Artificial intelligence-supported spatial scanning for enhanced real-time spectral analysis of heterogeneous media

Near-infrared (NIR) and mid-infrared (MIR) spectroscopy are increasingly used for rapid, non-destructive material analysis in agriculture and food industries.
A significant challenge in spectral analysis is dealing with sample heterogeneity, especially in granular samples like cereals, which create “spectrospatial noise.”
Researchers from Si-Ware Systems and other universities developed an enhanced method using AI to analyze heterogeneous samples effectively with MEMS-based FT-NIR sensors.
They employed a spatial scanning technique to integrate measurements over a larger area, reducing spatial noise while maintaining a high Signal-to-Noise Ratio (SNR).
The research revealed a trade-off between SNR and spot size: larger optical spot sizes spread out spatial noise but increase electrical noise.
Using spatial scanning, the team demonstrated substantial improvements in spectral repeatability and chemometric model accuracy by averaging many measurements over a scanning path.
The spatial scanning approach improved spectral repeatability and reduced protein and moisture RMSE significantly compared to stationary measurement modes.
This method provides a viable route for developing portable, high-precision sensors for precision agriculture and smart industry applications.

You may have missed