关键词: 信息融合, DempsterShafer理论, mass函数, 置信度, 衰减函数, 证据组合

Abstract: The localizability and credibility of distributed evidences are considered and, as a result, it is found that the reliability of an evidence weakens with increasing distance even if its source is reliable, and that a paradox is often possible if basing an evidence on the conventional D-S theory. The problem has been ignored until now and still regarded as that the evidence is reliable. To solve the problem, a distance function is used to estimate the validity of an evidence and it is further introduced in the mass function so as to express how the validity of the evidence changes. Then, the original evidences coming from the observing points at different distances are converted so as to combine them together, and the combination rules are compared with those in conventional D-S theory through experiments. The results show that the way proposed here has higher adaptability than conventional D-S rules for the possibility estimate of the extensive area outside the distributed observing points.