Multi-carbides segmentation and quantitative morphological parameters extraction with machine vision

Accurate phase segmentation is essential for effective phase identification and quantitative analysis. This paper presents a method for segmenting complex multi-carbides and extracting morphological parameters based on visual features derived from SEM image. Firstly, this method integrates thresholding and morphology algorithms, effectively addressing the challenge posed by a noisy background generated by a substantial presence of free alloys. Secondly, for the unimodal characteristic of grayscale in the SEM image, a segmentation framework is proposed that employs threshold for dual binary classifications in conjunction with bitwise AND operation. The segmented Cr-rich and V-rich carbides are consistent with the EDS, and the extracted contours effectively capture the edge details of the carbides. In addition, the analysis of the quantitative parameters indicates that as the phase area increases, there is a corresponding decrease in roundness and an increase in dispersion. This paper presents a novel and effective approach for achieving unbalanced multi-carbides segmentation, as well as for extracting quantitative morphological parameters of microscopic carbides.