Landscan: Query Driven Recognition System
Abstract
LandScan was intended to be an interface not to a database but to an active (and interactive) visual recognition system. That is, rather than searching a body of exiting facts about the domain, the system drives a vision component that will process data supplied to it by two cameras and respond with identification and analysis of objects found in the scene. We currently are looking at a scale model of a city block that is part of the University of Pennsylvania campus. Obviously, knowledge about language, the world and visual properties of objects is needed for this, and will have to reside in the various components of the system, but the system will be gathered in response to the user's requests.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 1988
- Accession Number
- ADA205040
Entities
People
- Ruzena Bajcsy
Organizations
- Moore School of Electrical Engineering