This system was developed to determine the exact banana colour ripeness with only a tablet computer. This customer has a banana riping facility where the colour of the ripened bananas is crucial to send them to the stores.
Classification of colour is subject to the human perception. This is not just classifying between green or yellow. This is selecting between 7 colourstages from yellowish green up to greenish yellow. Only trained quality engineers can see the difference. These quality engineers already carry tablet computers with them so this would be ideal to do the classification.
Colour classification is quite hard, but coping with numerous lighting conditions is even harder. We developed a way to correct the image so that there is almost no influence of the ambient light, whatever the source might be. This gives the advantage that the operator can do a classification in a poorly lit banana riping cell or in a very bright logistics hall with lots of fluorescent lights. The classification result will be the same when classifying a box of bananas in both lightingconditions.
Since holding a tablet computer in your hand is not quite the same as a static highly conditioned calibrated machine vision system, we had to make sure that the image is as good and as conditioned as possible. But can you do that when every operator has a different bodylength, has other 'dynamic behaviour'? We developed a procedure that makes sure that every image is taken at the same height with the box of bananas centered and aligned, with the correct focus and the right brightness.