Classification of Watermelon Ripeness Levels Using HSV Color Space Transformation and K-Nearest Neighbor Method

Abstract

Watermelons had high appeal due to their sweet taste, refreshing nature, and numerous benefits. However, consumers often faced difficulties in selecting suitable fruit because of the subtle differences between fully ripe and half-ripe watermelons. One important indicator of a watermelon’s ripeness was the yellowish pattern on its skin. In this study, the proposed use of digital image processing methods, specifically the HSV Color Space Transformation, was aimed at extracting watermelon images and employing the K-Nearest Neighbor (K-NN) method to classify them into two categories: "Ripe" and "Half-Ripe." HSV (Hue Saturation Value) was a color extraction method used to convert colors from the RGB model. The Hue component indicated the type of color, Saturation measured the purity of the color, and Value measured the brightness of the color on a scale from 0 to 100%. In this research, the K-Nearest Neighbor (K-NN) method was applied to classify watermelon images based on the extraction of skin color features. This method compared a new image (test data) with training images to determine classification based on the nearest distance with a parameter of k=3. The data used consisted of 120 images, with 92 images used as training data and 28 images as test data. Experimental results showed an accuracy of 89%, with 25 images correctly classified and 3 images misclassified.