Tomato Ripeness Identification Using Principal Component Analysis and K-Nearest Neighbor Based on Color Image

Abstract

Manually determining the level of ripeness of tomatoes has weaknesses because the standards are subjective and time consuming. This research aims to identify ripeness of tomatoes based on Hue Saturation Value (HSV) color representation using Principal Component Analysis (PCA) as feature extraction and K-Nearest Neighbor (KNN) for classification. This research uses 400 images with a spatial resolution of 400x400 which are grouped into 5 levels of maturity, namely green, turning, pink, light red and red. The data is divided into training data and test data with a ratio of 80:20. The scenario applied is a division of color space data, namely Hue (H), Saturation (S), Value (V), Hue-Saturation (HS), Hue-Value (HV), Saturation-Value (SV) and HSV. The values ​​of k as a neighbor in KNN used as a scenario are 1, 3, 5, 7, 9 and 11. The principal component values ​​applied are 5, 10, 15 and 65 with a variance ratio of 95%. The research results show that with K=7 and PC value =5 it produces the highest accuracy value with a percentage of 94% in HV testing. The results of this research show that by classifying test data of 80 image data, the results obtained were 75 accurate data and 5 inaccurate data.