Laser-induced breakdown spectroscopy (LIBS) technology and genetic algorithm optimization based error back propagation (GA-BP) neural network are used to classify and recognize 9 common plastics in this paper. Plasma spectra are generated by laser-induced breakdown of the plastic surfaces, and 100 sets of spectral data are collected for each plastic with a spectrometer. The national institute of standards and technology (NIST) atomic spectrum database is used as a reference to accurately calibrate the main element characteristic lines. In the experiment, 15 characteristic spectral lines are selected for analysis, and the dimension of spectral data is reduced by principle component analysis (PAC) method, and GA-BP neural network model is established. Experimental results show that the GA-BP neural network recognition efficiency is greatly improved after dimensionality reduction by PCA method, and the average recognition accuracy is 99.72%, which can identify a variety of plastics quickly and accurately.