Classification of cotton oil in the semi-refining process using image processing techniques: Image processing for industrial applications

This paper presents a steps model classification in the cotton oil semi-refining process by image processing. A simple system of efficient analysis was obtained, allowing its use on the factory floor, assisting in the process with respect to the assertiveness in the desired semi-refined state and its quality control. Two algorithms with different recognition patterns were analyzed: K-Nearest Neighbor and Quadratic Discriminant Analysis, with a validation strategy Leave-one-out and the neural network Extreme Learning Machine using K-fold validation technique. Both algorithms reached success rates of more than 90%, and the best result was obtained through the applied neural network, with more than 95% of accuracy.