Automatic damage detection Using pulse-coupled neural networks For the 2009 Italian earthquake

In this paper, we investigate the performance of pulse-coupled neural networks (PCNNs) to detect the damage caused by an earthquake. PCNN is an unsupervised model in the sense that it does not need to be trained, which makes it an operational tool during crisis events when it is crucial to produce damage maps as soon as the post-event images are available. The damage map resulting from PCNN was validated at a block scale of 120×120m using ground truth obtained by a combination of ground survey and visual inspection of the before- and after-event images. The comparison showed agreement between the change measured by PCNN on block scale and the damage occurred.