Weight prediction of intensively reared gilthead seabream Sparus aurata from morphometric traits measured in images

Abstract

AbstractThe objective of this study was to establish an accurate body weight (BW) prediction model for gilthead seabream Sparus aurata of 50–1000 g. Three thousand three hundred twelve (3312) fish were individually weighed and photographed. Traits measured from the images were total body length (TBL), fork body length (FBL), standard body length (SBL), body height (BH), head length (HL), eye diameter (ED), body area (BA, without fins), head area (HA), and eye area (EA). SBL, BH, BA, BA/SBL, and BA/BH showed a strong association with BW (correlation coefficients, r: 0.96–0.99). These traits were selected to proceed with the regression analysis. Simple, multiple linear, and 2nd-order polynomial regressions were applied to the whole data set and three BW subgroups of interest during gilthead seabream rearing (i.e., 50–100 g, 100–500 g, 500–1000 g). The prediction of BW from the whole data set was more accurate than from each BW subgroup. The models with the highest coefficient of determination (R2) and the lowest errors (mean absolute percentage error, MAPE) were either the power regression of BW with BA (R2: 99.0%, MAPE: 5.8%) or the multiple linear regression of BW with SBL, BA, BA/SBL, and BA/BH (R2: 98.6%, MAPE: 5.1%) as predictors. The accuracy of the two models is considered quite similar, and for reasons of simplicity, the power regression is advantageous, requiring only one trait to be measured (BA). The models identified in the present study can help to further develop the accuracy of machine vision-based systems for gilthead seabream BW measurement.