A bayesian approach to textile defect detection problem and a comparative analysis

Loading...
Thumbnail Image

Date

2010.

Journal Title

Journal ISSN

Volume Title

Publisher

Thesis (M.S.)-Bogazici University. Institute for Graduate Studies in Science and Engineering, 2010.

Abstract

The purpose of this thesis was to detect the defects on textile fabric images using the particle filters. We approached the problem from a Bayesian perspective and represented the model in a state space formulation. To describe the state space formulation; the texture models such as linear, 2-D linear and Markov Random Field models and the noise types like Gaussian, mixture of Gaussian and alpha-stable noise are investigated to find the best representation that is appropriate for our textile images. The implementation results are compared with Kalman, Extended Kalman and Unscented Kalman filters. Finally time and performance analysis of the filters is given.

Description

Keywords

Citation

Collections