Machine vision-driven manufacturing automation in the factory of the future

At Prodrive Technologies we believe in automation, automation of everything. Not only for our customers, but also for our own manufacturing facilities. The Prodrive Robotics Solutions (PRS) group develops generic robotic automation cells in-house that can work autonomously and robustly on many different products requiring no manual intervention.

Vision camera systems and algorithms, using 2D and 3D data, are developed for (1) detection and localization of components for automated pick and place actions, and (2) automated visual inspection. Because high-end electronics are manufactured, submillimeter accuracy is required for placement actions. Components that must be detected and localized range from small lenses with a width of just a few millimeters, up to housings of several hundred millimeters.

Automated visual inspections are performed to validate manufacturing processes and to ensure the highest production quality. Inspection steps range from sealant inspection, where dispensed material such as a glue is verified on position and thickness, to presence detection of components and correctness of wire placement. For this last inspection, deep learning algorithms are used to cope with the large variety of input data.

To speed up the integration and commissioning time, design files (CAD) can be used as input of machine vision algorithms as well. These files are used to generate data sets in a simulated environment for training and validation of machine vision and deep learning algorithms without requiring any manufacturing equipment or components.