Geometric PDE's and Invariants for Problems in Visual Control Tracking and Optimization
Abstract
In this proposal, we considered certain invariant flows to treat a number of key issues in controlled active vision, in particular visual tracking. The flows are all derived from some variational principle and so are physically very well justified. In fact, many of the partial differential equations (PDE's) in imaging are based on curvature driven flows from interfacial physics. They have been shown to be useful for a number of applications including crystal growth, flame propagation, and computer vision. We have extensively studied the problems of optimal transport and optical flow for problems in tracking. Optimal transport has appeared in econometrics, fluid dynamics, automatic control, transportation, statistical physics, shape optimization, expert systems, and meteorology. In particular, for the general visual tracking problem in controlled active vision, a robust and reliable object and shape recognition system is of major importance. We have based a new approach to this problem on the theory of optimal mass transport.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 03, 2005
- Accession Number
- ADA428955
Entities
People
- Allen R. Tannenbaum
Organizations
- Georgia Tech