岩性识别：方法、现状及智能化发展趋势

岩性识别是地质工作中一项基础而又重要的工作。传统的岩性识别方法过于依赖经验和地质专业知识积累，不仅耗时长、专业性强，还易受主观因素影响，导致准确率不理想。笔者等首先回顾了传统的岩性识别方法，之后总结了最新涌现的智能化识别方法，最后详细介绍了基于岩石图像、镜下图像、图像与元素信息融合等的智能识别方法。基于岩石图像的识别方法对于文中的岩石识别准确率可达90%以上，基于图像与元素融合的岩性识别方法可以降低图像相似度高、风化破坏表观特征等因素对识别准确度的影响。笔者等认为当前岩性智能化识别研究仍处于初级阶段。综合各类数据源的优势，利用机器学习深度挖掘岩石元素、矿物、光谱和表观特征间的内在关联性，有利于突破单源信息的局限性，实现岩性快速准确识别。

Lithology identification: Method, research status and intelligent development trend

Lithology identification is basic and important for geological work. Traditional lithology identification method is overwhelmingly dependent on manual experience and geological expertise, which is time- consuming and highly professional. Besides, it is vulnerable to subjective factors, resulting in unsatisfactory accuracy. We review the traditional lithology identification methods, then summarize the latest intelligent identification methods. Finally, we detailed introduce the intelligent identification methods based on rock images, microscopic images, images and element data fusion. The accuracy of intelligent lithology identification based on images can reach more than 90%. The lithology identification based on images and element data fusion can alleviate the influence of factors on accuracy such as high image similarity and apparent characteristics of weathering damage. We deem that the current research on intelligent lithology identification is still in its infancy and cannot meet the engineering needs. Combining the advantage of various data sources and using machine learning to deeply mine the internal correlation between rock elements, minerals, spectra and apparent features is conducive to breaking through the limitation of single source information and realizing rapid and accurate lithology identification.