Automated Quality Classification of Local Sunkist Oranges Based on Digital Image Processing Using the K-Nearest Neighbour Method
Cindy Hartita (a*), Luvyta L. M. Syawalia (b), Yulia Resti (c), Ismail Thamrin (b), and Irsyadi Yani (b)

a) Doctoral Program of Engineering Science, Faculty of Engineering, Universitas Sriwijaya, Indonesia
*cindyhartita1[at]gmail.com
b) Department of Mechanical Engineering, Faculty of Engineering, Universitas Sriwijaya, Indonesia
*irsyadiyani[at]ft.unsri.ac.id
c) Department of Mathematics, Faculty of Mathematics and Natural Science, Universitas Sriwijaya, Indonesia

Abstract

Indonesia, as one of the worlds largest tropical agricultural producers, faces considerable challenges in maintaining the quality and consistency of its horticultural commodities, particularly citrus fruits, amid rising domestic consumption and the growing complexity of modern distribution networks. Increased demand for local Sunkist in Indonesia, especially during certain periods, has led to a growing need for accurate and efficient fruit sorting processes. Manual sorting processes have the potential to cause classification errors, which can result in distribution losses, reduced fruit quality, and health risks for consumers. This study aims to develop an automatic local Sunkist quality identification system based on digital image processing using the K-Nearest Neighbour (K-NN) method. The dataset used consists of 90 images of local Sunkist grouped into three categories, namely unripe, ripe, and rotten. Image acquisition was performed using a webcam installed in a lighting chamber to maintain lighting stability and minimize background interference. The image processing involved cropping the centre area of the image to a size of 20x20 pixels, extracting the Red, Green, and Blue (RGB) colour intensity values, and normalizing the data using the Min-Max Scaling method. The classification process was carried out using the K-NN method with a K value of 3 and Euclidean distance calculation. The results showed that the system was able to classify the ripeness level with an accuracy rate of 92.6%.

Keywords: K-NN- RGB extraction- Image processing- Sorting system- Local sunkist

Topic: Applied Technology in Physics