0
Article ? AI-assigned paper type based on the abstract. Classification may not be perfect — flag errors using the feedback button. Tier 2 ? Original research — experimental, observational, or case-control study. Direct primary evidence. Detection Methods Marine & Wildlife Policy & Risk Remediation Sign in to save

Snapshot HoloSpec: dispersion-coded 4D feature learning for waterborne particle monitoring

Journal of Physics Photonics 2026
Jingyan Chen, Yaping Zhao, Yaping Zhao, Yuxing Li, Yanmin Zhu, Yanmin Zhu, Yuqing Cao, Edmund Y Lam

Summary

Researchers developed a compact optical instrument that simultaneously captures 3D shape and spectral (color) information from water-suspended particles in a single snapshot, enabling real-time identification of six particle types including microplastics without any sample preparation. The system achieved a 98.1% classification score and could run continuously in aquatic environments. Real-time, in-situ monitoring tools like this could transform how we track microplastic pollution in rivers, lakes, and oceans.

Abstract We present a compact, cost-effective, and highly accurate dispersion-based four-dimensional feature learning (3D spatial and 1D spectral) method for real-time monitoring of water-suspended microparticles, requiring neither sample preparation nor explicit reconstruction. By jointly leveraging digital holography and spectral imaging, the system uses a single prism to compress holograms with transmission spectral information onto an imaging sensor. Coupling an optical model with deep learning, our system recovers the 3D holographic information and class identity for six waterborne particle types, exemplified by microplastics, from a single snapshot. This enables all-in-one reconstruction of both geometry and identity. Both supervised and unsupervised pipelines outperform hologram-only baselines, achieving a macro F 1 of 98.10%. The F 1 score combines precision and recall, offering a comprehensive evaluation of classification performance. With the unsupervised feature-extraction route, our method improves F 1 by 15%, greatly reducing reliance on labeled data. The approach enables long-duration, real-time monitoring in dynamic aquatic environments without sample pretreatment or labeling, and it extends readily from transmission spectroscopy to other spectral modalities.

Share this paper