Generalized Similarity Measure for Multisensor Information Fusion via Dempster-Shafer Evidence Theory

Dempster-Shafer evidence theory (DSET) stands out as a mathematical model for handling imperfect data, garnering significant interest across various domains. However, a notable limitation of DSET is Dempster's rule, which can lead to counterintuitive outcomes in cases of highly conflicting evidence. To mitigate this issue, this paper introduces a novel reinforced belief logarithmic similarity measure ( $\mathcal {RBLSM}$ ), which assesses discrepancies between the evidences by incorporating both belief and plausibility functions. $\mathcal {RBLSM}$ exhibits several intriguing properties including boundedness, symmetry, and non-degeneracy, making it a robust tool for analysis. Furthermore, we develop a new multisensor information fusion method based on $\mathcal {RBLSM}$ . The proposed method uniquely integrates credibility weight and information volume weight, offering a more comprehensive reflection the reliability of each evidence. The effectiveness and practicality of the proposed $\mathcal {RBLSM}$ -based fusion method are demonstrated through its applications in target recognition and pattern classification scenarios.