Announcement
Starting on July 4, 2018 the Indonesian Publication Index (IPI) has been acquired by the Ministry of Research Technology and Higher Education (RISTEKDIKTI) called GARUDA Garba Rujukan Digital (http://garuda.ristekdikti.go.id)
For further information email to portalgaruda@gmail.com

Thank you
Logo IPI  
Journal > Jurnal Ilmiah Ilmu Komputer > Probalilistic Neural Network Based on Multinomial Model and EM Algorithm In Classification, Fusion and Change Detection Contex of Optical and SAR Images

 

Jurnal Ilmiah Ilmu Komputer
Vol 3, No 2 (2005): Jurnal Ilmiah Ilmu Komputer
Probalilistic Neural Network Based on Multinomial Model and EM Algorithm In Classification, Fusion and Change Detection Contex of Optical and SAR Images
Article Info   ABSTRACT
Published date:
18 May 2010
 
This paper presents the results of our continuing study on multidate multisensor image classification. In our previous study, we have recommended a neuro-statistical scheme in the framework of multitemporal optical-sensor image classification. The scheme consists of probabilistic nerual network (PNN) classifier to compute the posterior probabilities, expectation maximization (EM) method to optimize prior joint probabilities, and compound probabilities to produce thematic image and change image. This paper reports the results of extending the scheme for multidate multisensor image classification.For each sensor image classifier, two schemes have been evaluated. The first scheme has used the co-occurrence matrix texture feature images or original tonal images as the input data and the Gaussian kernel for the PNN classifier. The second scheme has used the original tonal image as the input data and the multinomial co-occurrence matrix kernel for the PNN classifier. The results are also compared to the use of back propagation (BP) classifier. Based on this study we have proposed a scheme for multidate multisensor image classification. Keyword : Probablistic Neural Network, Multinomia Model, Expectation Maximization
Copyrights © 2010