Extracting Dwellings in Refugee Camps Using Multifractal Analysis and Mathematical Morphology Based Descriptors

This article presents an automatic procedure for detecting and counting dwellings in refugee/internally displaced persons camps. Very high resolution (VHR) satellite images are used to monitor camps, especially in inaccessible to “in-situ” measures areas. We develop a new algorithm to analyze these images, with the aim of improving detection accuracy and computing performance. The algorithm is based on local multifractal analysis and mathematical morphology, two methods that are becoming increasingly popular in the image analysis community. Our procedure translates the visual characterization of the desired structures into a morphological image processing chain. However, morphological filtering is not performed on the original image <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">per se</i>, but on the image expressed by the Hölder exponent. Proposed method is applied to a set of VHR satellite images (GeoEye-1, WorldView-2, -3, -4 and JL-1GF02A) of two camps in Africa. Our technique is compared with results obtained by visual interpretation. The correlation coefficient between the two methods is 0.98, with an omission error of 7.98% and a commission error of 4.54%.