An Improved Method of Combining Conflict Evidences

摘要

Dempster-Shafer evidence theory is used extensively in multi-sensor data fusion. How to effectively combine highly conflict evidences becomes one of the hot points in this field. The previous study shows that the combination of conflict coefficient K with pignistic distance is a feasible solution. However, pignistic distance has some drawbacks which may affect its real application. In this paper, we propose an improved combination method. We first analyze the disadvantages of pignistic distance and bring in evidence distance d_J combined with conflict coefficient K which can efficiently represent the conflict degree between evidences. Then, we propose a new combination rule based on the representation of conflict. Finally, we use several experiments to illustrate the validity of the proposed method. The result shows that our model can outperform the existing approaches: while the conflict between the evidences is small, the result is close to Dempster combination rule; while the conflict between the evidences is large, the result of the proposed method is more reasonable.