundation for video exploitation and analytics is a Bayesian framework that models the video pixel statistics for image background, target (foreground) appearance, target shape, and target kinematics. Modern non-parametric sampling methods are used to allow each of underlying models to accurately fit observations and adapt to changing environmental conditions. The background and foreground models capture the spatial and temporal pixel dynamics; the shape model allows for target deformations that occur in both airborne and fixed-camera scenarios; the kinematic prediction model uses (when available) traffic flow-state information and knowledge of the environment and traffic patterns to provide dramatically improved accuracy.