An Evidence Fusion Method with Importance Discounting Factors based on Neutrosophic Probability Analysis in DSmT Framework

Qiang Guo; Haipeng Wang; You He; Yong Deng; Florentin Smarandache

Authors

Qiang Guo Institute of Information Fusion Technology Department Naval Aeronautical and Astronautical University Yantai, China
Haipeng Wang Institute of Information Fusion Technology Department Naval Aeronautical and Astronautical University Yantai, China
You He Institute of Information Fusion Technology Department Naval Aeronautical and Astronautical University Yantai, China
Yong Deng School of Computer and Information Science Sourthwest University, Chongqing, China
Florentin Smarandache Math. & Sciences Dept., University of New Mexico, Gallup, U.S.A.

Keywords:

Information fusion;, Belief function;, Dezert-Smarandache Theory;, Neutrosophic probability;, Importance discounting factors.

Abstract

Abstract: To obtain effective fusion results of multi source evidences with different importance, an evidence fusion method with importance discounting factors based on neutrosopic probability analysis in DSmT framework is proposed. First, the reasonable evidence sources are selected out based on the statistical analysis of the pignistic probability functions of single focal elements. Secondly, the neutrosophic probability analysis is conducted based on the similarities of the pignistic probability functions from the prior evidence knowledge of the reasonable evidence sources. Thirdly, the importance discounting factors of the reasonable evidence sources are obtained based on the neutrosophic probability analysis and the reliability discounting factors of the real-time evidences are calculated based on probabilistic-based distances. Fourthly, the real-time evidences are discounted by the importance discounting factors and then the evidences with the mass assignments of neutrosophic empty sets are discounted by the reliability discounting factors. Finally, DSmT+PCR5 of importance discounted evidences is applied. Experimental examples show that the decision results based on the proposed fusion method are different from the results based on the existed fusion methods. Simulation experiments of recognition fusion are performed and the superiority of proposed method is testified well by the simulation results.