A comparison of methods for quantitative analysis of sonar imagery and bathymetry.

Abstract

The usefulness of synoptic seafloor surveys has been limited in the past by the minimal extent to which the user can interact with the resulting survey data with any confidence. An interactive tool for co-registration and quantitative analysis of seafloor acoustic imagery and bathymetry is presented. The technique of stochastic back projection [W. K. Stewart, Proc. ASME Conf. on Off Shore Mechanics, Houston, TX (1988)] is used to produce mosaics of imagery and bathymetry. This application of a uniform mosaicking procedure allows the creation of data ‘‘layer cakes,’’ containing uniformly gridded, co-registered layers of all data available for a given location. This ‘‘layer cake’’ can then be accessed through a mouse-driven screen-interactive tool for selection and analysis of whatever information might be available in the grid. From gridded imagery one acquires a suite of first-order statistics (mean, variance, skewness, kurtosis, etc.), image histogram, image texture, and image fractal dimension. Raw side-scan data yield probability distribution functions (PDFs) indexed by slant range, incidence angle, slope, dip direction, and look direction. From bathymetry, slope statistics, variograms, and surface fractal dimension estimates are extracted. Given this partial or complete descriptive vector, classification maps are created of bottom type, capable of distinguishing to a first order lithologic differences seen in bottom photographs.