Image analysis and mass petrography

点击翻译按钮获取中文摘要

Petrography - the optical microscopic analysis of thin sections of materials - is the fastest and most cost-effective tool for investigating the mineralogy and microtexture of rocks, ceramics, and other solid materials. Rock, soil and archaeological thin sections have been studied for over 150 years, with samples in archives around the world. The data produced from petrographic analysis underpins the exploration for and processing of natural resources - including the metals we will need for renewable energy technologies - materials used in construction, and forensic studies of objects both recent and ancient.Petrographic analysis has traditionally required an appropriate optical microscope, access to the physical thin section samples, and an expert user to classify, measure and interpret the specimen. Samples could be viewed, and thus analysed, only one-at-a-time. Expert users can obtain qualitative data within minutes using a basic microscope, but quantitative analysis is time consuming. However, ground-breaking new optical microscopy technology (see figure) allows multiple sections to be scanned simultaneously to produce high quality images very rapidly. We will pair this advance with computer-based image analysis for classification and quantification, enabling thin section archives and previously-unavailable derived data to be brought into a digital environment.This PhD will develop a new, highly automated, workflow that will unlock the full value of optical analysis and combine human levels of identification and classification with superhuman levels of speed and reproducibility. These approaches will allow for quantitative and statistically-robust approaches to mineralogy and petrology, and for the rapid development of reference databases for archaeological materials. These in turn unlock the potential for new research in geology, archaeology and material science, with deeper insight into textural and mineralogical diversity of natural and archaeological materials.