Publication detail

Pattern Recognition By Use Of Functions Of Invariants

HORÁK, K. KALOVÁ, I.

Original Title

Pattern Recognition By Use Of Functions Of Invariants

English Title

Pattern Recognition By Use Of Functions Of Invariants

Type

conference paper

Language

en

Original Abstract

Visual system is the basic tool for non-contact inspection in industries. Such system usually has to detect defects and reject faulty products automatically. This article deals with autonomous visual system for glass bottle inspection, which has to detect non-homogenous regions in ROI in images. Additional system is able to distinguish between regions belongs to defect from regions belongs to manmade patterns, so that products can be made in high quality and cost-effective way. Snag is that synthetic patterns (like a word-sign) and defects (dirt, crack etc.) have similar image characteristics, hence it is difficult to recognize between them. In order to solve mentioned task we introduce novel method based on moment invariants. Each detected non-homogenous region is wilfully modified by so called "circle degradation" algorithm in several steps. Moment invariants corresponding to modified regions are then computed and subsequently functions of these invariants are created. On the basis of these functions, both synthetic patterns and defects can be recognized in a usual classification way.

English abstract

Visual system is the basic tool for non-contact inspection in industries. Such system usually has to detect defects and reject faulty products automatically. This article deals with autonomous visual system for glass bottle inspection, which has to detect non-homogenous regions in ROI in images. Additional system is able to distinguish between regions belongs to defect from regions belongs to manmade patterns, so that products can be made in high quality and cost-effective way. Snag is that synthetic patterns (like a word-sign) and defects (dirt, crack etc.) have similar image characteristics, hence it is difficult to recognize between them. In order to solve mentioned task we introduce novel method based on moment invariants. Each detected non-homogenous region is wilfully modified by so called "circle degradation" algorithm in several steps. Moment invariants corresponding to modified regions are then computed and subsequently functions of these invariants are created. On the basis of these functions, both synthetic patterns and defects can be recognized in a usual classification way.

Keywords

recognition, pattern, moments, invariants

RIV year

2008

Released

23.10.2008

Publisher

DAAAM International

Location

Vienna, Austria 2008

ISBN

978-3-901509-68-1

Book

Annals of DAAAM for 2008 & Proceedings of the 19th International DAAAM Symposium

Pages from

605

Pages to

609

Pages count

2

Documents

BibTex


@inproceedings{BUT32578,
  author="Karel {Horák} and Ilona {Janáková}",
  title="Pattern Recognition By Use Of Functions Of Invariants",
  annote="Visual system is the basic tool for non-contact inspection in industries. Such system usually has to detect defects and reject faulty products automatically. This article deals with autonomous visual system for glass bottle inspection, which has to detect non-homogenous regions in ROI in images. Additional system is able to distinguish between regions belongs to defect from regions belongs to manmade patterns, so that products can be made in high quality and cost-effective way. Snag is that synthetic patterns (like a word-sign) and defects (dirt, crack etc.) have similar image characteristics, hence it is difficult to recognize between them. In order to solve mentioned task we introduce novel method based on moment invariants. Each detected non-homogenous region is wilfully modified by so called "circle degradation" algorithm in several steps. Moment invariants corresponding to modified regions are then computed and subsequently functions of these invariants are created. On the basis of these functions, both synthetic patterns and defects can be recognized in a usual classification way.",
  address="DAAAM International",
  booktitle="Annals of DAAAM for 2008 & Proceedings of the 19th International DAAAM Symposium",
  chapter="32578",
  howpublished="print",
  institution="DAAAM International",
  year="2008",
  month="october",
  pages="605--609",
  publisher="DAAAM International",
  type="conference paper"
}