Design and Implementation of an Embedded Vision System for Industrial Inspection

Abstract

Nowadays, the advantages offered by image processing and deep learning increased their efficiency popularity. Thus, vision systems are widely motivating researchers to develop new protocols and features to optimize existing ones. Of course, technical challenges do not lack since the integration of image acquisition and processing units industrial environment poses considerable problems. In context, we adopted in our work the hybrid approach combining hardware design and software development. This approach makes the system compact, robust and reliable, especially in industrial field to ensure several operations quality inspection and verification. The proposed solution is to design an industrial embedded vision system that matches scalable hardware architectures to adaptable algorithms. this paper, we propose an efficient model to automate quality control in an industrial production line. This work aims to integrate the concept of the multi-tasking image processing in the manufacturing field by offering a whole pack of various inspection operations, from surface to dimensional inspections, based on simple hardware implementations, optical setups, and deep learning algorithms.