Spectral coherence applied to vessel tracking

Multi-Chromatic or spectral Analysis (MCA) of SAR images consists in splitting wide band SAR signals into sub-bands to generate several lower resolution images from a single acquisition. This splitting allows performing a spectral analysis of observed scatterers. Spectral coherence is derived by computing the coherence between sub-images issued from a single SAR acquisition. It was shown that in the presence of a random distribution of surface scatterers, spectral coherence is proportional to sub-band intersection of sub-images. This model is fully verified when observing spectral coherence on open seas areas. If the scatterers distribution departs from this distribution, like for manmade structures, spectral coherence may be preserved to a certain degree. We investigated the spectral coherence to perform vessel detection on a sea background by using spotlight images acquired on the Venice Lagoon. Sea background tends to lead to very low spectral coherence while this latter is preserved on the targeted vessels, even for very small ones. A first analysis shows that all vessels observable in intensity images are easily detected in the spectral coherence images which can be used as a complementary information channel to constrain vessel detection.