Parameterization of 3D AFORO otolith surfaces

Abstract

The 3D otolith shapes recently included in the AFORO database are defined by means of clouds of points of their 3D surfaces. Automatic retrieval and classification of 3D natural objects becomes a difficult task when the number of the elements in the database increases because of the great amount of information involved in the description of each object. In order to simplify that task, in this contribution we propose a new method for compacting data from 3D shapes by extracting, for each shape, a 3D characteristic curve that works as a 3D object signature. By means of that 3D signature the morphological shape information can be preserved but the number of points required for the shape representation is drastically reduced. The method allows us to define a 3D characteristic curve with a variable precision, so if we need 3D objects with a low precision the number of points diminishes also. The important issue is that, by construction, those signatures always form a 3D closed curve and their corresponding x, y, z coordinates can be treated as periodic functions. This property lets apply Elliptical Fourier descriptors on each coordinate in order to achieve a bigger information compression if we select a reduced set of them (the ones that bring more information). So, with a reduced set of parameters, the main information of the original 3D shape can be preserved. The 3D otolith shapes of AFORO are digitalized following an acquisition protocol and therefore it is not necessary that their descriptors have the rotation invariance property. However, from the proposed set of parameters, it is possible to develop a new set of parameters that are invariant to rotations. The presented parameterization was applied to sagitta otolith of Mediterranean sciaenids but can also be applied to describe other biological origin shapes