Abstract

In this paper, the classification of ornamental dolomitic marble stone tiles, in regard to their aesthetical value, was studied based on the rock’s texture. The stone tiles examined are of a dolomitic marble variety commercially known as Lais Grey. Twenty four (24) texture descriptors and seven (7) machine learning models were tested in order to find the best performing combination. The experimental study was conducted with an in-house dataset consisting of three tile classes containing digital images selected by an expert. A second dataset was compiled by applying clustering using the k-means algorithm, towards defining the tiles’ quality based on texture information. This process produced a dataset with two classes. The results revealed that the XCS-LBP texture descriptor joined by the XGBoost classifier achieved the best performance for screening the tiles into three (with 65.06% F1-score) or two (with 99.43% F1-score) quality classes.

Citation

G. Sidiropoulos, A. Ouzounis, G. A. Papakostas, I. Sarafis, A. Stamkos, G. Solakis, “Texture Analysis for Machine Learning Based Marble Tiles Sorting,”IEEE 11th Annual Computing and Communication Workshop and Conference (CCWC).