Pomelo maturity classification from field-acquired images using oil-gland morphology and a rule-based image-processing pipeline

Abstract

Pomelo maturity assessment in commercial orchards relies predominantly on vi sual inspection and harvest age records, which introduce inconsistency in post harvest grading. Non-destructive alternatives such as near-infrared spectroscopy and acoustic sensing have been reported, but typically require specialised instruments and controlled acquisition conditions. This study investigates the feasibility of oil-gland morphology as an interpretable maturity indicator, implemented as a rule-based image-processing pipeline executable on standard CPU hardware without model training. A hierarchical rule-based framework was developed to classify pomelo maturity from gland count features extracted under natural outdoor illumination. Thirty-three Citrus maxima samples (Khao Yai cultivar) representing three maturity stages were analysed in this proof-of-concept study (n = 11 per stage). The pipeline integrates adaptive thresholding, subregion segmentation, multi-scale morphological detection, and threshold-based classification. Detection reliability was verified on synthetic dot-pattern images prior to real-sample evaluation. On the collected dataset, the framework achieved an overall accuracy of 78.8% with a macro-averaged F1-score of 0.784. No mis classification occurred between the immature and mature groups; errors arose exclusively between adjacent stages. Mean processing time was 57 seconds per image on a consumer-grade laptop. Given the limited sample size and single cultivar scope, these results represent methodological feasibility rather than validated generalisation, and establish a baseline for morphology-based maturity assessment in pomelo.