Identification of Precious Metals using Deep Learning and Image Processing Methods

Abstract

There are nearly ninety types of minerals in the world. Gold, platinum, palladium, and silver are classified as precious metals. These metals are used in a wide range of fields, including industry, technology, and jewelry. Due to its very high electrical conductivity, gold is used in sensitive technologies such as chips and microprocessors. Platinum and palladium, on the other hand, are used in the coating of aircraft and jet engines, dentistry, and the production of medical devices because they are resistant to high temperatures. The strategic importance of precious metals and their wide range of applications underscore the significance of this study. The primary objective of this study is to perform segmentation of gold, platinum, and palladium metals present in images obtained under a light microscope from samples taken from soil or rock. Through the deep learning-based system developed in this study, the goal is to achieve benefits in terms of time, labor, and cost by performing a preliminary evaluation of precious metals by an expert before initiating testing on advanced and expensive sample analysis devices. Image segmentation was performed using the U-NET architecture, one of the deep learning methods. The results of the system, developed based on the presence of precious metals indicated by the expert, were evaluated using a confusion matrix, achieving an accuracy of 83.84%.