Automatic fish age estimation from otolith images using statistical learning

Abstract

In this paper, we investigate the use of statistical learning techniques for fish age estimation from otolith images. The core of this study lies in the definition of relevant imagerelated features. We rely on the characterization of a 1D signal summing up the image content within a predefined area of interest. Fish age estimation is then viewed as a multiclass classification issue using neural networks and SVMs. A procedure based on demodulation and remodulation of fish growth patterns is used to improve the generalization properties of the trained classifiers. We also investigate the combination of additional biological and shape features to the image-related ones. The performances are evaluated for a database of several hundred of plaice otoliths.

The age at death assessment in a multi-ethnic sample of pelvic bones using nature-inspired data mining methods

“…However, most of the applications of computational intelligence methods for data processing in biology concern the classic Feed-Forward Networks (FFN) trained by the Back-Propagation (BP) algorithm [61][62][63][64][65] or self-organising maps (SOM) [59,66].…”

Digital imaging techniques in otolith data capture, analysis and interpretation

“…Machine learning is a paradigm that seems to deliver the best results in terms of performance for automatic ageing, and neural network and statistical frameworks are popular implemen tations. Approaches that use machine learning paradigms often derive features from spatial and frequency do main analysis of 1-D transect signals, sometimes combined with 2-D features extracted from the image (Robertson & Morison 1999, Fablet et al 2004, Fablet & Le Josse 2005, Fablet 2006a) and other measurements such as weight (Fablet 2006b, Bermejo 2007.…”