Lithological Classification of Brazilian Pre-salt Rocks with GLCM Features and Machine Learning

Summary Lithological classification represents a method applied to discern and interpret unique rock structures, offering crucial insights into their petrophysical, morphological, textural, and geological characteristics. This process holds particular significance in the carbonate sedimentary rocks within petroleum basins due to their substantial capacity to store natural gas and oil, forming an intrinsic correlation with oil reservoir productivity. We introduce an automated framework for the lithological classification of specific carbonate rocks, categorizing them into seven distinct classes. The approach involves a comparative analysis of two established machine learning algorithms and a conventional feature extractor. Through experiments conducted on a proprietary dataset from a Brazilian petroleum company, the results indicate that Random Forests combined with GLCM exhibit promising potential for this classification task.