Distributed hierarchical decision fusion with cluster validity

The decision problem that arises in complex engineering problems where many separate sensors have to be used in parallel to measure a large set of objects, such as target detection/recognition, is addressed. A distributed hierarchical decision fusion method based on the validity of fuzzy clustering is proposed. Each sensor makes a local decision using a fuzzy clustering method and sends its decision about the number of clusters and the validity of the decision about the number of clusters and the validity of the decision, measured using a well-defined criterion, to the central station. The central station fuses the decisions collected from each sensor and makes a team decision optimizing all the local decisions. After clustering based on the optimal feature, the process is repeated until all object are properly classified. The result is a hierarchical decision tree that represents a description of a set of a priori unknown objects. This method has been applied to scenes where each object appears in a specified region