Advances in Rock Petrography: Image Processing Techniques for Automated Textural Thin Section Analysis

Abstract

Abstract Point counting is time consuming and requires extensive geologist/petrographer effort. In addition, point counting results are subjective and depend on the petrographer's knowledge and expertise. In this work, we introduce a fully automated workflow for thin section textural analysis in clastic rocks, using high resolution petrographic images of the thin sections acquired with a digital camera mounted on an optical microscope. This innovative workflow reduces the thin section textural analysis turnaround time and provides an objective and consistent analysis. The strength of this workflow resides in its high level of automation, which offers thin section analysis tool in much less time compared to the conventional point counting. The workflow is fully automated to process and analyze the entire thin section without manual involvement. The kernel of this workflow is based on a region growing algorithm for individual grain identification. An iterative loop, built on the top of this kernel, allows the completely automated scan of the entire thin section. The workflow was first rigorously validated for a single thin section. Grain by grain, results from the automated analysis are compared to the petrographer (point counting) analysis. Excellent agreement between the two analyses was obtained (porosity and grain size). The efficiency of the analysis was largely in the favor of the automated approach (3 minutes) compared to the 2 hours needed by the petrographer for this counting exercise (approximately 150 grains). This first validation test proved the workflow's accuracy and the efficiency. This workflow was then extensively validated using large set of thin sections (50 thin sections) showing an excellent qualitative agreement with conventional point counting. This second validation test proved the robustness and the efficiency of the workflow.