Stainless Steel Slab Surface Defect Detection Using High-Definition Images and Machine Learning Tools for Classification
(Room 204)
09 May 17
10:00 AM
-
10:30 AM
Detecting defects upon the creation of the slab at continuous casting can avoid future product degradation and product losses. If the defect is detected early, countermeasures can be taken to fix it in time. Ultrahigh-resolution images on both sides of the slab are processed at high speed, taking advantage of modern GPU and parallel programming. A graphic analysis tool was designed to be used by a plant process engineer to classify defects, and to create a set of images meant to train a deep neural network, which achieves impressive recognition rates, for on-line automatic defect detection and classification.