Automatic quality evaluation of fruits using Probabilistic Neural Network approach

Quality and safety are the important factors in food industry. Quality assessment of fruits and vegetables is done based on the analysis of external features such as color, size, shape, texture and presence of damage. The pallet of possible damages to fruits, particularly in apple, is extremely extensive and is often a criterion of quality determination methods. Our purpose, in this study, is to develop a non-destructive method to classify the apple fruits based on the external quality. By studying the damages inflicted on apple fruits, we have presented various feature extraction methods, the output of which were applied as input to train Probabilistic Neural Network (PNN) classifier. We have considered 20 color images of healthy fruits and 45 images of fruits with various damages for training and testing the classifier. The presented supervised classifier is able to distinguish defective fruits from non-defective ones with 86.52% and 88.33% accuracy for different set of extracted features.