Enhancing mango quality control: A novel approach to spongy tissue inspection through image clustering and machine learning models via X‐ray imaging
Patil Rajvardhan Kiran,Pramod Aradwad,T. V. Arunkumar,Roaf Ahmad Parray
DOI: https://doi.org/10.1111/jfpe.14664
2024-06-13
Journal of Food Process Engineering
Abstract:X‐ray imaging‐based analysis aims to detect Spongy Tissue disorder in Alphonso mangoes. By employing diverse augmentation techniques and K‐means clustering segmentation, our approach promises precise identification, signifying a significant advancement in machine learning‐based binary classification for non‐destructive assessment and improved mango quality assurance. The globally renowned Alphonso mango is esteemed for its exceptional flavor, fragrance, texture, and extended shelf life. However, it faces a substantial challenge with spongy tissue (ST) disorder, affecting up to 30% of a single lot. This issue leads to the rejection of entire batches during export due to delayed detection, causing significant produce loss. Additionally, the current destructive assessment method is inadequate, as physically inspecting only a few mangoes within a batch does not ensure overall quality, emphasizing the need for a non‐destructive approach. To address these challenges, this study proposes a non‐destructive approach utilizing X‐ray imaging to identify ST in Alphonso mangoes, employing machine learning (ML) for binary classification. We captured and augmented 216 X‐ray images, utilizing various augmentation techniques. K‐means clustering effectively distinguished healthy and ST‐affected areas in mango fruits. For automated classification, ML models were fine‐tuned through grid search hyperparameter optimization within a 5‐fold cross‐validation framework. The CNN model achieved impressive accuracies of 99.20% and 96.82% at threshold values of .5 and .95, respectively, on the test dataset, while the SVM model attained overall accuracy rates of 95.23% and 87.35% for the same thresholds. Our study demonstrates the potential of machine learning in X‐ray imaging for detecting internal fruit issues, improving inspection for disease and damage identification, and enhancing quality control across the food industry. Practical implications include the implementation of non‐destructive inspection methods to reduce produce loss and ensure consistent quality in mango lots. Future research could focus on validating the model's efficacy with external datasets, evaluating its robustness in detecting various internal irregularities, and exploring advancements in X‐ray imaging technology for broader applications in agriculture. Practical applications The suggested non‐destructive approach provides a practical solution for quality control, especially in export scenarios. The versatility of the ML‐powered X‐ray imaging technique to examine internal irregularities in various fruits beyond mangoes broadens its potential application in the fruit industry for sorting and grading purposes. This innovation contributes to heightened consumer satisfaction, providing a cost‐effective and scalable solution for widespread implementation. In summary, the findings of the study show promise for revolutionizing mango quality control, particularly for the esteemed Alphonso variety, and propelling advancements in fruit inspection methodologies across the industry.
food science & technology,engineering, chemical