Bayesian Feature Fusion Using Factor Graph in Reduced Normal Form

In this work, we investigate an Information Fusion architecture based on a Factor Graph in Reduced Normal Form. This paradigm permits to describe the fusion completely probabilistic framework and information related different features are represented as messages that flow network. way build sort of context for observed conferring solution great flexibility managing type with wrong missing values required by many real applications. Moreover, modifying opportunely into network, obtain effective condition inference reliability each source or presence single unreliable signal. The proposed has been used fuse detectors identity document classification task but its flexibility, extendibility robustness make it suitable scenarios where signal can be wrongly received missing.