A deep one-dimensional convolutional neural network for microplastics classification using Raman spectroscopy

Microplastics have emerged as major global environmental contaminations. Finding accurate and effective identification methods for microplastics is of great significance. In this paper, we propose a method to identify microplastics using Raman spectroscopy combined with a one-dimensional neural network(1D-CNN). Raman spectra of ten microplastics were collected and pre-processed. Both raw and pre-processed data were augmented. The 1D-CNN model reached the highest accuracy of 96.4 % based on raw data and 96.2 % based on pre-processed data. Then the K-Nearest Neighbors (KNN), Support Vector Machine (SVM), Multi-layer Perceptron (MLP) and AlexNet classifiers were trained and tested with the same dataset for contrast. Experimental results showed that our method has the best classification performance with an average accuracy of 95.8 % based on raw data and 95.5 % based on pre-processed data. This study proves that Raman spectroscopy combined with 1D-CNN can classify microplastics effectively and accurately whether the spectra data are pre-processed or not, which can further shorten the recognition time and provide a reference for microplastics detection in the future.