Visible imaging to convolutionally discern and authenticate varieties of rice and their derived flours

Abstract

In this research, more than 27,000 images (samples) of five different types of rice (Oryza sativa L.) have been used to design and validate a deep learning-based system to carry out their classification. A typical photographic camera was used to obtain images from five different varieties of rice, which will be used for their classification after proper treatment. The resulting photographs were processed by convolutional neural networks (CNNs), which have been trained and optimized using these images to identify different types of rice. Finally, the model was successfully validated using images which were initially isolated from the training database. The result was an algorithm capable of detecting and classifying all five rice types accurately. Therefore, CNNs have shown to be a compelling tool for the evaluation of this type of cereal, including for quality purposes, thanks to traits which include high sensitivity, speed, and not requiring highly specialized personnel to implement the optimized version of the algorithm.