Ripeness Level Classification of Banana Fruit Based on Hue Saturate Value (HSV) Color Space Using K-Nearest Neighbor Algorithm

Abstract

Many types of bananas are cultivated locally in Indonesia, including the Muli Banana or Musa Acuminata Linn. During the post-harvest period of banana fruit, there is a problem in the sorting process of bananas based on their level of maturity. The fruit sorting process manually uses the human eye, but it is ineffective due to decreased vision and the large quantity of fruit. Therefore, we need a system that can quickly classify the ripeness of the banana fruit. This study aims to create a system that can organize the maturity level of the banana fruit. The classification system designed using the HSV color feature extraction method and the K-Nearest Neighbor classification algorithm. After going through the testing phase, the system can classify bananas into three classes: unripe, ripe, and rotten. System testing used 30 test data images, and the results show 2 test images whose classification results are wrong and 28 other test images whose classification results are correct. Based on calculations, the accuracy achieved by the system is 93.333%.

Classification of six banana ripeness levels based on statistical features on machine learning approach

“…In the next step, we evaluate by calculating accuracy, precision, and recall for the Naive Bayes classifier and SVM using ( 10), ( 11), ( 12), (13), and ( 14). The overall evaluation results are shown in Table 6.…”

“…In the next stage, we evaluate the model by calculating the value of precision, recall, and accuracy (Acc) in each category using (10), ( 11), ( 12), (13), and ( 14) [36]- [38].…”

“…Based on the research results obtained, an accuracy rate of 73.33%. Research from Kamelia et al [13] uses the HSV and k-nearest neighbor color spaces to classify banana ripeness. The data used are 60 banana images divided into 30 training and 30 test data.…”

“…Furthermore, previous studies such as [13]- [15] classified the maturity level of bananas using 3 or 4 classes. The number of classes used in this study is still relatively small, so in this study, we also propose to use more banana ripeness levels than previous research, i.e., 6 classes.…”