Mineral Research

This project was divided into two periods, with the first period focusing on microscopic mineral pictures and the second period on mineral hyperspectral.

EI Compendex Paper (Published in Chinese):

http://www.earthsciencefrontiers.net.cn/EN/abstract/abstract5961.shtml

​

Mineral classification plays an important role in many research fields. Intelligent mineral identification based on deep learning brings a new development direction to these fields, it can effectively save labor costs as well as reducing classification errors. The purpose of this paper is to study an accurate, efficient and versatile intelligent mineral identification method by deep learning. We trained and tested this method on five kinds of minerals: quartz, hornblende, biotite, garnet and olivine. We used the convolution neural network, commonly applies to image analysis, to establish the model and designed the model structure based on residual network (ResNet-18). In order to support deep learning, we collected microscopic imaging data sets of five kinds of minerals independently, and used them to train, verify and test the model. Besides, we also expanded the data sets for training through reasonable data augmentation. In terms of structural design of the convolutional neural network, we selected ResNets-18 as the framework and finally trained a successful mineral identification model achieving 89% accuracy in the test.

Keywords: deep learning, mineral classification, computer vision, convolutional neural network, residual neural network

With the development of science and technology, human beings are gradually able to monitor the ground conditions in real-time through remote sensing technology, such as disaster warnings, exploration of the composition, content, and distribution of surface minerals. The remote sensing image detected by hyperspectral satellite contains abundant data information. With the help of spectral data, the surface material in the image can be effectively recognized by human beings. In recent years, the research on intelligent mineral identification is increasing, and the methods are various.

Based on deep learning, we showed the mineral hyperspectral identification and generation technology in this paper. First, we collected and summarized the method of previous studies by other people. On the basis of existing studies, we collected hyperspectral data and used a convolutional neural network containing residual structure to identify minerals. In order to do experiments on simulation and data argumentation, we used the generative adversarial network to generate mineral hyperspectral data. The main work and the innovation of this paper include three aspects: 1. Combining remote sensing technology, a deep learning method was applied in the identification and generation of mineral hyperspectral; 2. For mineral hyperspectral identification, the convolutional neural network including residual structure was used, and the mathematical model of cosine similarity was established as a comparison reference, which achieved remarkable results. 3. For a mineral hyperspectral generation, the generative adversarial network as an unsupervised learning method could effectively generate reasonable spectral data.

The selection of mineral varieties, samples, basic models, and parameters all have important influences on the result of deep learning. Finally, the high accuracy and efficiency of the model proved that deep learning could play a good role in mineral hyperspectral recognition and generation, and had a very broad application prospect.

​

Keywords: deep learning, convolutional neural networks, generative adversarial network, mineral hyperspectral, mineral classification