A New Approach for Calculating Texture Coefficients of Different Rocks With Image Segmentation and Image Processing...

Area of Science:

• Geology
• Materials Science
• Computational Science

Background:

• The texture coefficient (TC) is vital for analyzing rock microstructures and predicting mechanical properties.
• Current methods for estimating rock TC are often time-consuming and lack sufficient accuracy.
• Computational tools are increasingly used, but limitations persist in rock TC determination.

Purpose of the Study:

• To develop a novel, efficient, and accurate method for calculating the texture coefficient (TC) of rocks.
• To leverage deep learning and image processing for automated TC estimation.
• To reduce the time required for TC analysis in geological and materials science applications.

Main Methods:

• Acquisition and segmentation of thin-section images from 20 diverse rock types (igneous, metamorphic, sedimentary).
• Implementation of a Python-based software integrating deep learning for image segmentation.
• Development of a Python algorithm for calculating TC values from segmented images.

Main Results:

• Achieved high segmentation accuracy with an Intersection over Union (IoU) score of 0.97.
• Successfully computed TC values for various rock types using the novel approach.
• Reduced the computation time for TC estimation to approximately one minute per rock sample.

Conclusions:

• The proposed deep learning-based image processing method provides a highly accurate and rapid solution for rock texture coefficient (TC) estimation.
• This Python-based approach overcomes the limitations of existing time-consuming methods.
• The technique offers a significant advancement for microstructural analysis and mechanical behavior prediction in rocks.