Yellowfin Tuna (Thunnus Albacares) Meat Real-Time Mass Grading System Using OpenCV

Ensuring food safety and quality in the fish and seafood industry is crucial for consumer health and economic stability, especially for perishable exports like tuna. The study developed a non-destructive, real-time, and mass grading system for Yellow fin tuna meat using the Philippine National Standard as the basis for grading using an edge detection algorithm, contour detection, and deep learning to process RGB color space for quality grading. One thousand one hundred seventy-three tuna images were used for model evaluation, with 821 images for training and 352 for testing. The model demonstrated high accuracy across all quality classes, ranging from 95.08% for the Good class to 97.62% for the Very Poor class, although the Poor class had the lowest recall rate at 83.51 %. The QDA results from the DOST-FNRI sensory laboratory produced 336 data points for analysis. The model supports individual and mass testing of tuna samples, enabling researchers to conduct 671 comparisons. The model accurately classified 606 instances, with 65 misclassifications, resulting in an accuracy rate, a weighted recall, a weighted F1-score of 90.31 %, and a weighted precision of 90.34 %.