Detection of Surface Topographical Anomalies On Glossy Metallic Surfaces Using Machine Vision with Unique Systems Of Illumination
Abstract
With the advent of automation in every aspect of manufacturing, the Quality Control process still remains labor, time-intensive and prone to human errors. This research paperwork describes a machine vision-based system for identification of cracks and other three dimensional topographic anomalies in glossy metallic surfaces using two unique modes of lighting: Cloud Day Illumination(CDI) and Running Zebra Lighting. The proposed system is specific to a critical automotive component named ‘contact pointer’. The system uses industrial-grade machine vision cameras for Image acquisition. The acquired image is pre-processed (filtering, pattern detection, threshold matching) and the defects are detected and classified using deep learning based computer vision algorithms. Under ideal conditions, the system could bring down the labor dependency at a proposed ratio of 4:1 also being more efficient and less error rate than their manual counterparts achieving prediction accuracy of ~90%.