Detection and classification of microplastics in simulated shoal environments using hyperspectral imaging technology

Microplastic pollution poses a global environmental threat, yet conventional detection methods suffer from inefficiency and single-point measurement limitations. Hyperspectral imaging (HSI) offers a promising alternative but faces challenges in complex shoal environments where soil-water matrices and variable microplastic concentrations complicate detection. But existing HSI studies focus on single-environment matrices, neglecting shoals and the impact of distribution states on spectral signatures. This study addresses these gaps by developing a non-destructive, efficient HSI-based method to detect and classify 0.5 mm microplastics in simulated shoal environments under two different concentration scenarios. For high-concentration samples, the study employed Otsu image segmentation to obtain binary masks, combined with SG smoothing and SNV preprocessing for spectral feature extraction. Furthermore, density and spectral correlation analysis was performed based on the physical characteristics of the microplastics. This study constructed two types of models based on spectral information and the integration of spatial and spectral information, respectively, with results demonstrating superior performance of the spectral-spatial joint model, achieving an accuracy of 99.52 %. In the detection of low-concentration samples, an improved image segmentation method based on the second principal component was employed. The experiment revealed that the MRCNN model achieved an accuracy of 93.85 % in the detection of low-concentration samples. This accuracy was further enhanced to 96.09 % upon the introduction of the channel attention mechanism. This study demonstrates that hyperspectral imaging technology can effectively classify microplastics in shoal environments, even under varying concentration conditions. By employing a spatial-spectral joint modeling approach in conjunction with a channel attention mechanism, accurate differentiation of three types of microplastic samples was achieved, thereby offering a novel technological means for monitoring microplastic pollution.