Technical Objectives

Calibration-free augmentation of live video streams , without knowing anything about the geometric content of the scene, using a non-Euclidean mathematical framework that has been the focus of some recent computer vision research. The effect is to see graphic renditions of objects, activities or natural features in their appropriate location in a live video stream in which they may be physically obscured or invisible. A hierarchical approach to non-Euclidean spatial representation can break through current augmented reality limitations on spatial scalability, allowing freedom of cooperating viewpoints from close ground observation to aerial views.
Adaptive, predictive, context-based tracking , that can use whatever geometric and dynamical target properties are known to maintain accurate track and thus support reliable and accurate view registration.
Robust recognition of natural, human, and mechanical activities, using spatio-temporal features and fairly standard pattern recognition techniques. Activity recognition can focus attention, describe content, and ignore distracting clutter.

Back to Main Page

This page is maintained by Mike Van Wie.

Last update: 11/11/96.