Classification of Date Fruit Varieties Based on the Use of Capacitive and Force Transducer Responses

AbstractIn many date palm gardens, different varieties of date fruits are commonly grown. Each of these varieties has different physical and chemical properties. Therefore, in the first step of developing intelligent sorting systems and quality monitoring systems, it is highly desirable to develop an automatic method for identifying the type of date fruit variety. This study aims to develop a classification algorithm of date fruit varieties using electronic responses from capacitive sensors and a force transducer. To simultaneously obtain the capacitive properties including the frequency and analog responses of date fruit and the force parameter related to the weight of the date fruit, an electronic platform was designed and constructed. In this platform, the capacitive responses and force parameters were provided by setting up the IC 555 and calibrating a 1 kg load cell. 120 date fruits from four different varieties, including Zahedi, Qhasb, Mazafati and Medjool were considered for the model development. A total of 30% of the samples were separated for final classification evaluation.  Decision Tree (DTs) as a non-parametric supervised learning method was chosen for classification. To tune, the classification model with the best hyper parameters fitting 3 folds for each of 2204 candidates, totaling 6612 fits were examined. The evaluation of the developed model resulted in a classification quality with F-scores of 64%, 50%, 100% and 100% for the four varieties Zahedi, Qhasb, Mazafati and Medjool. It was concluded that the developed capacitance base classification system can be successfully used to classify date fruit with acceptable quality.