A novel approach to targeted change detection

In several applications the objective of change detection is actually limited to identify one (or few) specific “targeted” land-cover transition(s) affecting a certain area in a given time period. In such cases, ground-truth information is generally available for the only land-cover classes of interest at the two dates, which limits (or hinders) the possibility of successfully employing standard supervised approaches. Moreover, even unsupervised approaches cannot be effectively used, as they allow detecting all the areas experiencing any type of change, but not discriminating where specific transitions of interest occur. In this paper, we present a novel technique capable of addressing this challenging issue by using the only ground truth available for the targeted land-cover classes at the two dates. In particular, it jointly exploits the expectation-maximization algorithm and an iterative labeling strategy based on Markov random fields accounting for spatio-temporal correlation. Experimental results confirmed the effectiveness and the reliability of the proposed method.