Multi-Source Information Fusion Using a Reinforced Belief Divergence Measure and Negation of a Reconstructed Basic Probability Assignment Matrix
Xun Sun, Jun Long · 2024
Present conflict fusion strategies cannot simultaneously handle the problems of conflict contradiction and fusion efficiency in imperfect world conditions. To efficiently extract information from multi-source data, this study suggests a Multi-Source Information Fusion approach based on Negation of Reconstructed Basic Probability Assignment Matrix and a Reinforced Belief Divergence Measure. The approach uses a Gaussian distribution function to fill the original BPA, preventing overfitting. Next, a reinforced belief divergence measure is used to convert the original BPA into nrBPA, so reducing the evidence conflict. Finally, the nrBPA-Matrix approach is used to calculate the conflict. This method computes the indexes of the components in the evidence to arrive at a more appropriate weight for the evidence. Finally, the strongest weight of evidence is used as the fusion result.