The fuzzy logic or how to compare apples with pears

by Jochen Strahl on 28.03.2008

It is sometimes difficult for machines: facts which are simple for humans have to be examined and verified by them repeatedly at first. Thus containers which are similar can be differentiated simply and clearly by humans but conventional inspectors can only examine certain features such as shape and colour. However HEUFT uses fuzzy logic in order to obtain the correct overall result from the individual results.

It is no problem for a human to tell the difference between an apple and a pear

Let us compare apples with pears as an example. A machine has to take several measurements and make a comparison with a reference sample in order to determine the type of fruit. The size, shape and colour of the reference apple correspond more closely to the pear than the small apple. It is solely the measurement of the upper area which gives away the pear. On the other hand the size, shape and colour of the small apple are considerably below the values of the reference apple but the small apple comes off better as regards the measurement of the upper area.

One could now compare the results and say: the pear meets the requirements of an apple in three out of the four criteria – in other words it is an apple. The results of the small apple are below the reference values in all areas without falling out of the framework of any measurement.

The human can identify an individual apple relatively certainly by means of shape detection. However the human will not be able to reliably remove the pears from apples which pass him at 50,000 apples per hour because he is overtaxed by the speed. The machine on the other hand knows how to deal with this but the human has to programme it so that it works with methods which are as similar to the human brain as possible in order to achieve an optimal detection accuracy.

HEUFT achieves this by using fuzzy logic where individual criteria such as diameter, height and colour are not evaluated as good or bad. Each measurement is given a percentage here which expresses to what extent the available fruit meets the requirements of an apple. A total probability is calculated from all the measurements using a complex algorithm. The object which is looked at is not of the same type as the sample if this value does not exceed the 50% level – it is not an apple in our example.

It is the same in the case of the bottle material used during production. There are many similar containers but they do not meet the requirements of the current production under any circumstances. These have almost the same attributes as the current brand and therefore the fuzzy logic has to separate the wheat from the chaff without rejecting any "good" bottles.

The HEUFT inspectors have been using this type of verification successfully for many years now. This is possibly one of the reasons why HEUFT stands out against its competitors by a long way particularly in this field or to say it in fuzzy logic terms: it is extremely probable that HEUFT fulfils all the criteria and provides the best inspection result. Make a comparison!

Did you find this report helpful??
not very