Early Detection of Bruises in Khasi mandarin (Citrus reticulata Blanco) for the Assessment of Post Harvesting Losses: A Machine Learning Approach

Abstract

This paper explores the application of machine learning (ML) and thermal imaging (TI) for early detection of bruises in Khasi mandarin (Citrus reticulata Blanco), aiming to reduce supply chain losses by identifying damaged fruit before deterioration becomes visually apparent. Leveraging the principle that materials emit distinct infrared radiation based on their physicochemical properties, thermal imaging is used to differentiate bruise from unbruise Khasi mandarins. A machine learning model is applied for classification, successfully analyzing thermal images to identify subtle variations indicating early damage. The thermal images showed that a temperature difference of more than 0.5°C between the bruise and unbruise areas enhanced the detection process. The results demonstrate the possibilities of combining thermal imaging and ML for non-destructive and efficient fruit quality monitoring. This approach offers a reliable method for early identification of fruit damage, enabling timely interventions that prevent further deterioration and minimize post-harvest losses. The study underscores the possibilities for integrating advanced imaging and machine learning techniques in agricultural quality control. Future research with larger datasets can improve model accuracy, benefiting stakeholders across the fruit supply chain and supporting industry sustainability.