Performance Modeling of Feature Aided Trackers

The bandwidth of current and future image analysts is insufficient to ingest the vast quantity of data generated by staring sensing platforms. Because of this, signature exploitation algorithms need to play a role in the data processing chain to cue analysts to specific regions of interest within the imagery. For systems which utilize exploitation algorithms, there is a need to understand the ability and limitations under various conditions. We propose a feature aided tracker that is able to model its performance based on the current operating conditions. This performance model is a crucial component of any layered sensing architecture. BENEFIT: EO based tracking algorithms for security applications, crowd monitoring, consumer characterization, and traffic flow analysis all suffer from limited self diagnostics. Algorithms which are able to predict performance corresponding to specific operating conditions would be able to 1)optimize sensor collection 2)mitigate false alarms 3)allow fuzzy decision making and 4)allocate additional resources when needed.