Information Fusion Methods

Information Fusion Methods

No universal method is both able to fuse these two elements and to deal with all possible cases. However, it is possible to find a large range of schemes suiting each case.

Let us examine each case to assess the different methods to apply:

1.: First, let us assume that the two sources are reliable. We can assume it as long as sets E₁ and E₂ intersect widely and in particular if E₁ = E₂. In this case, there is a consensus and we can reasonably think that $x \in E_1 \cap E_2$ .
2.: If the two sources do not agree, i.e. if $E_1 \cap E_2 = \varnothing$ , then the reliability hypothesis is no longer credible. This suggests that one source is wrong, and the other is reliable. But which is the reliable source? We do not know which one is reliable and which one is wrong. As a precaution, all the information is kept and we assert $x \in E_1 \cup E_2$ .

It is obvious that the first information fusion method, $x \in E_1 \cap E_2$ , is the most informative because the information is refined to the intersection of sets given by each source. It is also the most ``risky'' approach because the real value of x is assumed to be inside a smaller set than the two initial sets.

The second method, $x \in E_1 \cup E_2$ , is more reliable since all the information given by the two sources is preserved. The drawback of such an approach is a loss of accuracy since the set assumed to contain x, is larger than each of the initial sets.

The fusion of uncertain information is equivalent to finding a compromise between a too accurate result which is certainly false and a sure result which is too imprecise.

In the following part, we will summarize some basic concepts about fuzzy logic and possibility theory. Then, the property of the operation applied to fuzzy sets ([Klir and Yuan, 1995]) will be presented.

We will introduce then two adaptive fusion methods which use a method of fusion (conjunctive and disjunctive) which is really suited to the kind of information to aggregate.

At last a fuzzy priority rule is also given, where preference is given to the information delivered by a very reliable source rather than information delivered by less reliable sources.

These last three information aggregating operators have been proposed by Dubois and Prade [Dubois and Prade, 1994b] [Dubois and Prade, 1994a].

IRIT-UPS