Prediction of Banana Quality Using Sensor Network

Abstract

Efficient classification of fruits and vegetables is one of the major challenges of the fruit industry. In the case of large-scale production of fruits and vegetables, storage and maintenance are the main challenges. Some of the techniques developed to do this based on image processing and hyperspectral imaging. Available techniques can classify fruits and vegetables based on their external appearance. It minimises the probability of getting correct prediction results. There is need of automated framework, which should consider not only external factors but also internal properties of fruit sample. This work aims to develop and analyze framework to predict fruit quality using sensors and machine learning. To achieve this, non-destructive experimental setup is designed to extract fruit samples gaseous sensory attributes. Banana fruit samples considered to carry out this experiment. Extracted attributes from fruit samples are stored and clustered into three classes. These three clusters named as, Eatable, Not Eatable and May be Eatable. After clustering, fuzzy inference system is developed based on fuzzy rule-based construction. The designed framework could predict banana samples quality, approximately 88% correctly.