A contribution to the problem of mapping seabed transition zones.

The paper proposes a framework for discriminating between distinct sea-bottom regions on the basis of the shape of the objects present on them. Supervised (Bayesian) and unsupervised algorithms are presented that allow the definition of shape classifiers. These classifiers define partitions of a non-Euclidean shape space which is an extension that imposes robustness with respect to circular permutations and mirroring, additionally to the translational, scale and rotational invariances already contemplated by the original Kendall definition. The algorithms proposed are applied to real data corresponding to the shapes induced on side-scan images by two kinds of underwater bottoms: sand ripples and Posidonia fields. Application of the proposed theory to the problem of on-line bottom discrimination is illustrated on fields simulated using statistical shape models identified from real data