A minimum - cost thresholding technique for unsupervised change detection

This article maybe used for research, teaching and private study purposes. Any substantial or systematic reproduction, redistribution , reselling , loan or sub-licensing, systematic supply or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material. Abstract. We propose an automatic thresholding technique for diOE erence images in unsupervised change detection. Such a technique takes into account the diOE erent costs that may be associated with commission and omission errors in the selection of the decision threshold. This allows the generation of maps in which the overall change-detection cost is minimized, i.e. the more critical kind of error is reduced according to end-user requirements. 1. Introduction Change-detection maps are a useful support for decision making (risk assessment, natural disaster monitoring, damage evaluation). Most unsupervised change-detection techniques proposed in the remote-sensing literature aim at minimizing the overall change-detection error, without taking into account the practical objective for which maps are produced (Singh 1989, Bruzzone and Serpico 1997). However, in many applications, commission errors (which occur when unchanged pixels are detected as changed ones) and omission errors (which occur when changed pixels are identi ed as unchanged ones) involve diOE erent consequences in terms of decision making. In these cases, the minimization of the overall error is not the most suitable goal for the change-detection process. For example, let us assume that the objective of change detection is the identi cation of damages caused by a natural disaster (e.g. wild re, ood). In this context, the end-user (e.g. civil protection) may prefer a change-detection map where all the possible damaged areas are pointed out (even if some of them are wrong), rather than a map where some damaged areas are missing. This means that omission errors are more costly than commission ones. Consequently, the goal of change detection should be the generation of a map in which omission errors are minimized. This objective can be attained by associating diOE erent weights (i.e. costs) with …