In industrial automation applications, there exist a large variety of different component that can be found inside an industrial switchgear.

These components can differ by Brand Manufactures or typologies but even in size, color and more.

Given these challenges, it is difficult to design a vision system capable to identify these components without requiring the creation of a large dataset and the use of extensive human labelled data.

The proposed vision system exploit the detection of features (e.g. holes on the component’s surface) and match the information with the cad to extrapolate the component’s position to precise localize the components positions.

The detection precision is lower than 1mm and with the proposed method the robots are capable to center holes with dimension lower than 2mm without any effort.