Loading...
Pre-filtering micrograph data to train machine learning algorithms to optimize medical metals
Burke, Reese Catchings, Avary E. Lowe, Beatrice Lowe, Terry C.
Burke, Reese
Catchings, Avary E.
Lowe, Beatrice
Lowe, Terry C.
Citations
Altmetric:
Advisor
Editor
Date
Date Issued
2025-04
Date Submitted
Keywords
Collections
Research Projects
Organizational Units
Journal Issue
Embargo Expires
Abstract
Slip band analysis plays a crucial role in understanding material deformation and failure. Our research focuses on developing a program to automate the measurement of average slip band spacing within metal grains. The goal is to create a filtering algorithm that differentiates individual grains and identifies the average distance between slip bands within each grain. Our current approach involves image preprocessing using the Frangi filter, pixel equalization, scaling, thresholding, and then applying the Canny filter to isolate slip bands. We then apply the Hough Line detection algorithm to identify slip bands and compute their angles relative to the x-axis. By determining the grain width perpendicular to the average slip band angle and counting the slip bands, we estimate their average spacing. While this method has been successfully implemented for a single grain, further validation is required across multiple grains before conclusive analysis can be performed. Initial tests with alternative edge detection techniques, such as Sobel and Canny filters, revealed that Frangi filtering provides the most effective segmentation. A key challenge remains in defining crystal boundaries, as slip bands within a single grain align consistently, but outlining grain boundaries is complex. To address this, we are exploring blurring, labeling, thresholding, and masking techniques to enhance boundary detection. This research has broader implications for materials science, particularly in predicting metal failure and optimizing biocompatible materials. Our findings can apply to another project on our research team focused on improving metal/human tissue interfaces for medical applications.
Associated Publications
Rights
Copyright of the original work is retained by the author.