A ROBUST COMPUTER VISION–BASED METHODOLOGY FOR AUTOMATED DETECTION AND QUANTIFICATION OF M/A CONSTITUENTS, ACICULAR FERRITE, AND FERRITE WITH SECONDARY PHASES IN WELDED STEELS

Abstract

Abstract: Quantitative metallographic evaluation of welded steels remains largely dependent on manual interpretation, resulting in limited reproducibility, strong operator dependency, and insufficient morphological characterization of critical microstructural constituents. In particular, the reliable identification of martensite–austenite (M/A) constituents, acicular ferrite (AF), and ferrite with secondary phases (FS) in weld metal microstructures presents a persistent challenge under conventional optical microscopy. This paper presents a robust and fully automated computer vision–based methodology implemented in C++/OpenCV for objective phase segmentation and quantitative analysis of etched weld metal micrographs.