Separation of fish stocks by otoliths: Image representation, Fourier approximation and discrimination

Abstract

In this thesis the entire process of Fourier Contour Analysis have been investigated for potential sources of bias. No potential biases has been found when creating the contour, however it has been shown there is a slight bias in which otoliths get invalidated by breakage. A probablity based sampling strategy has been suggested instead of the previously used stratification based. It has been shown that both numerically solving the case of best alignment fit for the otolith contours and that using the absolute value of the fourier coefficients improve the final discriminant results