Motion Pattern Anomaly Detection System (MPADS)

The basic goal of this research is to improve the capability to detect, locate, and characterize emerging abnormal activities in cluttered environments based on motion pattern analysis. Parietal Systems, Inc. (PSI) proposes to do this by developing novel closed-loop pattern detection and sensor management algorithms to detect significant motion anomalies in normal background activity and to manage sensor resources to optimize anomaly detection performance. The anomaly detection algorithm will utilize techniques and results of scan statistics to process large MTI data sets efficiently without requiring well-formed tracks. This statistical approach works equally well with MTI radar, video, or other types of sensors capable of detecting moving objects. The sensor resource management algorithm will employ an innovative approach based on generalized multi-armed bandits to develop an online algorithm with minimal processing requirements. Because this approach accounts explicitly for future sensor observations, it is able to generate non-myopic resource decisions that achieve near-optimal performance. The research will also apply innovative performance analysis methods to characterize the expected anomaly detection and resource management performance of this approach as a function of sensor capabilities, resource usage, and the contrast between anomalies and normal background activity.