Cognition-driven Tool for Visual Countermeasure Evaluation

We propose to design a model of visual target acquisition that can predict human performance in detecting low-contrast targets in complex scenes. This model will build on current state-of-the-art computational models of early human vision and visual search, and use proven cognitive modeling methods (belief networks for situation assessment and expert systems for knowledge-based reasoning) to model some of the top-down cognitive effects that come into play in real viewing situations. Recognizing that early vision does not accept symbolic top-down inputs because it does not have the capability to process symbolic data, we will implement the top-down effects as a modulation of various visual functions by changing the values of the free parameters of those functions (such as parameters for relative weighting of processing channels). A limited software prototype will be implemented to demonstrate model feasibility, to be followed up in Phase II with a full scale implementation and validation through psychophysics experiments. The proposed model enables the development of design tools that predict the impact of new designs on operator/system performance, before expensive real-time operator-in-the-loop simulation studies are conducted. Such tools have great potential in designing systems for automatic target cueing; medical image anomaly cueing; car collision avoidance; and highway safety.