Skip to Main Content
Seabed type has a major influence on mine countermeasures (MCM) operations. In this paper an accurate and efficient seabed characterization method is presented. The basic classification is performed by wavelet-based decision trees, while a data fusion algorithm based on Dempster-Shafer Theory of Evidence is used to combine the information provided by multiple observations from the same or different sensor platforms. Markov Random Field filtering of the resulting belief maps is finally used to provide cleanly delimited seabed segmentations. The technique is demonstrated using data obtained by NURC during recent sea experiments with NATO partner nations.