Görüntü işleme ve yapay sinir ağlari kullanarak mineral tanima

The different methods and algorithms can be used to identify the minerals. One of the methods of identifying minerals is to use the the microscopic / poloarization colors of minerals. Each color has a pixel value in the color spectrum. The pixel values of RGB (Red, Green, Blue) color spectrum are used as parameters to populate the the input data for this study. In the first phase of this study, a lot of images are taken from the thin sections of the minerals by using a digital camera mounted to a microscope. These polorized images are transmitted into the testing and development computer as Joint Photographic Experts Group (JPEG) images. The data which is obtained from the pixel values of one of the images is loaded to the matrixes for further image clustering process. The K-Means algorithm is used to classify the loaded image data based on the pixel values of the mineral. The image data is divided into three clusters using K-Means to avoid the local minima. In the second phase of this study, the output data obtained from the first step is used to train a neural network with forward propagation. Once a neural network with forward propagation training is completed, the neural network identified the minerals. The Alcali Feldspar, Plagioklas, Pyroxene and Quartz Minerals which are used in this study proved with approximately between %12 and %80 accuracy whereas the olivin, muscovite, hornblende, and biotite minerals proved with over than 64% accuracy.