MULTISENSOR INTEGRATION FOR UNDERWATER SCENE CLASSIFICATION

We describe a new approach for the classification of a seafloor that i s imaged with high frequency sonar and optical sensors. Information fr om these sensors is combined to evaluate the material properties of th e seafloor Estimation of material properties is based on the phenomeno logical relationship between the acoustical image intensity, surface r oughness, and intrinsic object properties in the underwater scene. The sonar image yields backscatter estimates, while the optical stereo im agery yields surface roughness parameters. These two pieces of informa tion are combined by a composite roughness model of high-frequency bot tom backscattering phenomenon. The model is based on the conservation of acoustic energy travelling across a fluid-fluid interface. The mode l provides estimates of material density ratio and sound velocity rati o for the seafloor. These parameters serve as physically meaningful fe atures for classification of the seafloor. Experimental results using real data illustrate the usefulness of this approach for autonomous an d/or remotely operated undersea activity.