Publication detail
Image Processing Using Neural Network
ŠKORPIL, V. ŠŤASTNÝ, J.
Czech title
Zpracování obrazu pomocí neuronové sítě
English title
Image Processing Using Neural Network
Type
conference paper
Language
en
Original abstract
This contribution describes the problem of the object classification by using artifical neural network. This is the question of the Back-propagation, Kohonen and ART network algorithms comparison. Back-propagation algorithm is based on minimization of neural network energy, Kohonen network is one of the self-organizing neural networks and characteristic feature of ART network is its ability to switch the variable and stable mode without damaging information learned. The application part includes the total evaluation of given algorithms on the bases of time learning and success in classification of given objects of two-dimensional technologic scene. The real technologic scene was simulated with digitization of two-dimensional pictures of real objects.
Czech abstract
Příspěvek se zybývá problematikou klasifikace objektů pomocí neuronových sítí. Byly porovnány algoritmy Back-propagation, Kohonenovy sítě a sítě ART. Vyvinutá aplikace zahrnuje uvedené algoritmy, aplikované na dané objekty dvoujozměrné technologické scény. Reálná technologická scéna byla simulovaná digitalizací dvourorozměrných obrazů reálných objektů.
English abstract
This contribution describes the problem of the object classification by using artifical neural network. This is the question of the Back-propagation, Kohonen and ART network algorithms comparison. Back-propagation algorithm is based on minimization of neural network energy, Kohonen network is one of the self-organizing neural networks and characteristic feature of ART network is its ability to switch the variable and stable mode without damaging information learned. The application part includes the total evaluation of given algorithms on the bases of time learning and success in classification of given objects of two-dimensional technologic scene. The real technologic scene was simulated with digitization of two-dimensional pictures of real objects.
Keywords in English
Neural Networks, Image Processing, Pattern Recognition
RIV year
2002
Released
05.09.2001
Publisher
Foerster
Location
UTKO VUT Brno
ISBN
80-214-1962-8
Book
Proceedings of the Conference TSP 2001
Edition number
1
Pages from–to
22–27
Pages count
6
BIBTEX
@inproceedings{BUT4168,
author="Vladislav {Škorpil} and Jiří {Šťastný},
title="Image Processing Using Neural Network",
booktitle="Proceedings of the Conference TSP 2001",
year="2001",
month="September",
pages="22--27",
publisher="Foerster",
address="UTKO VUT Brno",
isbn="80-214-1962-8"
}