Reflectance and UV Induced Fluorescence Imaging for Red Betel Chlorophyll and Anthocyanin Prediction
Retno Damayanti*, Yusuf Hendrawan, Sandra, Bambang Dwi Argo, Rut Juniar Nainggolan, Mitha Sa^diyah

Department of Biosystems Engineering, Faculty of Agricultural Technology, Universitas Brawijaya, Malang 65145, Indonesia
Email: *damayanti[at]ub.ac.id

Abstract

Non-destructive measurement of phytochemical content is essential for quality control in herbal industries. Piper crocatum Ruiz (red betel) contains high levels of chlorophyll, flavonoids, phenols, and anthocyanins, yet conventional analytical methods are destructive, time-consuming, and labor intensive. This study presents an integrated system combining reflectance imaging, UV-induced fluorescence imaging, and computer vision with multi-output artificial neural networks (ANNs) to simultaneously predict chlorophyll and anthocyanin contents in red betel leaves. Images were acquired under controlled illumination for both modalities, and 100 color texture features from multiple color spaces (RGB, HSV, Lab, grayscale) were extracted, ranked, and selected using multiple feature evaluators. Selected features were used to train and optimize ANN models, with performance assessed by mean squared error (MSE) and correlation coefficient (R). Results showed distinct phytochemical patterns across leaf maturity stages, chlorophyll increased from top to bottom leaves, while anthocyanins decreased. Both reflectance and fluorescence models achieved high prediction accuracy (R validation = 0.95), with fluorescence slightly outperforming reflectance (MSE validation = 0.12783 vs. 0.17227) but requiring more training iterations (29 vs. 5). The multi-output ANN approach improved accuracy, efficiency, and prediction consistency for predicted chlorophyll and anthocyanin content in Red Betel Leaves.

Keywords: Anthocyanin- Chlorophyll- Fluorescence- Red betel- Reflectance

Topic: Agricultural and bioprocess engineering