This paper, proposed a classification approach that utilizes the high recognition ability of Hidden Markov Models (HMM s) to perform high accuracy of classification by exploiting the spatial inter pixels dependencies ( . the context ) as well as the spectral information. Applying unsupervised classification to remote sensing images can provide benefits in converting the raw image data into useful information which achieves high classification accuracy. | International Journal of Computer Networks and Communications Security C VOL. 1, NO. 4, SEPTEMBER 2013, 165–172 Available online at: ISSN 2308-9830 N C S Land Cover Classification Using Hidden Markov Models Dr. GHAYDA A. AL-TALIB1 and EKHLAS Z. AHMED2 1 Assist. Prof. Computer Science Department, College of Computer Science and Mathematics, Mosul University, Mosul, Iraq 2 M. Sc. Student, Computer Science Department, College of Computer Science and Mathematics, Mosul University, Mosul, Iraq E-mail: 1ghaydatalib@, 2ekh_aa2007@ ABSTRACT This paper, proposed a classification approach that utilizes the high recognition ability of Hidden Markov Models (HMM s) to perform high accuracy of classification by exploiting the spatial inter pixels dependencies ( . the context ) as well as the spectral information. Applying unsupervised classification to remote sensing images can provide benefits in converting the raw image data into useful information which achieves high classification accuracy. It is known that other clustering schemes as traditional k-means does not take into account the spatial inter-pixels dependencies. Experiments work has been conducted on a set of 10 multispectral satellite images. Proposed algorithm is verified to simulate images and applied to a selected satellite image processing in the MATLAB environment. Keywords: Hidden Markov Models (HMM), Land cover, Multispectral Satellite Images, Clustering, Unsupervised classification. 1 INTRODUCTION In this paper the Hidden Markov Models (HMM s) for unsupervised satellite image classification has been used. An HMMs were extensively and successfully used for texture modeling and segmentation (., classification), this is majorly due to their ability to model contextual dependencies and noise absorption [1]. The land cover is an important geospatial variable for studying human and physical environments and is increasingly used as input data in spatially .
