We present the current state of the Kinetic Visualization Project, a two year research project combining the efforts of BBN Technologies, Prof. Ronald Pickett of UMass Lowell, and Prof. Colin Ware of the University of New Hampshire, and supported by ARDA. The goal of the project is to develop motion-based visualization techniques that will enable analysts to rapidly perceive and understand patterns embedded in high-dimensional and complex data, and to convey their findings effectively to policy makers. While other groups have focused on animation, the presentation of time-varying data using motion, we are interested in the use of motion to explore and understand patterns in complex but static data, such as hyperspectral imagery and sociological, financial and computer link networks.

     A key characteristic of the human visual system is its highly evolved ability to respond to and interpret motion in the visual field. The perception of motion by the human visual system allows people to detect objects obscured by clutter (e.g., a tiger in the grass), to "directly" perceive the forces acting on objects (e.g., the tension in a sail, the wind blowing across a field of grass), and to change their focus of attention quickly to look at details of one object without losing the context provided by the larger field of view (e.g., the actions of the quarterback executing a football play). We have spent the past year developing software and performing experiments that demonstrate the effectiveness of visualization techniques that exploit the human perceptual response of motion-induced grouping to address key problems associated with exploring massive, complex, high-dimensional data sets.

     This talk will present the results of the first year of this project, including demonstrations of 1) the use of two dimensional motion in link-analysis, 2) understanding relations between signature plots and overhead images by motion brushing of hyperspectral data, and 3) the use of three dimensional motion to make it possible to simultaneously visualize and understand patterns defined by eight dimensions of data in hyperspectral imagery.