Detection of starch content in maize kernel based on Raman hyperspectral imaging technique

Area of Science:

• Agricultural Science
• Spectroscopy
• Food Quality Analysis

Background:

• Maize kernel quality is significantly influenced by starch content.
• Maize aging negatively impacts kernel quality, necessitating monitoring methods.
• Raman hyperspectral imaging offers potential for non-destructive analysis.

Purpose of the Study:

• To develop and validate a non-destructive method for detecting starch content in maize during aging.
• To investigate the relationship between Raman spectral changes and the maize aging process.
• To establish a predictive model for maize starch content using spectral data.

Main Methods:

• Raman hyperspectral imaging was employed to collect spectral data from maize kernels at different aging stages.
• Two-dimensional correlation spectroscopy (2D-COS) was used to analyze spectral variations and determine the order of Raman peak changes.
• Various preprocessing, variable selection, and modeling methods (including Extreme Learning Machine - ELM) were applied to build a starch content prediction model.

Main Results:

• The study identified characteristic Raman shifts related to maize aging.
• The Extreme Learning Machine (ELM) model, using 35 characteristic Raman shifts and normalization preprocessing, achieved optimal prediction performance (Rc=0.8827, Rp=0.8502).
• The established model demonstrated high accuracy in predicting maize starch content during aging.

Conclusions:

• Raman hyperspectral imaging, combined with 2D-COS and chemometrics, provides a feasible and powerful non-destructive method for determining maize starch content during aging.
• This approach offers a theoretical foundation for advancing non-destructive detection techniques in the cereal industry.
• The findings support the use of spectral imaging for quality assessment and monitoring of agricultural products.