In this paper, the discriminative capability of a combination of biological and shape features for fish age classification are analyzed. In particular, the usefulness of otolith weight in several species, in combination with other features such as otolith shape features and biological features such fish length, weight and sex is evaluated. The classification performance for different state-of-the-art statistical learning classifiers (i.e. several non-linear, non-parametric classifiers such different types of multi-class support vector machines) using an Atlantic cod database has been tested in which otolith weight has shown to be a powerful characteristic for classification purposes but the greatest accuracy is achieved when it is used simultaneously with other features.

Peer Reviewed