Microstructure segmentation using multi-angle polarized light microscopy

• Published in: Materials characterization 2022-10, Vol.192 (C), p.112197, Article 112197
• Main Authors: McCabe, Rodney J., Savage, Daniel J., Bhattacharyya, Abhishek, Gogusetti, Renuka, Alexander, David
• Format: Article
• Language: English
• Subjects: Polarized light microscopy; Segmentation; Titanium
• ISSN: 1044-5803; 1873-4189
• DOI: 10.1016/j.matchar.2022.112197

Grain size and shape are important microstructure attributes affected by material processing and contributing to material properties. We develop automated tools to segment multi-angle polarized light microscopy (PLM) images of low symmetry metals such that statistics of grain size and shape can be automatically measured. To demonstrate the tools, we apply PLM to alpha‑titanium, an hcp crystal structure metal with axisymmetric optical properties and a good candidate for the technique. The segmentation routine uses a region growing algorithm based on similarity of the intensity variance of neighboring pixels for three sample rotations, and grain dilation that checks the proximity of pixels to grain boundaries and triple points. For relatively fine-grained Ti, the segmentation is accurate for 97.5% of the grain boundaries in the microstructure when compared to EBSD results, and the accuracy is expected to improve for larger grained materials. The technique does a good job of defining twin features in deformed materials that, due to twin size and optical limitations, exhibit substantial intensity gradients making segmentation challenging. •Automated multi-angle polarized light microstructure segmentation.•Segmentation of twinned titanium microstructures.•Better than 97.5% accurate compared to EBSD.•Large microstructure area segmentation.