Evolutionary Design of Local Binary Pattern Feature Shapes for Object Detection

Czech title

Evoluční návrh geometrických tvarů příznaků Local Binary Pattern pro rozpoznávání objektů

Type

conference paper

Language

English

Authors

Kadlček Filip, Ing., Ph.D.
Fučík Otto, doc. Dr. Ing. (DCSY)

Keywords

Local Binary Pattern (LBP), AdaBoost, Evolutionary design, feature shapes

Abstract

This paper deals with the evolutionary design of application specific feature shapes of Local Binary Pattern (LBP) features for object detection in image processing applications. LBP features are very often utilized in image classification systems which are used for pattern recognition. By using genetic algorithm the application of specific weak classifiers' feature shapes, which are highly optimized to achieve a better accuracy of the AdaBoost strong classifier, are being evolved.

Annotation

This paper deals with the evolutionary design of application specific feature shapes of Local Binary Pattern (LBP) features for object detection in image processing applications. LBP features are very often utilized in image classification systems which are used for pattern recognition. By using genetic algorithm the application of specific weak classifiers' feature shapes, which are highly optimized to achieve a better accuracy of the AdaBoost strong classifier, are being evolved.

Published

2012

Pages

1–8

Proceedings

2012 NASA/ESA Adaptive Hardware and Systems (AHS-2012) Conference

Series

CFP1263A-USB

ISBN

978-1-4673-1914-0

Publisher

IEEE Computer Society

Place

Nuremberg

BibTeX

@inproceedings{BUT91499,
  author="Filip {Kadlček} and Otto {Fučík}",
  title="Evolutionary Design of Local Binary Pattern Feature Shapes for Object Detection",
  booktitle="2012 NASA/ESA Adaptive Hardware and Systems (AHS-2012) Conference",
  year="2012",
  series="CFP1263A-USB",
  pages="1--8",
  publisher="IEEE Computer Society",
  address="Nuremberg",
  isbn="978-1-4673-1914-0",
  url="https://www.fit.vut.cz/research/publication/9987/"
}