Classification of Siam Orange Ripeness Level using K-Nearest Neighbors Algorithm and Features Gray Level Run Length Matrix

Currently, in evaluating the level of ripeness of Siamese oranges in the Indonesian industrial market, we still use manual methods by relying on human senses and word-of-mouth information. The results of these evaluations have the potential to be inconsistent and inaccurate. This research proposes a classification process that can determine the level of ripeness of Siamese orange fruit to overcome this problem. A ripeness level classification system for Siamese oranges based on skin characteristics was developed using HSV (Hue, Saturation, Value) colour features and Gray Level Run Length Matrix (GLRLM) texture features classified using the K-NN (K-Nearest Neighbors) method. The evaluation results used 300 images of Siamese oranges divided into three classes. Siamese orange fruit ripeness levels show the highest accuracy results of 69.41% obtained at texture degrees of 45, 90 and 135 from a value of K=2. From these data, it is concluded that the Siamese orange fruit ripeness level classification can identify the ripeness level of Siamese orange fruit well.