An Improved Multi-Source Data Fusion Method Based on a New Divergence Measure

Abstract

Nowadays, due to its efficiency in processing conflict and uncertainty information, multi-sensor data fusion method has received extensive attention in practical applications. Dempster-Shafer evidence theory (DSET) plays a vital role in solving complex and imprecise information problems. However, when dealing with highly conflicting evidence, DSET will produce an unreasonable result between distinct basic probability assignments (BPAs). Recent researches indicate that divergence measure is a valid method with similarity degree and information volume to confirm the weight of evidence. As a result, how to solve the counterintuitive result with the weight model has become an open question. In this article, an improved multi-source data fusion method model with a novel divergence measure was proposed. Within the framework of DSET, the proposed divergence measure method has been demonstrated to satisfy various expected characteristics of divergence measure. In addition, we combine the proposed fusion model and the divergence measure to preprocess the data to obtain the final weights. Compared with the existing experimental results, the proposed fusion model outperforms other correlation models, giving the highest belief value 99.38% to the right target.