Neurophysiological Based Methods of Guided Image Search
Abstract
Report developed under SBIR contract for topic NIMAO2-OO1 Complex analysis of intelligence imagery is crucial to the missions of intelligence organizations, yet remains constrained by labor-intensive, time-consuming visual search of large volumes of imagery. Many algorithms have been developed to automatically identify regions of interest in large, complex sets of imagery, yet the utility of such algorithms is limited by the fact that human analysts detect features in imagery with higher accuracy than existing methods. We developed a model of visual feature detection, the Neuronal Synchrony Model, based on neurophysiological models of temporal neuronal processing, to improve the accuracy of automatic detection of features of interest in complex natural imagery. The Neuronal Synchrony Model of image feature detection was applied to accurately identify and highlight regions of images that contain target features, thus automating the labor-intensive, "scanning" portion of imagery analysis. The accuracy of the Neuronal Synchrony Model was tested with natural images containing visually controlled, synthetic targets as well as with natural targets using a variety of overhead imagery background and target types. A proof-of-concept demonstration of the effectiveness of this model showed enhancement in the speed and accuracy of interactive, guided visual search of particular classes of imagery.
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 10, 2003
- Accession Number
- ADA414960
Entities
People
- Frank M. Marchak