Topic: Hierarchical Framework for Recognition Problems
By: M. Chu (MIT), S. K. Mitter (MIT)
The purpose of the proposed research is to provide a general framework for approaching recognition problems which are characterized by a large network of sensors. An individual sensor provides little information for the given task but the collaboration of a large group of these sensors yields a significant amount of information. We will propose a framework which embodies these recognition problems whose structure incrementally fuses the information of clusters of sensors to larger groups based on the objectives and constraints of the task. The framework will consist of a hierarchy with an associated algorithm which was motivated by attempting to combine the notions of bottom-up and top-down processing. Although this framework was originally derived from computer vision applications, we find that this framework is applicable to the issues of incrementally fusing information in a large network of sensors and handling sensor failures.
Questions and comments to the webmaster. This page is copyrighted by the Massachusetts Institute of Technology.