Estimation of density of trout in fish farm using image-processing technique

Fish is one of the major sources of animal protein that has an important role as a healthy food for people. To supply this useful food source, fish farming in ponds has been considered and developed. Since about 57% of the total cost of fish farms is spent on fish feeds, fish farmers would like to determine the mass of live fishes accurately within period of the rearing. So far, the weight and mass density of live fishes is estimated by statistical methods with an error of about 15 to 25 percent. Therefore, a new method for estimation of the live mass density of fishes was proposed using image-processing technique. A metal frame for installation of a digital still camera was designed and installed on the pool, and at six stages of growth season, several images were taken for different fish densities. An accurate scale was used to measure the actual weight of fishes. Images were then analyzed using MATLAB software. Results showed that at each growth stage, the area of live fishes (in pixel) and actual weight of fishes followed a linear relationship with coefficients of determination (R2) of 0.937, 0.962, 0.790, 0.988, 0.991 and 0.942 from the beginning of growth to final stage of fish harvest. Model was validated with new set of data, and results revealed no significant difference between actual and estimated densities at 5% level of significance.