Object Cueing Using Biomimetic Approaches to Visual Information Processing

Period of Performance: 09/09/2014 - 04/09/2015

$79.9K

Phase 1 STTR

Recipient Firm

Soar Technology, Inc.
3600 Green Court Array
Ann Arbor, MI 48105
Principal Investigator
Firm POC

Research Institution

University of Central Florida
4000 Central Florida Boulevard
Orlando, FL 32816
Institution POC

Abstract

Understanding imagery from unmanned automated systems in a timely fashion requires support systems for end users that can filter and preprocess massive data via complex vision and understanding. A new computer system that mimics both the bottom-up biological and top-down cognitive processes of the human visual system will provide breakthrough decision support for immediate imagery analysis. We propose FOCVS (pronounced focus ), the Full context Onboard Cognitive Visual System. In FOCVS, top-down cognitive perception based on context such as user goals and mission parameters will direct low-level visual components to search efficiently by defining areas of attention, target features, and desired level of detail. Bottom-up processing will iteratively search the live data in real time, processing scenes in the full context of multiple visual and nonvisual inputs through unified processing of space-centered vision. Finally, the cognitive component will interpret reports from the neural component and apply intelligent filtering to create an actionable understanding of the environment for the human user. The result of our research will be a decision aid that adapts its execution based on the goals of a user who is in direct contact with the UAS and the mission, enabling timely and responsive first-pass imagery understanding.