A COMPUTER VISION APPROACH FOR CLASSIFYING CALIFORNIA PAPAYA RIPENESS USING K-NEAREST NEIGHBOR

Abstract

Determining the ripeness level of California papaya is important for harvest decisions, sorting, distribution, and spoilage control. In practice, ripeness identification is still commonly performed visually and is therefore subjective. This study aims to develop a digital image-based classification system for California papaya ripeness using the K-Nearest Neighbor (K-NN) algorithm with Hue and Saturation features in the Hue Saturation Value (HSV) color space. The dataset consists of 90 primary images, divided into 60 training images and 30 testing images, with four ripeness classes: unripe, half-ripe, ripe, and rotten. All images were captured using a Xiaomi Mi A2 Lite smartphone and cropped to 1436 × 1000 pixels. Classification was conducted using Euclidean distance. The value of k was selected empirically through trial and error in the original study, and k = 9 was retained because it produced the most stable result on the available data while reducing the potential for class ties. The evaluation produced 22 correct predictions out of 30 test images, resulting in an accuracy of 73.33%. This revised manuscript strengthens the methodological reporting by clarifying parameter selection, documenting the data distribution and providing a literature-based comparison with alternative methods, such as Support Vector Machine (SVM) and Convolutional Neural Network (CNN). The findings suggest that K-NN with HSV features remains a feasible, low-cost baseline, although its performance should be improved through larger datasets, per-class evaluation reporting, and head-to-head comparisons on the same dataset.