Unsupervised change detection between multi-sensor high resolution satellite images

Gang Liu, Julie Delon, Yann Gousseau, Florence Tupin

Abstract: In this paper, we present a novel unsupervised framework for change detection between two high resolution remote sensing images. Thanks to the use of local descriptors, the method does not need any image co-registration and is able to identify changes even with images acquired from different incidence angles and by different sensors. Local descriptors are used to both locally align images and identify changes. The setting of thresholds as well as the final grouping of isolated changes are performed thanks to a contrario statistical procedures. This provides a complete and automatic pipeline, whose efficiency is shown through several challenging pairs of high resolution urban images, acquired through different satellites.