The proposed solution had to be developed as a prototype for sorting all the objects that left the extracting machine into up to four categories, while ensuring a minimum production of 75 kg of arils per hour.
Computer vision algorithms had to assess the colour of each object individually, and had to be capable of discriminating between arils and other undesired material, mainly composed of fragmented skin and internal membranes. The control algorithms had to synchronise image acquisition from two different cameras; analyse the images and the displacement of the objects on the conveyor belts; and activate the sorting unit.