Development of an image processing algorithm for recognition of selected indigenous fruits

Abstract

Fruits are extremely fundamental in our everyday diet as they contain the vast majority of significant nutrients, minerals, and antioxidants. Sorting and grading of fruit are important aspects of its processing. However, these separation operations are still largely done manually in many developing countries including Nigeria. This study therefore, developed an algorithm for identifying and classifying fruit types. The proposed method involved the use of an image acquisition device, which acquired the images of the selected fruits namely apple, onion, banana, pepper and tomato fruits. These fruit images were divided into training and testing data sets. The algorithm extracted the textural and colour features of the fruit images from the training data sets to serve as templates for the testing procedure, after which they were processed using MATLAB software with Support Vector Machine (SVM) algorithm as the classifier. The fruit recognition system classified the input fruit sample by determining the similarities between the attributes (colour and Gray Level Co-occurrence Matrix values) of input fruit samples and the templates obtained from the training data sets. The levels of accuracy of the proposed system for apple, banana, green pepper, red pepper and tomato fruits were 97.2, 97.0, 97.2, 98.1 and 97.2%, respectively. The proposed method proved to be very promising in classifying the selected fruit types based on their colours and textural characteristics.