Combination of granules, rough sets with evidence theory and its application in incomplete data fusion for belief estimation

This paper presents an approach to deal with multi sensor data fusion problem in incomplete circumstance using combination of granule idea, rough approximation and evidence theory. It deletes redundant sensors through rough set theory in selecting and reducing features, and forming dominant characters to form various granules. It applies these granules to establish belief functions to get different belief estimates. It extracts decision rules from incomplete system to identify targets. Experiments show this method can overcome slow problem in posing massive data set with fluctuant sensors and prove to be feasible and efficient.