Time in hyperspectral processing: A temporal based classification approach

This paper deals with the problem of classifying processes using the temporal information in the sequence of hyperspectral images that are obtained as they take place. That is, taking into account the temporal evolution of the process in the discrimination that must be made. To this end we have considered a particular type of artificial neural networks with trainable delays in their synapses. The classification scheme is studied and applied to the case of resin curing processes. Several test cases involving different proportions of resin components as well as certain environmental conditions such as humidity were created and the system was tested over them producing very promising results.