Knowledge-Based Vision Techniques for the Autonomous Land Vehicle Program
Abstract
The goal of this research was to develop techniques for automatically acquiring and representing knowledge about complex cultural and natural environments for such purposes as autonomous navigation, planning, intelligence analysis, and manipulation. Our research strategy was to develop representations and techniques for storing (or incrementally learning) semantic and geographic information about a specific geographic area to permit both mission planning and knowledge-based interpretation of sensed data, develop representations for natural and man-made objects, develop techniques to recommend distinctive features of these objects that can be used for recognition purposes, and develop techniques for building three-dimensional descriptions of an environment from data gathered by range or intensity sensors moving through this environment. In this report we describe our accomplishments in these areas. Near the end of the original contractual period a new task was added to evaluate the application of our cartographic stereo techniques to ground-level imagery. We characterized the strengths and weaknesses of the techniques, enhanced them based on the insights provided by the evaluation, and then evaluated the resulting systems. In this report we briefly describe the enhancements and our evaluation process.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 1991
- Accession Number
- ADA259801
Entities
People
- Martin A. Fischler
- Robert C. Bolles
Organizations
- SRI International