Multi-source registration using learning and global optimization methods

In this proposal, we describe a novel approach to real-time multi-source geo-registration. We solve the registration problem in conjunction with detection and classification of objects that are used for image registration. Target detection is used for solution validation. The problem is decomposed and solved iteratively using fast learning and optimization algorithms. Proposed innovations include Support Vector machine methods for classification reinforcement and global optimization methods via cooperative team of genetic algorithm agents.