Optimized recognition of microplastic ATR-FTIR spectra with deep learning

Microplastic pollution poses significant environmental challenges due to its persistence and harmful effects on ecosystems and human health. This study presents an optimized method for identifying microplastics using attenuated total reflection Fourier transform infrared (ATR-FTIR) spectroscopy combined with a one-dimensional convolutional neural network (1DCNN). The proposed 1D-CNN model features a multi-layer structure designed for robust feature extraction from spectral data, achieving exceptional classification accuracy. Through systematic optimization, including improvements in feature extraction for mixed samples, the model achieved a test accuracy of 0.9992. This study demonstrates the effectiveness of deep learning-based methods in handling complex spectral data and provides a fast, accurate, and reliable solution for microplastic identification, with potential applications in environmental monitoring and pollution control.