Portable Near Infrared Spectroscopy Study for Turmeric Adulteration Detection

Area of Science:

• Analytical Chemistry
• Spectroscopy
• Chemometrics

Background:

• Turmeric powder adulteration with cheaper materials is common for economic gain.
• Existing on-site detection methods are limited, posing a challenge for quality control.

Purpose of the Study:

• Develop a rapid, non-destructive method for detecting corn flour adulteration in turmeric powder.
• Utilize portable near-infrared (NIR) spectroscopy, chemometrics, and deep learning for on-site analysis.

Main Methods:

• Prepared 330 samples with 0-50% corn flour adulteration and collected NIR spectra (900-1700 nm).
• Compared various pre-processing methods and feature selection strategies (PCA, CARS, UVE).
• Developed a dual-attention CNN-LSTM network for simultaneous classification and content regression of adulteration.

Main Results:

• Optimized PLS model with D2 pre-processing and UVE feature selection.
• CNN-LSTM dual-attention model achieved 98.9% classification accuracy and excellent regression performance (R2 = 0.9927).
• Attention maps highlighted spectral features related to curcumin attenuation and starch enhancement.

Conclusions:

• Portable NIR spectroscopy combined with a multi-task deep learning model provides an accurate and efficient strategy.
• This approach enables rapid on-site screening of turmeric adulteration under controlled conditions.