Automatic Airport Detection with Line Segment Detector and Histogram of Oriented Gradients from Satellite Images

Airports are extremely critical targets in both economic and military areas. The earlier detection of these regions provides a very important intelligence information for making that regions unusable against a possible war. For this reason, a new approach has been proposed to automatically detect airports from satellite images. This approach consists of two stages. Firstly, straight line segments have been determined by Line Segment Detector (LSD) based approach and at the end of this, potential airport regions were identified. Secondly, Co-occurrence Histograms of Oriented Gradients (CoHOG) and Hess-CoHOG features were extracted from candidate regions. Extreme Learning Machine (ELM) was used to test the work. In order to evaluate the performance of the proposed method, extensive experiments were applied to satellite images located in different regions of the world. Accuracy, sensitivity and specificity criteria were used in the classification performance. The proposed method was compared with previous works and proved superior with the accuracy of 91%.