Indication fo conflict in data fusion

摘要

In possibility measure theory, several adaptatives fusion rules exist. All need a way to evaluate the similarity (or conflict) between the sources. Generally, thoses indications of conflict do not take into account the entire distributions and.or are time consuming as far as computation is concerned. After a review of the most popular similarity evaluating schemes, two new expressions are proposed. Some properties are investigated and experimental results are shown graphically. The notion of more pessimistic indication is explained. As a start point for further work, we prove that every existing adaptative data fusion rule which depends on the conflict cannot be commutative.