Determination of the quality classes of Elazig cherry marble with image processing

Abstract

Nature has provided the basic resources for the sustainability of human life throughout history. Natural stones have been used intensively by people for different purposes such as shelter, weapons and ornaments. Marble has attracted attention with its variety of colors, aesthetic appearance, durability and rich textural structure and has become one of the most preferred natural stones. This high usage rate has made marble a very valuable mineral in economic terms. Elazig Cherry marble (Rosso Levanto) is a rare type of marble with unique color, pattern and texture features, extracted only from the Elazig - Alacakaya region of Turkey. The evaluation and classification of marble quality is mostly carried out by experts based on observation today. However, this subjective method has important limitations such as high margin of error, economic risks and increased workload. Therefore, the classification process must be transferred to the digital environment; it must be fast and reliable. This study proposes the use of image processing techniques based on color analysis in order to perform the digital classification of Elazig Cherry marble. Based on expert opinions, quality classification metrics were determined and traditional A, B, C quality levels were technically defined. In addition, Class A marbles were divided into detailed subclasses (A1, A2, A3) and the classification process was made more sensitive. In the classification tests conducted using ResNet50 architecture and Support Vector Machines (SVM), which are among the deep learning models, 95.80% accuracy was achieved. The results obtained show that marble producers can use digital classification processes effectively, thus increasing both production efficiency and reducing the workload of employees, thus saving significant time.