Classifying slate tile quality using automated learning techniques

Abstract Slate classification has typically been performed manually by an expert on the basis of an assessment of standard defects that characterize the quality of the slate. We describe an innovative and objective approach to automated classification that eliminates subjectivity and human error. The slate is classified using machine learning techniques based on numeric variables obtained from 2D–3D images captured by a linear 2D camera and a 3D laser scanner, which provide the necessary information on the slate. With a view of obtaining an optimally efficient classification model, we implemented supervised machine learning techniques (support vector machines and multi-layer perceptron neural networks) and non-supervised techniques (cluster analysis and self-organizing maps). The results obtained in our research demonstrate that the error for automated classification was lower than for manual classification. Automated classification also removes the subjectivity implied by manual quality control in the slate classification process.