AUTOMATED QUANTIFICATION OF GRAIN SIZE AND PHASE RATIO IN DUPLEX STEEL MICROSTRUCTURES USING OPENCV

Abstract

The microstructural evaluation of duplex stainless steels plays a crucial role in determining their strength, toughness, and corrosion resistance. Conventional manual measurement of grain size, ferrite–austenite phase balance, and heat-affected zone (HAZ) width is labor-intensive and strongly dependent on operator perception, which often leads to inconsistency in results. In this work, an automated computer‐vision‐based methodology is proposed for fast and reproducible quantitative metallographic analysis using the OpenCV library.

The developed workflow includes image normalization, grayscale conversion, histogram equalization, adaptive thresholding, and morphological refinement to isolate ferrite and austenite regions. Grain boundary extraction is performed using Sobel and Canny operators, followed by watershed segmentation to accurately separate adjacent grains. Pixel measurements are calibrated to micrometers based on the scale bar present in the microscopic images, enabling precise dimensional quantification.