Mineral compositionanalysis of rock image based on deep learning feature extraction

Deep learning was used for rock identification.The image data of 15 kinds of common rocks were collected, and a deep learning model of rock recognition was constructed based on a convolutional neural network.The rock identification accuracy rate reached 63%.Analysis of rock identification results, dolomite, limestone, marble, gabbro and basalt rock which have close mineral composition were easy to mistake each other, so the characteristics of the mineral composition was very important for rock identification.Further more, the feature map produced in the learning process of convolution neural network was analyzed, and the minerals from various types of rocks were extracted successfully, such as quartz, feldspar and mica in granite, amphibole and plagioclase in diorite, sericite and other minerals in phyllite.The results showed that the deep learning method could effectively extract the mineral composition feature of rock, and the deep learning could effectively identify the rock, and it was helpful to identify the rock according to the mineral composition.