Limestone lithological classification, using image processing and pattern recognition technique

Image processing is a technique that simulates the human vision system. This technique enables applying every statistical or intelligent operation to recognize differences. In this way this new technique is used in quality control systems in most industries. Studying sedimentary rocks is very important from the economic point of view. Some units of sedimentary sequences involve different mineral deposits, hydrocarbon and water resources. Therefore it is important to identify and classify different sedimentary rocks. Descriptive classification of sedimentary rocks is usually based on visual and textural features and chemical composition of a sample. In this paper different samples of a limestone mine in central part of Iran are classified. The samples were collected from different parts of the mine and crushed down in size from 2.58 cm to 3.58 cm. The rock samples were labeled based on percentage of chemical and lithological compositions. Each sample was assigned to one of the distinguished groups. The images of the samples were taken in appropriate environment and processed. A total of 74 features were extracted from the identified rock samples in all images. In order to feature dimensional decrease, principal component analysis method was used. Then Bayesian statistical algorithm was used as a useful tool for classification. Classification Correctness Rate (CCR), calculated for the test data sets are %88, %68, %61 and %72 for the first to fourth class respectively. Therefore it can be inferred that the extracted features of images are appropriate indicators for different samples identification. These precise results besides the advantages of image processing technique, which are increasing speed of operation and decreasing cost, appears to be a desirable success.