OPTICAL CHARACTER DETECTION ON INDUSTRIAL PRODUCTS

1 PASKER Vít
Co-authors:
1 HLAVICA Robert 1 GRYCZ Ondřej 1 ŠPAČKOVÁ Hana 1 MENŠÍK Martin
Institution:
1 VSB - Technical University of Ostrava, Ostrava, Czech Republic, EU, vit.pasker@vsb.cz
Conference:
28th International Conference on Metallurgy and Materials, Hotel Voronez I, Brno, Czech Republic, EU, May 22nd - 24th 2019
Proceedings:
Proceedings 28th International Conference on Metallurgy and Materials
Pages:
1820-1824
ISBN:
978-80-87294-92-5
ISSN:
2694-9296
Published:
4th November 2019
Proceedings of the conference were published in Web of Science and Scopus.
Metrics:
34 views / 15 downloads
Abstract

The article deals with the acquisition of character positions from the input image and with modifications to increase the accuracy of recognition that precedes it. As the input image is meant a camera output or a photo of text information in the surface on different types of blanks. The input image itself contains many defects, for example noise, holes, light reflection, uneven surface caused by cutting or welding etc. All of these defects have to be eliminated to increase the accuracy of the algorithms for image recognition. In the article, there are described methods of improving the input image, which are suitable for all types of text, but it is mainly aimed at industrial products and materials that exhibit a large percentage of defects caused by metallurgical processes.

Keywords: Image recognition, image filters, blank defects, binarization, character detection.
Scroll to Top