Paperin karakterisointi konenäöllä (Machine vision based paper characterization). M, T. Ph.D. Thesis, 2002.
abstract   bibtex   
Paper characterization and automatic quality control are very demanding applications of machine vision. Different papers are visually highly similar even though they do not belong to the same quality class. This causes several requirements to the quality control system. Additionally, due to the large production capacity of a paper machine, quality control system must handle huge amounts of data. Fast processing of measured data is important in achieving a reliable and user-friendly quality control system. In this thesis, a non-supervised learning and clustering based approach for classifying papers that differ in their quality is presented. The system is simple to train and start using. Some texture features of paper images are calculated and images are clustered to a self-organizing map according to these features. Clustering is made in a non-supervised fashion and the need for human involvement in training is minimal. The method offers a very fast classifier and also a self-intuitive visual user interface. The approach has been tested with images that are taken from papers with different quality. Tests were made by comparing the resolution and classification accuracy of different kinds of texture features. Also the accuracy of previously used methods in paper inspection were studied and their results were compared to the new methods used in this thesis. Result of the work bare that the use of advanced texture features and non-supervised clustering are very effective and flexible to use in paper characterization. Local binary patterns (LBP) gave even 40 times better classification accuracy compared to the old methods. An interesting result was also that LBP features suite well to be used with self-organizing maps and non-supervised learning.
@phdthesis{
 title = {Paperin karakterisointi konenäöllä (Machine vision based paper characterization).},
 type = {phdthesis},
 year = {2002},
 id = {2608cbab-1ffe-35e1-8855-d5ec13435a0f},
 created = {2019-11-19T13:01:15.518Z},
 file_attached = {false},
 profile_id = {bddcf02d-403b-3b06-9def-6d15cc293e20},
 group_id = {17585b85-df99-3a34-98c2-c73e593397d7},
 last_modified = {2019-11-19T13:46:04.155Z},
 read = {false},
 starred = {false},
 authored = {false},
 confirmed = {true},
 hidden = {false},
 citation_key = {mvg:376},
 source_type = {mastersthesis},
 notes = {M.Sc. thesis, Department of Electrical and Information Engineering, University of Oulu, Finland, 71 p + App.},
 private_publication = {false},
 abstract = {Paper characterization and automatic quality control are very demanding applications of machine
vision. Different papers are visually highly similar even though they do not belong to the same
quality class. This causes several requirements to the quality control system. Additionally, due to
the large production capacity of a paper machine, quality control system must handle huge amounts
of data. Fast processing of measured data is important in achieving a reliable and user-friendly
quality control system.

In this thesis, a non-supervised learning and clustering based approach for classifying papers that
differ in their quality is presented. The system is simple to train and start using. Some texture
features of paper images are calculated and images are clustered to a self-organizing map according
to these features. Clustering is made in a non-supervised fashion and the need for human
involvement in training is minimal. The method offers a very fast classifier and also a self-intuitive
visual user interface.

The approach has been tested with images that are taken from papers with different quality. Tests
were made by comparing the resolution and classification accuracy of different kinds of texture
features. Also the accuracy of previously used methods in paper inspection were studied and their
results were compared to the new methods used in this thesis. Result of the work bare that the use of
advanced texture features and non-supervised clustering are very effective and flexible to use in
paper characterization. Local binary patterns (LBP) gave even 40 times better classification
accuracy compared to the old methods. An interesting result was also that LBP features suite well to
be used with self-organizing maps and non-supervised learning.},
 bibtype = {phdthesis},
 author = {M, Turtinen}
}

Downloads: 0