Abstract
Prediction relationships between weight and image features were established with a high correlation for whole aquacultured rainbow trout. Three hundred fish from three different farms were used. The fish were temporarily removed from the raceway, anesthetized, and their picture was taken by a digital camera. A reference square of known surface area and color was placed beside the fish. The fish were then returned to the raceway alive. The image was analyzed, and the view area of the fish was calculated using the area of the reference square. The average color of the fish was also determined (L*, a*, and b* values). The following equations were used to fit the view area (X) vs. weight (Y) data: linear, power, and second order polynomial. The R2 values for the used equations were: linear = 0.98; power = 0.99; polynomial = 0.98. Image analysis can be used reliably to predict the weight of whole aquacultured rainbow trout. In addition, color and other visual attributes can be objectively determined by image analysis to sort by visual quality.
Keywords:
The authors would like to thank the manager Mr. Tamer Demirkan, HAT Fisheries Co. Ltd.; the facilities engineer Mr. Ersin Çelik, Selina Fisheries Co. Ltd.; and the facilities engineer Mr. Serhat Topçu, Basar Trout Farms, for allowing the use of their plant and their fish for this study. The authors also thank Dr. Erkan Gümüs for his help in acquiring the pictures. Partial funding was provided by the University of Alaska Fairbanks and Akdeniz University.