Weight grading of freshwater fish based on computer vision

In order to facilitate the subsequent processing, it is necessary to develop a grading system for weight classification of freshwater fish automatically. In this study, 86 freshwater fish were collected as the test samples. With taking the image of each fish by the machine vision, and through the image processes：gray, binary conversion and contour extraction, the axis and the projected area of the crucian were extracted. By experiment, the proportional relations of the length with the weight of the head, the belly and the tail were found out, which were used to correct the projected area. Finally the prediction model was extracted by the regression analysis. The experiments showed that the weight of the fish was highly correlated with the projected area, the R2 was 0.9878; and the forecast model was verified. The mean relative error was 3.89% and the mean absolute error was 6.81 g. Results show that the computer vision can be used to grade the freshwater fish.