Abstract:
The problem of recognizing defects on the surface of hot-rolled metal is quite old, but technologies have only recently reached a sufficient level for the automation of t...Show MoreMetadata
Abstract:
The problem of recognizing defects on the surface of hot-rolled metal is quite old, but technologies have only recently reached a sufficient level for the automation of this process [1]–[3]. One of the most suitable methods is the application of convolutional neural networks [4]–[7]. As the qualitative dataset for network training we selected the NEU (Northeastern University) surface defect database, which is a dataset of the most identified cases of hot rolled defects. Also, these data were supplemented with images of a clean hot-rolled surface. An algorithm for recognizing defects was developed. Camera parameters for machine vision were calculated.
Published in: 2022 International Conference on Industrial Engineering, Applications and Manufacturing (ICIEAM)
Date of Conference: 16-20 May 2022
Date Added to IEEE Xplore: 09 June 2022
ISBN Information: