Automatic building extraction from digital ortho - photos remains an open research area in digital photogrammetry. This paper proposes a sequence of classification and linearization processes for the extraction of building features with different rooftops from high resolution Multispectral satellite images (e.g., IKONOS and Quickbird) in Middle East countries. In this paper the minimum distance from a variant mean classification followed by a linearization approach using Hough transform is used. The variant mean if performed using a part of the image as a training data set for each building. This training set iteratively updated at each process by adding a new color data to the training data set. A linearization process is then performed using the ability of Hough transform to detect straight lines.
The software has been designed and developed in Matlab programming environment and designed to be a user friendly since little interaction is required from the users is required. The proposed system works in four different phases: the first phase is the user interaction phase, the second phase is the colors classification phase, the third phase is Hough transform phase, and the fourth phase is the coordinate classification and polygons generalization phase. An image from IKONOS sensor is used for testing this system. The extraction results are compared with manually digitized ones. The comparison illustrates the efficiency of the proposed algorithm in which it can extract approximately 80% of buildings in the image properly.
|
