An automatic visual system for marble tile classification

The international market of ornamental stone has seen the use of materials quarrying from East Europe, Africa, Asia and South America, characterized by a low cost of labour. This event has embittered the competition in this field. Hence, it becomes fundamental to produce machined products of high quality in a more efficient way. This work aims to design and fulfil a prototype to automatically classify the Royal Perlato tiles of Coreno. It is based on artificial vision. The implemented hardware visual system is cheap because it is made up of components readily available and is easily obtainable by small and medium ornamental stone businesses. The developed software algorithm is based on a multifactor analysis of variance and a quadratic discriminant analysis. The experimental results show that the developed hardware and software system permits all the tiles of Royal Perlato of Coreno to be classified properly. Moreover, the developed system not only improves the efficiency of the selection but also reduces the controversies between the manufacturer and buyer, making tile classification more selective and increasing the number of qualitative classes taken into consideration. Finally, it allows the duplication on the selection line without altering data dispersion.