PROJECT TITLE :

Machine Learning Driven Approach Towards the Quality Assessment of Fresh Fruits Using Non-invasive Sensing

ABSTRACT:

Accurate moisture content (MC) information in fruits and vegetables in an automated manner might be critical for insightful quality and grade evaluation in agriculture science. To preserve a healthy sensory characteristic of fruits, a practical, feasible, and cost-effective technique for defect recognition employing prompt detection of MC in fruits and vegetables is required. Using terahertz (THz) waves with Swissto12 material characterisation kit (MCK) in the frequency range of 0.75 THz to 1.1 THz, we present a non-invasive machine learning (ML) driven technique to monitor fluctuations in MC in fruits. In this case, multi-domain characteristics from the time, frequency, and time-frequency domains are collected, and three machine learning techniques, such as support vector machine (SVM), k-nearest neighbor (KNN), and Decision Tree (D-Tree), are used to precisely analyze MC in both apple and mango slices. Using 10-fold validation and leave-one-observation-out-cross-validation procedures, the results revealed that SVM outperformed other classifiers. Furthermore, for days 1 and 4, all three classifiers had 100% accuracy, with 80 percent MC value (freshness) and 2 percent MC value (staleness) of both fruits' slices, respectively. Similarly, with intermediate MC values in both fruits' slices, an accuracy of 95% was attained on days 2 and 3. By combining machine learning with THz sensing at the cellular level, this study will pave the way for non-invasive real-time fruit quality evaluation. It also has the ability to maximize economic benefits by detecting the quality of fruits in a timely and automated manner.