An image features assisted line selection method in laser-induced breakdown spectroscopy

Abstract

Analytical lines play a crucial role in laser-induced breakdown spectroscopy (LIBS) technology. To improve the classification performance of LIBS, an image features assisted line selection (IFALS) method which based on spectral morphology and the characteristics of Harris corners was proposed. With this method, a classification experiment for 24 metamorphic rock samples was conducted with linear discriminant analysis (LDA) algorithm. The result showed that the classification accuracy was increased from 94.38% of the conventional classification model MLS-LDA (Manual line selection-linear discriminant analysis) to 98.54% of IFALS-LDA. Furthermore, the time required for the whole classification process was decreased from 2768.38 s of MLS-LDA to 4.36 s of the proposed method, thus the classification efficiency was greatly improved. In addition, compared with the existing automatic line selection method, the convergence rate of IFALS-LDA is significantly faster than that of ASPI (Automatic spectral peaks identification)-LDA. This study demonstrates that LIBS assisted with the image features in machine vision can promote the analytical performance of LIBS technology.