Building detection using HOG descriptors

One of the most important applications of remote sensing is detecting and analyzing man-made structures in aerial and satellite images. Buildings deserve specific consideration among man-made structures. Manual detection is the one way of labeling the buildings. Recent satellite image resolutions allow researchers manual detection. However, this approach is cumbersome and prone to errors. Therefore, an automated method is needed. Unfortunately, the solution is not straightforward due to various factors. In this study, we propose an adaptation of an earlier successful method, Histogram of Oriented Gradients, to building detection. We test our method on several Ikonos satellite images and provide the results.