previous    up   next

Algorithm


We use a Mycin-like uncertainty management but we include a factor imp which defines a relative importance of expert knowledge information source and grey levels image information source:

The algorithm is described below:

IF ( tex2html_wrap_inline1367 AND tex2html_wrap_inline1369 ) THEN

    tex2html_wrap_inline1373

ELSE IF ( tex2html_wrap_inline1375 AND tex2html_wrap_inline1377 ) THEN

      tex2html_wrap_inline1381

    ELSE IF ( tex2html_wrap_inline1385 AND tex2html_wrap_inline1387 ) THEN

    displaymath1389

      ELSE tex2html_wrap_inline1393

This algorithm is summarized by the following figure :




(for tex2html_wrap_inline1341 values -1, tex2html_wrap_inline1399 , tex2html_wrap_inline1401 , +1, giving tex2html_wrap_inline1405 for all possible three values imp, tex2html_wrap_inline1245 , tex2html_wrap_inline1257 ).

If more than one production rule can be activated for an identical conclusion (on class i for example), the conclusion will be assigned a combined certainty factor computed by the following commutative and associative operation.


For example :

Application of rule 1 gives class i ( tex2html_wrap_inline1415 ) and application of rule 2 gives class i ( tex2html_wrap_inline1419 )
then the combined certainty factor tex2html_wrap_inline1341 will be :


displaymath1423

displaymath1425

displaymath1427

displaymath1429

displaymath1431

At the end of the process, each point is classified in each class with updated certainty factors, so the final classification can take place.


      previous    up   next     
  
 IRIT-UPS