Classification of starfruit maturity using smartphone-image and multivariate analysis

Abstract

Grading starfruit samples based on the stages of ripeness can be facilitated by machine vision system (MVS) that requires high-quality images. The evolution of modern smartphone cameras can be of assistance in this regard. This study was carried out to examine if smartphone image-based processing can be applied to categorize starfruit samples based on their maturity. In this regard, images of starfruit samples at three different maturity stages were acquired through smartphone camera. The MATLAB platform was used to extract the color features of the images. Each channel of the three-color space model (RGB, HSV, L*a*b*) was extracted. Principal component analysis (PCA) was applied to quantify the existence of variance in three different maturity classes. In addition, classification models were created using Linear Discriminant Analysis (LDA), Linear Support Vector Machines (SVM), Quadratic SVM, Fine K-Nearest Neighbor (KNN), and Subspace Discriminant Analysis (SDA). The best classifier was found to be the Linear Discriminant Analysis (LDA) which had an accuracy of 96.2% for calibration and 93.3% for validation. Accurate classification of the ripeness indices of starfruit samples demonstrated by LDA indicates the potential for commercial application of this technology.

Keywords

• Image processing
• Machine vision
• Computer vision
• Color space model
• Fruit grading
• Linear discriminant analysis