A Quantitative Analysis of Void Detection in Scanning Electron Microscopic Images(SEM) using Machine Learning
Abstract
Image Processing is an advanced technique in extracting useful information from the image. Scanning Electron microscopic (SEM) produces images from the samples by scanning the surface. It provides microanalysis information but the experts are not able to make quantitative analysisin presence of voids, imperfections or contaminants in both metal and non-metal SEM images. Voids causes more damages in materials strength. To overcome this problem, supervised machine learning concept, classification and regression technique is used.It classifies according RGB format and sets a colour limit. The regression technique helps as output as voids occupied area and unoccupied area. It predicts the voids in statistical manner and provide accurate results to the experts. Here we used improvised classification and regression algorithm, which provides a better result in identification of voids in SEM images so the users can improvise the strength of material at earlier stage.
Keywords:Scanning Electron microscope, Supervised Machine Learning, Classification Algorithm, Regression Algorithm, Metal and Nonmetal Image.