Discrimination of Norwegian farmed, ranched and wild-origin Atlantic salmon, Salmo salar L., by image processing

Abstract  A method of distinguishing between farmed, ranched and wild-origin Atlantic salmon, Salmo salar L., using scale morphology is proposed. Circuli spacing and scale texture data, as expressed as a Fourier transform of transmission luminescent patterns. were extracted by image processing. Spacing patterns and texture features were most distinct for wild salmon compared with the other two groups. Three-group quadratic discriminant function models were developed using different combinations of data types. The most efficient model to separate the three groups had a classification efficiency of 74%. When models were simplified to two groups, farmed and wild, efficiency increased to 90%, thus reflecting the feature overlap between farmed and ranched groups. The method may be a useful tool for more objective and efficient classification of wild versus husbandry-origin salmon. However, it should be stressed that farmed salmon that escape at the smolt stage are still problematic.