Du, M. ; Guo, L. B. ; Hao, Z. Q. ; Li, J. M. ; Li, X. Y. ; Liu, J. G. ; Lu, Y. F. ; Shen, M. ; Yi, R. X. ; Zeng, X. Y.
Laser-induced breakdown spectroscopy; Characteristic spectral lines of non-metallic elements; Support vector machine; Plastics identification;ARTIFICIAL NEURAL-NETWORKS; INDUCED PLASMA; EMISSION; POLYMERS
Laser-induced breakdown spectroscopy (LIPS) combined with support vector machine (SVM) was adopted to identify 20 kinds of different colored industrial plastics from different manufacturers in open air. The experimental parameters of spectral acquisition were optimized firstly. 100 spectra recorded under optimum conditions were randomly and equally divided into training set and test set. 6 non-metallic characteristic spectral lines were used to avoid the interference with metallic lines. And the training time of SVM model was reduced. The results show that 996 of 1000 test spectra were identified correctly and the average classification accuracy is reached to 99.6%. The classification efficiency is improved with 6 non-metallic characteristic spectral lines. The research demonstrates that, when fewer of major non-metallic characteristic spectral lines are used, laser-induced breakdown spectroscopy technique with support vector machine can identify more kinds of plastics with high accuracy and efficiency.