Mackerel Tuna Freshness Identification Based on Eye Color Using K-Nearest Neighbor Enhanced by Contrast Stretching and Histogram Equalization

Purpose: The present study focuses on the development of a robust fish freshness classification system based on the application of different digital image processing techniques from mackerel tuna eye images toward better classification. Methods: Contrast stretching and histogram equalization were done to improve image quality before the classification. The system contained 250 training images in a dataset, for five freshness classes which are 3, 6, 9, 12, and 15 hours post-catch, with 50 test images. For classification, the K-Nearest Neighbor (KNN) algorithm was employed with a parameter setting of K = 5, ensuring effective differentiation between the various freshness levels based on the enhanced image features. Result: The results depicted very low MSE values after enhancement at 6 hours, as low as MSE = 0.0012606 and PSNR = 28.9944 dB for contrast stretching, and for 12 hours, histogram equalization gave the best results, MSE = 0.0030712 and PSNR = 25.127 dB. Further, classification done through the KNN classifier with K=5 gave results with accuracy as high as 100% was achieved on the testing data, proving that the model was successfully able to identify the classes of freshness. Novelty: The novelty in the present research work is the integration of advanced image-processing techniques, which allow the achievement of an improved level of detection of fish freshness and a very useful solution to the seafood industry in view of product quality and safety assurance. Generally, the paper epitomizes an important milestone in the application of machine learning and image processing for the assessment of the quality of foods.