Combined Application of Decision Tree and Fuzzy Logic Techniques for Intelligent Grading of Dried Figs

In this study, the quality assessment of dried figs was performed using a machine vision system and data mining techniques. Images from five different classes were obtained using a color CCD camera. After preprocessing and segmentation of the images, 52 features including 6 from size and shape, 4 from texture and 42 from color information were extracted. To find and select the best features for figs grading, the correlation-based feature selection was utilized. It was found that five features (Mouth Area, Homogeneity, Variance value for R, Variance value for B and Kurtosis value for B/R+G+B) surpassed the other features in the attribute selection process. Afterwards, a combined decision tree-fuzzy logic (DT-FL) technique was developed to classify the dried figs based on the superior features. Comparison of validation stage of the utilized DT classifiers indicated that the DT with REP algorithm was the best classifier with an accuracy of 91.74%. Practical ApplicationsManual grading is a costly and time-consuming method for dried figs grading. Besides, human inspectors' judgment in quality inspection is usually incompatible with each other. Another method for dried figs sorting is mechanical method. This type of product sorting is carried out based on shape properties. It is evident that use of such methods cannot guarantee the precise sorting of figs because the other important quality indices such as decay and defects cannot be controlled. Therefore, utilization of new technologies such as machine vision and artificial intelligence can be a suitable solution for automated inspection of figs. In this study, the quality grading of dried figs was performed using a machine vision system and data mining techniques. Using such a combined classifying model in the form of an automatic sorting system, it is effectively possible to sort different grades of dried figs with high levels of reliability, speed and accuracy.