Segmentation-free image processing and analysis of precipitate shapes in 2D and 3D

Abstract

Segmentation based image analysis techniques are routinely employed for quantitative analysis of complex microstructures containing two or more phases. The primary advantage of these approaches is that spatial information on the distribution of phases is retained, enabling subjective judgements of the quality of the segmentation and subsequent analysis process. The downside is that computing micrograph segmentations with data from morphologically complex microstructures gathered with error-prone detectors is challenging and, if no special care is taken, the artifacts of the segmentation will make any subsequent analysis and conclusions uncertain. In this paper we demonstrate, using a two phase nickel-base superalloy microstructure as a model system, a new methodology for analysis of precipitate shapes using a segmentation-free approach based on the histogram of oriented gradients feature descriptor, a classic tool in image analysis. The benefits of this methodology for analysis of microstructure in two and three-dimensions are demonstrated.

Investigating local orientation methods to segment microstructure with 3D solid texture

“…HOG has been frequently used for pose normalisation, 2D image description, and classification, like in the popular scale-invariant feature transform that relies on aggregated gradients [10]. Different methods have been proposed to extend HOG to the 3D case [11][12][13] and recent work on simple microstructure classification has used 3D HOG descriptors [11]. HOG is popular to recognise or retrieve shape when good contrast occurs at object boundaries.…”

Texture-Based Metallurgical Phase Identification in Structural Steels: A Supervised Machine Learning Approach

“…Most of the engineering metallic alloys like dual-phase steels, α-β brass, α-β titanium, etc., possess multiphase polycrystalline microstructures [2]. Such microstructures are characterized by grain sizes, distinct phases and their volume fractions, and morphology [3]. Under different mechanical and thermal manufacturing and operating conditions, these microstructural features undergo changes that result in modified bulk properties of the metal [4].…”