Hypothesis Integration in Image Understanding Systems.
Abstract
The goal of this research is to develop a robust control strategy for constructing image understanding systems (IUS). This paper proposes a general framework based on the integration of 'related' hypotheses. Hypotheses are regarded as predictions of the occurrences of objects in the image. Related hypotheses are clustered together. A 'composite hypothesis' is computed for each cluster. The goal of the IUS is to verify the hypotheses. We constructed an image understanding system, SIGMA, based on this framework and demonstrated its performance on an aerial image of a suburban housing development. Keywords include: image understanding systems, SIGMA, and robust control strategy.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 1985
- Accession Number
- ADA160129
Entities
People
- L. S. Davis
- T. Matsuyama
- V. S. S. Hwang
Organizations
- University of Maryland