Automated Skin Defect Identification System for Fruit Grading Based on Discrete Curvelet Transform

The purpose of this study was to develop a methodology for assessing fruit quality objectively using texture analysis based on Curvelet Transform. Being a multiresolution approach, curvelets have the capability to examine fruit surface at low and high resolution to extract both global and local details about fruit surface. The fruit images were acquired using a CCD color camera and guava and lemon were analyzed by experimentation. Textural measures based on curvelet transform such as energy, entropy, mean and standard deviation were used to characterize fruits’ surface texture. The discriminating powers of these features for fruit quality grading is investigated. The acquired features were subjected to classifiers such as Support Vector Machines (SVM) and Probabilistic Neural Networks (PNN) and the performance of classifiers was tested for the two category grading of fruits namely healthy and defected. The results showed that best SVM classification was obtained with an accuracy of 96%. The study concludes that curvelet based textural features gives promising insights to estimate fruit’s skin damages.