Detection of Urban Features From High Resolution Satellite Images

One common approach to detect urban features from high resolution images is, using automatic classification methods. The purpose of this paper is to demonstrate the applicability of the developed algorithm (DA) by systematically evaluating its performances in comparison to other popular classifier, support vector machine (SVM). The detection performance of algorithms is evaluated by an object-based criterion. Considering consistency, the same set of ground truth data which is produced by labeling the building boundaries in the GIS environment is used for accuracy assessment. The method is applied to two different Quickbird images for complex urban patterns. In evaluation of object based accuracy assessment it is shown that, while, SVM provide higher rates of correct detection it provides higher rates of false alarms. DA, on the other hand, providing tolerable rates of correct detection and lower rates of false alarm.