Classification of Rice varieties using DMLP-PCA inspired features with MVE Classifier

The classification of varieties of rice from images is one of the most difficult tasks in computer vision. This complex process is normally applied in automating packaging systems used in production companies. Most companies opt for this option because the alternative way of doing it manually is time consuming, monotonous and prone to expensive errors. To refine the process, there is need to interrogate the features because a classification algorithm is as good as the features used. Therefore, this paper presents Deep Multi-Layer Perceptron generated features. In addition, the paper also introduces the Majority Voting Ensemble (MVE) classifier. The technique was evaluated against CNN (Convolution Neural Network) generated features as well as other traditional classifiers. The proposed solution performed better than the other methods including end-to-end deep learning models.