Fluorescence Machine Vision-Based Rapid Quantitative Characterization of Microplastics

Abstract

Rapid detection of microplastics (MPs) is crucial for reducing their risks to the environment and human health. Recently, advanced optical technologies such as Fourier transform infrared spectroscopy and Raman spectroscopy, combined with artificial intelligence (AI) algorithms, have improved the efficiency of MP detection. However, traditional high-end instruments have limitations, including high cost, low efficiency, and susceptibility to signal interference, which result in insufficient data volume and quality, limiting the generalization and robustness of AI models. In this study, fluorescent microscopic images of coumarin 6 (C6)-stained MPs were collected, and machine vision models were trained for counting, size, and shape recognition. Results show that C6 provides stable and broad-spectrum staining, yielding high signal-to-noise image data. Also, a well-trained YOLO v11, which introduces a lightweight attention mechanism, further enhances the robustness and accuracy of the model. Meanwhile, a commercial AI-assisted detection platform named FluoPlastVision has been developed for use with fluorescence microscopes, enabling real-time counting, particle size analysis, shape classification, and data export of MPs. This method combines high-quality fluorescence images with high-performance deep learning models, enabling high-precision MP identification and showing potential as a standardized quantitative detection technology.