author(s)
El habouz Youssef, Es-saady Youssef, El yassa Mostafa, Mammass Driss, Nouboud Fathallah, Chalifour Alain, El habouz Hammou and Manchih Khalid

abstract
The external form of otolith is one of the main characteristics of fish species recognition; which is a major issue in several marine ecological studies; such as the determination of the food spectrum by the otoliths recovered from the stomach or feces. In this paper, we present an automatic identification system of fish species based on otolith shape analysis. Our proposed system consists of three main phases: pre-processing phase: image de-noising and enhancing grayscale contour. Feature extraction phase: we extract the median distance vector of the contour, which is used in the third phase, and this one is based on a multi-layer perceptron classification method. The efficiency of the new system was proved on two bases: AFORO database and a first national database, which is collected and prepared during the present work. Compared to Elliptic Fourier descriptors, Complex Fourier descriptors and Geodesic-based method, the correct recognition rate obtained was the higher, with 98.33 % for the first dataset and 95.6% for the second.