An Intelligent Computer Vision System to Rock Classification in Oil and Gas Industry

This paper explores the use of a hierarchical neuro-fuzzy model for image classification of macroscopic rock texture. The relevance of this study is to help Geologists in diagnosing and planning the oil reservoir exploitation. The same methodology can be also applied to metals, in order to classify the different types of materials based on their grain texture. We present an image classification for macroscopic rocks, based on these texture descriptors and on a neuro-fuzzy approach. Then, a neural network is used, and the results obtained by both approaches are compared. To evaluate the system performance we used 50 RGB images, for each rock class and subclass, thus producing a total of 800 images. The classes of igneous rocks that make up the image database are: gneiss (two classes), basalt (four classes), diabase (five classes), and rhyolite (five classes). For each image were extracted: Hurst coefficient for gray and color images (a coefficient for each RGB channel); spatial variation coefficient (gray and color); entropy and cooccurrence matrix. Tests converged to optimum solution for the classification taking into account the fuzzy rules extraction which had a good performance.