This machine sorts a set of small aluminum parts that have little variations in size or even the same dimensions but a different number of extremely small holes.
There are no standard shapes, that means that the system has to learn new models at each batch.
Part recognition is performed using blob analysis functions on a backlight table.
The main issues were the way to distribute those small items on the 2D table, and to learn new model at each batch.
One solution was to use a feeder that allows remote control to distribute parts at different speeds and amplitudes.
Our camera unit inspects at each cycle parts distribution and decides if the parts have to be moved in order to allow piece recognition.
Thanks to the flexibility of our camera unit, the vision system communicates perfectly with the different machine components and performs sorting task at a rate of 1 piece/second.