An evidential sine similarity measure for multisensor data fusion with its applications

Abstract

It remains challenging in managing uncertain and imprecise information in multisensor data fusion. Dempster–Shafer evidence theory (DSET), which has a strong appeal for modeling uncertainty and imprecision, has received tremendous popularity. Whereas, highly conflicting pieces of evidence often lead to counterintuitive results of Dempster’s rule. To address this issue, we propose a new evidential sine similarity measure ($\mathcal{E}\mathcal{S}^2\mathcal {M}$) that reasonably calculates discrepancies between the pieces of evidence by considering the Pignistic probability transform. Furthermore, we prove that the $\mathcal{E}\mathcal{S}^2\mathcal {M}$ satisfies properties of bounded, symmetry, and non-degeneracy. On top of $\mathcal{E}\mathcal{S}^2\mathcal {M}$, we develop a new multisensor data fusion method, which considers both the credibility and the information volume to reflect the importance of each piece of evidence. Finally, we demonstrate the effectiveness and rationality of the proposed method in target recognition and pattern classification applications.