Pattern recognition for texture classification of ornamental stone slabs

The recognition of similarities in the texture and in the characteristics of ornamental stone slabs is of fundamental importance to determine their proper utilization for ornamental purposes and it is of great interest in applications. Its economic importance is well known, to determine the quality of a shipment and its commercial value. The problem of evaluating in a routine fashion these aesthetic characteristics, as the above aspects shall be called, is a challenging problem, which at present has not been satisfactorily solved. These aspects may be evaluated by petrographic analysis and by a series of ad hoc and subjective criteria. These approaches result in a non reliable classification and the consideration of a large number of characteristics of the objects, each measured in an arbitrary way, with uncertainty as to its significance in the overall evaluation. Pattern recognition approaches have also been attempted, both using the statistical and the structural approach, but their success has been meagre. In this paper, a combined approach, presenting both statistical and structural elements will be attempted. The purpose of this paper will be to define a procedure to recognize similar patterns and to classify suitably the ornamental stone slabs. An algorithm will be described and preliminary results will be given for a large sample of over 300 stone slabs.