Title :
A novel contextual classifier based on SVM and MRF for remote sensing images
Author :
Ali Masjedi;Yasser Maghsoudi;Mohammad Javad Valadan Zoej
Author_Institution :
K. N. Toosi University of Technology, Geomatics Engineering Faculty, Photogrammetry and Remote, Sensing Department
fDate :
7/1/2015 12:00:00 AM
Abstract :
This paper proposes a novel method for classification of Remote Sensing images. In this method, the popular Maximum Likelihood Classifier (MLC) combined with the Support Vector Machine (SVM) classifier. This method computes the energy function of Markov Random Field (MRF) in the neighborhoods of the test pixels. Then, relates the Markovian energy-difference function to the SVM classifier. Therefore, the salt-and-pepper effect on the classified map is reduced using the proposed contextual classifier. In this paper, two datasets include a hyperspectral and a multispectral image are used. In order to evaluate the proposed method, classification results of this method are compared with MLC and SVM. Experimental results demonstrate that the proposed classification system significantly outperforms other approaches for both hyperspectral and multispectral images.
Keywords :
"Support vector machines","Accuracy","Spatial resolution","Training","Markov random fields","Hyperspectral sensors"
Conference_Titel :
Geoscience and Remote Sensing Symposium (IGARSS), 2015 IEEE International
Electronic_ISBN :
2153-7003
DOI :
10.1109/IGARSS.2015.7326794