Abstract Mineral identification is a basic skill in geological studies, and is useful for characterizing rocks and tracing diagenesis and mineralization processes. Traditional methods of observation under a microscope are subject to many complex factors such as the limitations of resolution and magnification, so they are poor in qualitative analysis, and inefficient. With the expansion of geological prospecting, it is necessary to provide information for all minerals, pores and trace elements in rocks. So, mineral identification has started to rely on advanced microbeam mineral analysis techniques. This paper summarizes the common mineral analysis techniques such as Raman spectroscopy, X-ray fluorescence spectrometry (XRF), X-ray diffraction (XRD), Scanning electron microscopy (SEM), and Automated mineralogy (AM) systems. These microbeam technologies now approach a semi-automated analysis process, and most of these methods mainly detect the chemical composition of the mineral, rather than the mineral's optical characteristics which are the most basic properties of minerals. Therefore, this study proposes a method that can use mineral's optical features for automatic classification, mineral recognition based on convolutional neural network (CNN) and face recognition technology. The feasibility, research status and outlook of this method are also discussed. The proposed method uses convolution neural network technology to automatically extract the optical characteristics of minerals for mineral identification. Successful application of these techniques will have profound application value by reducing the cost and time needed to process and identify minerals.

中文翻译：

---

矿物识别回顾及其未来

摘要 矿物鉴定是地质研究的一项基本技能，可用于表征岩石和追踪成岩成矿过程。传统的显微镜下观察方法受分辨率、放大倍数限制等诸多复杂因素影响，定性分析能力差，效率低下。随着地质勘探的扩大，需要提供岩石中所有矿物、孔隙和微量元素的信息。因此，矿物鉴定开始依赖先进的微束矿物分析技术。本文总结了常见的矿物分析技术，如拉曼光谱、X 射线荧光光谱 (XRF)、X 射线衍射 (XRD)、扫描电子显微镜 (SEM) 和自动化矿物学 (AM) 系统。这些微束技术现在接近于半自动化的分析过程，并且这些方法中的大多数主要检测矿物的化学成分，而不是矿物的最基本性质的矿物光学特性。因此，本研究提出了一种可以利用矿物的光学特征进行自动分类、基于卷积神经网络（CNN）和人脸识别技术的矿物识别方法。还讨论了该方法的可行性、研究现状和展望。该方法利用卷积神经网络技术自动提取矿物的光学特征进行矿物识别。这些技术的成功应用将通过降低处理和识别矿物所需的成本和时间而具有深远的应用价值。