Deep learning approaches for bruised mandarin orange classification via fluorescence hyperspectral imaging

Authors: LEE, AHYEONG - National Institute Of Agricultural Sciences (RDA); HONG, SUK-JU - National Institute Of Agricultural Sciences (RDA); Baek, Insuck; KIM, JINSE - National Institute Of Agricultural Sciences (RDA); Kim, Moon

Submitted to: Postharvest Biology and Technology
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 6/12/2025
Publication Date: 6/19/2025
Citation: Lee, A., Hong, S., Baek, I., Kim, J., Kim, M.S. 2025. Deep learning approaches for bruised mandarin orange classification via fluorescence hyperspectral imaging. Postharvest Biology and Technology. 230: 113724. https://doi.org/10.1016/j.postharvbio.2025.113724.
DOI: https://doi.org/10.1016/j.postharvbio.2025.113724

Interpretive Summary: Widely consumed around the world, citrus fruits are susceptible to bruising, which is not always easily discernable by eye but is detrimental to quality. When discovered, bruise damage in the fruits will disappoint customers to the detriment of consumers’ perception of general fruit quality for future purchases, as bruises are not only unappealing but also can be vulnerable areas for pathogens to affect the fruit. This study examines the potential use of non-destructive hyperspectral fluorescence imaging as a tool for post-harvest detection of bruised mandarin oranges, specifically utilizing three multivariate data analysis methods and three deep learning models. The models demonstrated over 99% classification accuracies, showing that hyperspectral fluorescence imaging with deep learning models is feasible for developing automated sorting technologies for the fruit. Development of such non-destructive bruise detection methods for automated post-harvest sorting to remove damaged fruit before it reaches the fresh market for consumers will be of great benefit to both producers and consumers with respect to market economics and to public food safety.

Technical Abstract: Citrus fruit is extensively consumed worldwide, and the bruising of these fruits significantly affects their quality, impacting consumers’ purchasing decisions. Detecting such damage during post-harvest operations is crucial. However, the bruised area does not exhibit distinct color differences compared with normal regions, complicating visual inspection and rendering it time-consuming. Consequently, this study investigates the potential of fluorescence hyperspectral imaging to discern bruised mandarin oranges. Hyperspectral images were acquired following illumination with a pair of 365 nm UV lights. Three multivariate data analyses—decision tree, support vector machine, and partial least squares discriminant analysis—and three deep learning models—ResNet50, EfficientNetB0, and MobileNet—were employed for the classification of bruised mandarins. Preprocessing steps, including dark and white correction, spectra preprocessing, and region of interest (ROI) extraction, were conducted prior to model development. Classification accuracy was determined through model training. Among the models, ResNet50 with nine principal component images exhibited high classification accuracy: 99.65% for the training group, 100% for the validation group, and 100% for the test group. GradCAM visualization further confirmed the successful formation of heatmap over bruised areas. Thus, the classification of bruised mandarins using fluorescence hyperspectral imaging and deep learning is feasible, laying the groundwork for automated sorting technologies.