Hyperspectral change classification based on direction pursuit detection

Godefroy Brisebarre, M. Guillaume, L. Denise

Abstract: Change detection is an important area of interest within the hyperspectral community. Some recent works already showed the performances of direction pursuit in this area. In addition to giving a good change detection those methods allows a distinction between pixels detected on different directions. Unfortunately, when applied to real interesting data, those methods lead to an irrelevantly high number of different detection directions. We present in this paper a direction merging approach based on an hypothesis test, in order to obtain significant classes. We experimentally show the efficiency of such an approach to classify the detected change pixels.

External IDs:dblp:conf/whispers/BrisebarreGD13