A Scientific Inquiry fusion theory for high-level information fusion

摘要

The Joint Directors of Laboratories fusion model is adequate as a functional description, but falls short as a formal design guide for the application of logical inference under uncertainty for high-level information fusion. We propose a formal construct called the Scientific Inquiry Fusion Theory, with three stages of explanation, prediction and generalisation aligned with the corresponding inferences of abduction, deduction and induction. We first define fusion as formal models without uncertainty, in which the corresponding logical inference patterns can be used for solving fusion problems. Then, we extend these formal models with uncertainty through probabilistic graphical models, where fusion processes are realised by statistical queries and learning algorithms based on the sound unification of Probability Theory, Mathematical Logic and Machine Learning. Finally, we demonstrate the application of this formal high-level information fusion framework with an example of automated maritime security situation and threat assessment.