Strong Conflict

Strong Conflict

With at least two sources giving a little agreeing among all those available (figure 33), areas where the value of the parameter is more certain than elsewhere (area C), are sought.

Renormalized conjunctive fusion makes it possible to give more weight to these areas. However, the approach is still disjunctive outside areas C because there is always a strong conflict between all the other sources. The mode of fusion remains mainly disjunctive.

**Figure 33:** Quantified adaptive fusion of four sources in strong conflict.
$\begin{figure}\begin{center}\fbox{\epsfbox{c3-adq2.eps}}\end{center}\end{figure}$

In example 33, there are two areas C for which two sources agree:

it concerns intersection of distributions $\pi_1$ and $\pi_2$ on the one hand, and
$\pi_3$ and $\pi_4$ on the other hand.

In this case, preference is given to the area where the height of the intersection is greatest, i.e. the area for which consensus is highest.

Here h(n) can be at

0.2 (intersection height of $\pi_1$ and $\pi_2$ ), or
0.1 (intersection height of $\pi_3$ and $\pi_4$ ).

But h(n) is set at 0.2 (height of intersection of distributions $\pi_1$ and $\pi_2$ ) because it is the most plausible area on the one hand, and it is necessary to remain within the limits of interval [ 0, 1 ] for renormalization of $\pi_n(\omega)$ by h(n) on the other hand. As renormalization of conjunctive fusion is carried out with h(n) = 0.2, intersection of $\pi_1$ and $\pi_2$ is favoured compared to that of $\pi_3$ and $\pi_4$ .

IRIT-UPS