Rock classification in petrographic thin section images based on concatenated convolutional neural networks

Rock classification plays an important role in rock mechanics, petrology, mining engineering, magmatic processes, and numerous other fields pertaining to geosciences. This study proposes a concatenated convolutional neural network (Con-CNN) method for classifying geologic rock types based on petrographic thin sections. Plane polarized light (PPL) and crossed polarized light (XPL) were used to acquire thin section images as the fundamental data. After conducting the necessary pre-processing, the PPL and XPL images as well as their comprehensive image developed by principal component analysis were sliced into small patches and were put into three CNNs, comprising the same structure for achieving a preliminary classification. Subsequently, these patches classification results of the CNNs were concatenated by using the maximum likelihood method to obtain a comprehensive classification result. Finally, a statistical revision was applied to fix the misclassification due to the proportion differences of minerals that were similar in appearance. In this study, there were 92 rock samples of 13 types giving 106 petrographic thin sections and 2208 petrographic thin section images, and finally 238,464 sliced image patches were used for the training and validation of the Con-CNN method. The 5-folds cross validation showed that the proposed method provides an overall accuracy of 89.97% and a kappa coefficient of 0.86, which facilitates the automation of rock classification in petrographic thin section images.