0
Article ? AI-assigned paper type based on the abstract. Classification may not be perfect — flag errors using the feedback button. Sign in to save

Advanced Microplastic Identification in Marine Environments via Hybrid Deep Learning

2026
D. Deepa, Nivedha S S, Nivetha Angel R

Summary

Researchers propose a hybrid deep learning architecture combining 3D convolutional neural networks and Vision Transformers applied to hyperspectral imagery to detect and classify microplastics in turbid marine environments, capturing both local spectral signatures and global contextual patterns that single-model approaches miss.

Body Systems
Study Type Environmental

Microplastics, which are defined as plastic particles less than 5 mm in diameter, can be considered a serious issue to the marine ecosystem because it is persistent, toxic and bioaccumulative in the food web. Nevertheless, the majority of the conventional detection algorithms, such as manual microscopy and simple-enhancement models suffer a multitude of shortcomings insofar as they lack great accuracy, are lacklustre in scaling, and inefficient in ocean environments, in particular turbid and noisy ones. The proposed project is going to introduce a new design of hyperspectral-based microplastic detection through a combination of 3D Convolutional Neural Networks and Vision Transformers as a system that dynamically works to raise the alarm. The suggested system records the spectral information of hundreds of wavelengths at high spectral resolution, preprocesses the data on noise reduction and spectral clarity, partitions and clustering spectral patterns, and implements a hybrid deep learning network to classify microplastics precisely. This method is new, as it is the combination of spatial-spectral features with contextual learning on a global scale, which allows accurately detecting the microplastics, including identifying the type of plastic, in complicated water conditions.

Share this paper