Using Hyperspectral Imaging in Food Analysis

In a recent study published in Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, researchers from the Beijing Technology and Business University used hyperspectral technology to analyze different types of wheat flour for quantification and quality (1).

wheat ears, grains and bowl of flour on a wooden table | Image Credit: © Nitr - stock.adobe.com

Wheat flour is an important ingredient in many food products around the world. Its quality and consistency are important because it ensures the quality of food products. The researchers combined several methods including hyperspectral technology, advanced data analysis methods, and machine learning to distinguish between five different types of wheat flour (1).

In this study, the researchers use hyperspectral imaging, which captures information across a broad spectrum of wavelengths. The research team established an analysis model based on the reflectance of wheat flour samples at wavelengths spanning from 968 nm to 2576 nm (1).

To enhance the accuracy of their model, the researchers applied several preprocessing techniques, including multivariate scattering correction (MSC), standard normalized variate (SNV), and Savitzky-Golay (S-G) convolution smoothing (1). These preprocessing steps effectively reduced the impact of noise in the original spectral data.

To streamline the model further, the study employed advanced feature extraction algorithms, including competing adaptive reweighted sampling (CARS), successive projection algorithm (SPA), uninformative variable elimination (UVE), and the UVE-CARS algorithm (1). These algorithms played a crucial role in identifying the most informative wavelengths for wheat flour grade discrimination.

The classification models that the researchers used in this study implemented two different approaches. The first approach was partial least squares discriminant analysis (PLS-DA), and the second approach is support vector machine (SVM) (1). The researchers also utilized the particle swarm optimization (PSO) algorithm to fine-tune the parameters of the SVM model, such as the penalty coefficient (c) and the regularization coefficient (g) (1).

The study's results underscored the superiority of non-linear discriminant models over their linear counterparts. Specifically, the MSC-UVE-CARS-PSO-SVM model exhibited the highest forecasting accuracy, achieving an impressive 100% accuracy rate in both the calibration and validation sets (1). These findings validate the effectiveness of hyperspectral reflectance technology coupled with SVM discriminant analysis in accurately determining wheat flour grades (1).

Furthermore, the study showcased the advantages of nonlinear models in spectral data analysis, highlighting their ability to effectively distinguish between different wheat flour grades. The PLS-DA method was particularly effective in discerning variations and identifying influential variables contributing to these differences (1).

The SVM algorithm's ability to excel in machine learning problems with limited sample sizes further solidified its role as a preferred method for processing spectral data and modeling analysis. This feature makes it particularly well-suited for the complex task of wheat flour grade classification (1).

The study's findings offer a compelling argument for the adoption of hyperspectral technology in wheat flour grade detection. As hyperspectral technology continues to evolve, it holds great promise for ensuring the quality and consistency of food products, ultimately benefiting both producers and consumers.

This research can be further improved by expanding the sample size and exploring additional wheat flour grades (1). The researchers also suggest delving deeper into the selection of effective wavelengths to broaden the model's scope of application (1).

Reference

(1) Zhang, S.; Yin, Y.; Liu, C.; Li, J.; Sun, X.; Wu, J. Discrimination of wheat flour grade based on PSO-SVM of hyperspectral technique. Spectrochimica Acta Part A: Mol. Biomol. Spectrosc. 2023, 302, 123050. DOI: 10.1016/j.saa.2023.123050

This article was written with the help of artificial intelligence and has been edited to ensure accuracy and clarity. You can read more about our policy for using AI here.