Image-Based Flaw Detection and Process Life In Production Products Using Deep Learning

Abstract

The identification of flaws in manufactured goods is a crucial part of quality control in production. The manufactured products must satisfy two things viz., the industry specifications and customer specifications and accordingly the defects are classified. Defects through human inspection is too demanding and may always calls for an additional defect known as overlook defects. The advent of machine vision and learning made the human efforts simpler and the journey from human’s to machines for quality control is proved to be a successful learning curve and highly demanding as well. The technique of machine vision and image annotation brought the change required in the field of quality control. This paper provides an insight of deep learning techniques for defect identification with a case study. To begin, defects and its significance is brought forward that might appear in manufactured products and then the characteristics, benefits, and shortcomings of customized approaches are put forth. The article, outlines the fundamental concepts and methodologies of deep learning with machine vision approach such that defects in manufacturing can be identified. A fractional order singular value decomposition (FSVD) algorithm with deep neural networks is adopted for defect detection and python platform is used for its implementation