Classifying Scanning Electron Microscope Images Using Deep Convolution Neural Network

The research aims to classify high-temperature materials with wide applications such as electronic, re-entry vehicles, and semiconductors. The challenging act is to extract unique features as the images are microscopic with different resolutions. The images captured from the SEM (Scanning Electron Microscope) machine are classified according to their crystal type, for SiO2, CCC, silica tile, carbon fiber, CeZrO2 using Convolutional Neural Network (CNN), which is a deep learning framework. Images obtained by XRD (X-ray diffraction) machines are classified according to the crystal structure (such as crystalline, amorphous, and tetragonal) irrespective of the material. An ensemble-CNN-based classifier is designed to train and classify (SEM and XRD) images with accuracy.