Discrimination between raised and wild European sea bass through image analysis.

This study aimed at verifying if computer image analysis could represent a good technique to discriminate between reared and wild European sea bass (Dicentrarchus labrax). Images were acquired from sea bass of known origin (n=47 reared and n=13 wild). Images were taken using a digital camera and standardized procedures and were analysed using ImageJ, an image-processing program. Each fish was described by 7 morphometric, 4 shape and 3 colour descriptors that were automatically measured by the soft- ware. The data resulted normally distributed and were submitted to one way-ANOVA that considered the production system (PS) as fixed effect. Linear discriminant analysis (LDA) was used as clas- sification method to identify sea bass PS. Any morphometric parameters (i.e., area, length) were different between PS, while solidity (0.93 vs 0.91, P<0.01), mean gray value (94.7 vs 76.0, P<0.001) and median gray value (89.9 vs 72.6, P<0.001) were higher in reared if compared to wild sea bass. Solidity, the ratio between area and convex area of an object, could be possibly relat- ed to fat distribution, muscles development and growth condition. Gray values have a range between 0 (black) and 255 (white), meaning that caught sea bass was darker. Colour differences could be explained by dietary and environmental factors such as water temperature and chemical parameters. Analyzing all the 14 fea- tures using LDA method led to a 3.3% error of classification after cross-validation. Solidity, mean and median gray values resulted in a 93.3% right classification. Image analysis could be an effective tool to discriminate between reared and wild sea bass, even though further research is required to confirm its on-line applica- tion.