Automatic building detection from satellite images using spectral graph theory

Abstract

Approach of detection of buildings on satellite images is presented in this paper. Actuality of work is shown. Possible areas of the use of the research results are presented. The detection is based on joint use of height, color, shape. The height of the object is determined based on energy minimization within the region. The image is segmented previously. The area of vegetation and the shadows are detected based on color invariants. Regions of the image are grouped based on the properties using spectral clustering. The shape and size of objects are analyzed additionally. The results of the algorithm are given.