A Vision-Based Framework For Enhanced Quality Control In A Smart Manufacturing System

Dimensional metrology is an integral part of quality control in manufacturing systems. Most existing manufacturing systems utilize contact-based metrology, which is time consuming and not flexible to design changes. There have been recent applications of computer vision for performing dimensional metrology in manufacturing systems. Existing computer vision metrology techniques need repeated calibration of the system and are not utilized with data analysis methods to improve decision making. In this work, we propose a robust non-contact computer vision metrology pipeline integrated with Computer Aided Design (CAD) that has the capacity to enable control of smart manufacturing systems. The pipeline uses CAD data to extract nominal dimensions and tolerances. The dimensions are compared to the measured ones, computed using camera images and computer vision algorithms. A quality check module evaluates if the measurements are within admissible bounds and informs a central controller. If a part does not meet a tolerance, the central controller changes a program running on a specific machine to ensure that parts meet the necessary specifications. Results from an implementation of the proposed pipeline on a manufacturing research testbed are given at the end.