A novel combination approach of evidence based on Pignistic probability distance

Abstract

DS theory is widely used in the fields of decision-level information fusion. To suppress the counter-intuitive results such as "one vote rejection" and "trust bias" encountered when utilizing DS theory of combination, as well as the conflict issue between elements within the evidence, a novel method for effectively handling with conflicting evidence is proposed. The method first introduces the representations of self-conflict and mutual-conflict to characterize conflicts in evidence combination. And then according to Pignistic probability distance, an algorithm for correcting self-conflict based on the simulated annealing adaptive algorithm is designed. A correction method for mutual-conflict is next presented on the basis of Pignistic similarity. Finally, the combination result can be obtained utilizing the DST combination rule. Some numerical examples provided show the efficiency and rationality of the proposed method.

Keywords

• target recognition|ds theory|evidence combination|pignistic probability distance|simulated annealing adaptive algorithm