Y.K. Firdaus, J. Prasetyo, Y. Hendrawan, D.F. Al Riza
Reflectance and fluorescence imaging were employed to assess maturity in California lemons (Citrus limon cv. California) based on skin color and texture, alongside internal quality indicators. Fluorescence imaging outperformed reflectance alone in predicting maturity levels, likely due to enhanced capture of biochemical and textural variations in lemon surface. Machine learning analysis using both support vector machines (SVM) and k-nearest neighbors (k-NN) revealed that SVM models trained on fluorescence images provided the most accurate classification. Specifically, fluorescence imaging data processed with SVM (without scaling) achieved 100% accuracy in training and 92% in testing, surpassing other model and feature configurations. These findings showed the utility of fluorescence imaging for potential, nondestructive lemon maturity classification, offering a promising framework for broader applications in citrus and other horticultural commodities. © 2025 Association of Agricultural Technology in Southeast Asia. All rights reserved.
Department of Biosystem Engineering, Faculty of Agricultural Technology, Universitas Brawijaya, Malang, Indonesia
