As an indispensable material widely used in various important fields such as information, energy, industry, agriculture, transportation, and even aerospace and marine development, plastic particle raw materials have penetrated into all aspects of human food, clothing, housing and transportation. China is a large importer country of plastic raw materials. The existing test methods often cost too much time and barely achieve on-site testing. Therefore, the development of a discriminating model for waste plastic particles used in the field is of great significance for fast clearance and anti-smuggling in customs. Raman spectroscopy has the advantages of fast, non-destructive, small sample consumption, non-pre-treatment and strong adaptability, and has been widely used in rapid on site identification. Firstly, this research establishes a Raman spectroscopy reproducibility test method. On the basis of ensuring the real and effective Raman spectroscopy data, Raman spectroscopy combined with chemical discrimination method is applied to the identification of waste plastic materials. Two kinds of actual customs clearance plastic materials with similar composition were selected, each including 40 standard and wasted products. The Raman spectrum information of the samples was collected by NCS Smart 200 Raman spectrometer. A total of 640 samples of data of plastic raw materials were collected. The original Raman spectra of the two kinds of plastic materials were compared and analyzed. To further explore the composition changes of waste plastics, 1 001 cm^-1 was selected as the normalized reference peak position. The relative peak intensity changes of waste plastic raw materials and standard plastic raw materials were compared. The changes of relative peak intensity indicated that the waste plastic raw materials had chemical aging causes a change in its molecular structure and composition. Based on the principal component analysis (PCA), the original Raman spectroscopy and pre-processed Raman spectroscopy are subjected to dimensionality reduction. The first two principal component spaces of the original Raman spectroscopy have intertwined, which is difficult to completely separate. The spatial separation of the first two principal components of the pre-treatment Raman spectrum conducts well. Therefore, by performing background subtraction and smoothing pre-processing on the original Raman spectral data, the influence of the fluorescence background and noise on the discrimination can be reduced, and the accuracy of the discrimination can be improved. Half of the sample is divided into a calibration set for model building, half is divided into prediction sets for model verification, and partial least square discriminant analysis (PLS-DA) is used to build a waste plastic raw material identification model. The correctness rate is 100% for the modeling training set and 99.06% for the model verification set. The research shows that based on Raman spectroscopy technology, combined with test data pre-processing and partial least squares discriminant analysis method, it can effectively achieve the on-site, fast and accurate identification of plastic raw materials, and provide theoretical reference for the development of on-site testing equipment and methods.