Research On Computer Vision-based Fillet Color Sorting Of Atlantic Salmon (salmo Salar) And Measuring Of Their Feeding Activity

In recirculating aquaculture systems (RAS), the optimization of fish processing and feeding managing is the key to suppress the production cost. In our study, we discussed the computer Vision-Based sorting of Atlantic salmon (Salmo salar) fillets according to their color level, and the measurement of feeding activity of Atlantic salmon (Salmo salar) by computer vision-based method. The following two parts describe our work in detail.(1) Computer Vision-Based Sorting of Atlantic Salmon (Salmo salar) Fillets According to Their Color LevelComputer vision method was used to evaluate the color of Atlantic salmon (Salmo salar) fillets. Computer vision-based sorting of fillets according to their color was studied on2separate groups of salmon fillets. The images of fillets were captured using a digital camera of high resolution. Images of salmon fillets were then segmented in the regions of interest and analyzed in red, green, and blue (RGB) and CIE Lightness, redness, and yellowness (Lab) color spaces, and classified according to the Roche color card industrial standard. Comparisons of fillet color between visual evaluations were made by a panel of human inspectors, according to the Roche Salmo Fan TM lineal standard, and the color scores generated from computer vision algorithm showed that there were no significant differences between the methods. Overall, computer vision can be used as a powerful tool to sort fillets by color in a fast and nondestructive manner. The low cost of implementing computer vision solutions creates the potential to replace manual labor in fish processing plants with automation.(2) Measurement of feeding activity of Atlantic salmon (Salmo salar) in RAS by computer vision-based methodsIn recirculating aquaculture systems (RAS), the over-feeding events impact the welfare of Atlantic salmon and the operation of systems. This study built a computer vision-based method for measuring the feeding activity of the Atlantic salmon (Salmo salar) school. The feeding procedures of Atlantic salmon were recorded by a CCD camera that mounted above the rearing tank. Feeding activity analysis was based on the difference frame due to the fish motion. An overlap coefficient was defined to calibrate the inaccuracy of the calculation caused by overlapping of fish bodies in images. And an approach was proposed to filter out the influence of the light-reflection on image analysis. Based on these data, the computer vision-based feeding activity index (CVFAI) was determined through measuring the total intensity change of the difference frames caused by the fish motion. To assess the reliability of CVFAI, a manual observation feeding activity index (MOFAI) was determined by scoring each kind of the recorded feeding behaviors. The CVFAI and MOFAI presented a linear relationship at a correlation coefficient of0.915. Thus CVFAI is a potential indicator to measure the feeding activity of Atlantic Salmon school, with a low-cost and rapid way.