Charles River Analytics is developing a system for the Air Force Research Laboratory Sensors Directorate (AFRL/SN) for detecting and tracking moving objects. Since image exploitation algorithms for Intelligence, Surveillance, and Reconnaissance (ISR) and weapon systems are sensitive to differences between the operating conditions (OCs) under which they are trained and the extended operation conditions in which they are tested, new algorithms must be developed to adjust for these differences. These algorithms must be able to identify the OC in which they are operating, and then tune their parameters and models accordingly. For instance, a system designed to track vehicles on highways would not adjust well to driving conditions in a parking lot, as variables such as speed, stopping, or changing directions vary greatly.
Charles River Analytics is responding to this variability with ALPS for Real-Time Video Tracking, which adjusts for the problems associated with the sensitivity to OC differences. “ALPS automatically detects moving objects of interest and tracks them,” said Neal Checka, a scientist at Charles River Analytics. “It tracks targets across multiple frames using a multiple hypothesis tracker tightly coupled with a particle filter.
Checka explained additional capabilities of ALPS: “It can detect the operating condition by analyzing the size, appearance, and motion traits of the objects being tracked. For example, our system can distinguish object type (pedestrians versus vehicles) and terrain type (highways versus parking lots) using statistics such as aspect ratio, maximum and average speed, and maximum rate of change.”
For the current phase of development, Charles River Analytics is creating a fully-functional, real-time prototype that can perform across various operating scenarios, vehicle types, and motion patterns. Charles River Analytics is uniquely qualified for this project with its extensive experience in parameter adaptation of computer vision algorithms, automatic target recognition, target tracking, event detection and recognition, and image processing technologies for imaging modalities.