DBPapers
DOI: 10.5593/SGEM2014/B23/S10.009

BUILDING EXTRACTION FROM HIGH RESOLUTION SATELLITE IMAGES USING MATLAB SOFTWARE

S. Dahiya, P. K. Garg, M. K. Jat
Wednesday 1 October 2014 by Libadmin2014

References: 14th International Multidisciplinary Scientific GeoConference SGEM 2014, www.sgem.org, SGEM2014 Conference Proceedings, ISBN 978-619-7105-12-4 / ISSN 1314-2704, June 19-25, 2014, Book 2, Vol. 3, 71-78 pp

ABSTRACT
Now-a-days, automatic building extraction from high resolution satellite images is a very active field of research. During the past decade, a considerable research has been done for automatic extraction of buildings from satellite images. In India, building extraction for satellite images is a difficult task because the building doesn’t follow a specific pattern and the individual building covers a very small area on the ground. In addition, the reflectance of buildings and roads is almost similar on satellite images for many regions which results in error in traditional digital classification. In such cases, differentiation between buildings and roads by traditional digital classification of satellite image becomes a difficult task. In this paper, an algorithm has been proposed to extract the urban buildings from high-resolution multispectral IKONOS image. In the proposed method, an IKONOS image of Dehradun City, Uttarakhand, India has been used. For the accurate extraction of buildings, a program is developed in Matlab. In the first step, an image enhancement method is applied on the multispectral IKONOS image. Then NDVI is applied individually on all bands of enhanced image to remove the vegetation cover. After removing the vegetation, all the bands are stacked, and the image is converted from RGB color space to L*a*b* color space. The colors in ’a*b*’ space are classified using KMeans Clustering. Using the results from K-Means, all the pixels in the image are labeled, and thereafter the segmentation is done. In next step, morphological operators – erode and dilate are applied. The small holes of other objects present in the buildings segments are filled up. Later, the segments of our interest are filtered out on the basis of area and shape. At the end, some cleanup operators, like smoothness and orthogonality are applied on the image. The buildings present in the IKONOS image are also manually digitized using ArcMap 10. The buildings extracted from the proposed method are compared with the manually digitized buildings to carry out the accuracy assessment.

Keywords: Building extraction, Segmentation, Clustering, Morphological operators, Matlab