A NEW PROMPT FOR BUILDING EXTRACTION IN HIGH RESOLUTION REMOTELY SENSED IMAGERY
Keywords: Building extraction, SRM segmentation, Shape-factors, Clustering, Remote sensing
Abstract. Remote sensing technology has proved a magnificent role in urban development. High resolution remotely sensed imagery provides remote sensing mapping, GIS data acquisition and automatic updates, and also supports for extraction train features such as building. In this work, a new approach for building extraction in colored high resolution remotely sensed imagery is proposed. The approach includes Statistical Region Merging (SRM) segmentation, boundary tracing and clustering based on new shape-factors. After performing SRM, the boundary tracing algorithm is carried out. Moore neighbour contour tracing algorithm is used for this purpose. New shape factors are proposed based on geometry of buildings and other terrain features. The correlation of shape factors between buildings and road features is significantly low. The shape factors are based on circularity, elongation and compactness. Afterwards K-means classifier is utilized in order to discriminate between buildings and other objects. In this step squared Euclidean distance has been opted. Two clusters have been set for separating buildings from roads and complex features. In order to evaluate the capability of this method, two images of Worldview2 sensor has been used. The fine result demonstrates the proficiency of shape factors and remarkable and satisfactory performance of the new prompt.