Genetic algorithm based parameters adjustments for micron-order image analysis of metal fracture

Abstract

A brittle fracture, which is a kind of metal fractures, accompanies little or no plastic deformations of the metal, The origin of the fracture is one of the most important clues to identify the cause of it, and can be located by observing the surface. Here, fractography is an area of study to analyze fracture mechanisms and/or causes by using electron microscope images. The method to locate the origins has been established through it. However, they have been done mainly by visual observations. At MHS2007, we proposed a method to improve its measurement accuracy to micron order by introducing computer vision techniques on Scanning Electron Microscope, or SEM, digital images [1]. However, some parameters were manually determined and required to be automated. This is a continued research of it. We propose, in this paper, a new method to determine proper parameters for image segmentation using Genetic Algorithms (GAs.) Through experiments, we proved the scheme worked properly, and suboptimal parameters were determined so that the greatest number of cleavage steps was obtained. It is expected that these proposed methods will enable non-experts to analyze metal fractures accurately in micron scale.