Contextually Adaptive False Alarm Mitigation

Logos Technologies proposes to develop a processing chain that creates local background models based on spectral and texture based features to improve the detection and false alarm performance of change detection, tracking, and target detection systems for wide area persistent surveillance. Our innovative approach extends proven methods in hyperspectral image processing to provide updatable context-dependent models that support mitigation of important sources of false alarms and track misassociation for wide area persistent surveillance data. It does not need a detailed urban model and/or complex rules for target behavior but builds models based on intrinsic properties of the scene. These rules can be combined with other context dependent factors, such as pre-existing traffic or proximity to buildings. We will provide analytic support for the approach, develop all key elements of the processing chain, and demonstrate improved performance using relevant data from EO sensors. We will develop this concept in a form applicable to sensor architectures that provide only limited spectral information, including the anticipated ARGUS-IS system. As part of our commercialization strategy, will demonstrate its applicability to future variants of the Light Hawk wide area persistent surveillance and exploitation system.