Texture Analysis and Support Vector Machine for Classifying SEM images of VA-CNTs

Abstract

Dispersion of Carbon Nanotubes (CNTs) is one of the most substantial indicators designed to verify the effectiveness of the proposed methods in synthesizing CNTs. In recent years, various approaches have been suggested to synthesize nanostructures in which Scanning Electron Microscopy (SEM) has been used to demonstrate the quality of the CNTs. The SEM images of CNTs contain critical information with high resolution on a nanometer scale. The dispersion degree detection is one of the challenges in the quality of dispersion’s CNTs. If SEM images of CNTs have uniform and agglomerated distributions of particles, they are called sparse and dense images, respectively. Thus, the CNT images can be classified into two categories, including dense and sparse images. In the present study, a new algorithm has been developed to classify vertically aligned CNTs (VA-CNTs) based on texture analysis and Support Vector Machine (SVM). In this regard, these images were first transformed to time series, and their specifications were investigated through time series and Gray-Level Co-Occurrence Matrix (GLCM) analysis methods. Then, statistical specifications of these series were extracted. In this case, eight features have been extracted to classify VA-CNTs. The extracted specifications were used as inputs of an SVM algorithm to classify SEM images of CNTs into two groups, entailing dense and sparse. This algorithm has been tested on 80 VA-CNTs images with identical sizes. The findings showed a precision above 98 percent, which proved the validity of the proposed algorithm.