Image processing methods and their applicatoins to marble production

Image processing is a field of study that has started to be used in many different fields in recent years and its importance is increasing. As in many other fields, studies have begun to be carried out in the field of marble processing for the application of image processing methods and techniques in recent years. These applications started with transferring the marble colors to the computer environment, and then studies that could classify marbles of different origins have been realized. After that, the color selection studies for marbles produced from the same block were made by computers by using Artificial Intelligence and Artificial Neural Networks (ANN) techniques. However, there are still some issues to be overcome in order for these studies to turn into applications. First, ANN models can be created and run with expensive and complex professional programs. Their integration into the factory is still problematic. Another issue is processing time. Even it is based on ANN, the color selection process must make a separation at a time suitable for the normal processing time of the factory. In order to make marble color selection with computers, both professional ANN software must be adapted to the factory environment and the processing times must be adapted to the factory processing times. This thesis focused on the second part of the problem; studies and suggestions were made about shortening the processing time and first, a model that makes marble color selection with image processing techniques supported by ANN was created and successful results were obtained. Then, two different techniques were proposed to shorten the processing time, for each technique, a total of 29 ANN models were reconstructed for 14 and 15 different operating conditions, respectively, to find the best option. At the end of the study, the processing time reduced from 103 seconds to 34 seconds with a model correlation of 0.929 by using the suggested technique named K-40. With this suggested technique, the processing time of ANN-based marble color classification decreased by 67%.