Computational Vision for Generic Target Recognition
Abstract
During the tenure of this grant we have solidified the early stages of edge detection, and a computer code remains publicly available for this. Active interactions with several DoD- and independently-funded industrial organizations have been established. During the course of this grant we focused our research on two directions, one in early vision and the other in intermediate to high-level vision. On the early vision side we developed curve detection algorithms that function in high-noise situations and are extending our differential-geometry-based algorithms into texture, shading, and stereo. We have started to analyze the interactions between shading and edge structures. These latter projects will be continuing. In support of our high-level, shape recognition work, systems have been developed to generate generic descriptions of visual shapes, and these are matched to databases using graph matching algorithms. The mathematics are being extended to deal with subgraph homeomorphisms. Both function within a continuous dynamical systems framework, which has important biological analogs as well as implementation advantages.
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 08, 2002
- Accession Number
- ADA401208
Entities
People
- Steven W. Zucker
Organizations
- Yale University