A hierarchical oil depot detector in high-resolution images with false detection control

Oil depot detection in high-resolution images is a challenging task due to the complicated background. This paper aims at further investigating this problem and presents an approach to detect oil depots in a hierarchical manner. Firstly, the Ellipse and Line Segment Detector (ELSD) which guards against false positives is applied to detect elliptical arcs in the image. Afterwards, the Histograms of Oriented Gradient (HOG) are extracted based on the elliptical arc candidates and input into the AdaBoost classifier in order to get the detection of oil tanks. Finally, the Depth-First-Search (DFS) is used to cluster the detection of oil tanks and determine the final oil depot area. Experimental results on real database indicate that the hierarchical algorithm is robust under complicated background and shows good performance against false positives.