The Automated Dunham Classification of Carbonate Rocks Through Image Processing and an Intelligent Model

摘要

An automated model for classifying the carbonate rocks based on Dunham classification is presented. The proposed model works based on the integration of image analysis and artificial neural network. Images of thin sections, captured by a digital camera attached to an optical petrographic microscope, were employed as the inputs of the model, while the outputs were four classes of Dunham classification. To get this goal, several parameters of each image were studied to investigate their worth in the process of classification and network training. These parameters were automatically extracted from each image based on image processing techniques. The palpitant heart of the automated model is feature extraction step, which specifies the difference of images for the neural network. For training the neural network, images of 138 thin sections were used, whereas for investigating the accuracy of the model, images of 44 thin sections were employed. The high accuracy of 81.6%, for the previously unseen test samples, confirmed the validity of the proposed model to classify automatically textures of carbonate rocks.