1 - Segmentation bathymétrique d'images multispectrales SPOT

This paper addresses the analysis of multispectral SPOT images in order to update nautical charts and to control nautical data. We have developed a segmentation approach based on two Markovian modeling steps. The first one is based on Markov chain (1D) modeling, whereas the second step involves a hierarchical process, Markovian in scale. Each of them includes the unsupervised estimation of the parameters. The model parameters are automatically calibrated whereas the noise parameters are estimated in the context of generalized distribution mixtures. An adaptive bathymetric inversion model is then derived in order to recover the water depth near the coasts. This bathymetric estimation has been validated on real data, for which control points are available that correspond to bathymetric measures supplied by previous hydrographic campaigns.