Implementation and potentials of a machine vision system in a series production using deep learning and low-cost hardware
Hubert Würschinger, Matthias Mühlbauer, Michael Winter, Michael Engelbrecht, Nico Hanenkamp
Abstract
For manufacturing processes there is a need to ensure an efficient production and to fulfill the increasing quality requirements. To handle these challenges, Machine Vision Systems can be used for process monitoring and quality control. In this paper the implementation and thereby the potentials of such a system in a series production using Transfer Learning with low cost hardware is introduced. The necessary steps, from the hardware implementation, the data acquisition, the preprocessing over the optimization and the application are depicted. Finally we show that the proposed solution can fulfill defined requirements and can compete with a professional Machine Vision System.