Content-Based Image Retrieval Framework for Classification of Cocoa Beans

Abstract

Companies that process cocoa into chocolates and other cocoa products want fully fermented beans. Fermented beans are of excellent flavour and brownish, while other bean classes do not meet this standard. Apart from using chemically-based methods to evaluate the quality of cocoa beans, visual inspections are widely used at the primary level and have been successfully used to classify beans. The challenge of this method is that it is usually done by humans, which sometimes leads to the misclassification of beans. Computer vision has been proposed for the inspection of beans with an acceptable degree of performance. However, most machine learning-based approaches present models that make it impossible for users to inspect beans that influence machines’ decisions and provide proper feedback to improve performance. This paper proposes a Content-Based Image Retrieval (CBIR) framework that can classify and display beans from the database that influences the classification decision. The framework was able to predict with 100% accuracy as Support Vector Machine (SVM). The performance of the framework stood superior to Naive Bayes (NB), Decision Tree (DT) and Discriminant Analysis Classifier (DAC). In addition, it offers users opportunities to see images that are helpful to decision making..