A Review of Data Analysis Techniques Used in Near-Infrared Spectroscopy

Abstract

Although the analysis of the structure of objects and the components that makeup them has been done for decades, it is one of today's research topics to do this analysis quickly and without damaging the sample. Near-infrared spectroscopy is used in many areas due to its non-contact measurement, fast analysis, and high accuracy features. Near-infrared spectroscopy is used in the classification or quality analysis of products, especially in the agriculture and food sector, due to the chemical bonds interacting in this region. The most critical part of achieving a successful result in near-infrared spectroscopy is pre-processing and analyzing the spectral data using the correct method. In this review, we perform a survey of recent studies that use near-infrared spectroscopy in food production and agriculture. Since there are many studies in this field in the literature, the survey is limited to cover works in the last five years. The review's main question is the pre-processing and data analysis methods used in these studies and the main features of these methods. Among the examined studies, the most frequently used pre-processing method was standard normal variate, and the most frequently used analysis method was partial least squares regression. In addition, the software tools and the spectrum range were also examined within the scope of the study.

Rapid and cost‐effective assessment of nutrients in pistachio (Pistacia vera L.) leaves through Fourier transform near‐infrared spectroscopy (FT‐NIRS)

“…Root mean square error of prediction (RMSEP) and coefficient of determination ( R 2 ) values were employed to evaluate the prediction performance of the models (Esbensen, 2009). The regression point displacement (RPD), or, in another name, the ratio of performance to deviation (RPD), was also utilized in the evaluation of the results (Çataltas & Tutuncu, 2021; Kljusurić et al., 2016). The program calculates the RMSEP values using the following equation (Esbensen, 2009): $$$$\begin{equation}RMSEP\ = \sqrt {\frac{{\mathop \sum \nolimits_{i = 1}^N {{\left( {{y}_i - {y}_{i,ref}} \right)}}^2}}{n}} ,\end{equation}$$$$where y i is the predicted value, y i,ref is the measured value, and n is the number of samples.…”

Integrated LSPR Biosensing Signal Processing Strategy and Visualization Implementation

“…The former mainly involves setting the acquisition parameters of the sensor, completing the spectral acquisition, and receiving the transmitted data, while the latter aims to achieve comprehensive processing of the spectral data. This includes data processing and analysis according to user needs, performing operations such as smoothing [23], peak finding, baseline correction, spectral analysis, etc., to ensure the accuracy and reliability of the results. Such a comprehensive spectrometer software system will help to improve the efficiency and operability of spectro-scopic experiments and meet the diverse needs of users in different fields.…”

Detection of protein, starch, oil, and moisture content of corn kernels using one-dimensional convolutional autoencoder and near-infrared spectroscopy

“…Therefore, the NIR spectrum needs to be preprocessed as it contains sensor, light, or converter-induced distortions ( Mishra et al, 2020 ). Although dozens of methods are used in the literature, the most commonly used preprocessing methods can be said as mean scatter correction (MSC), standard normal variate (SNV), Savitzky-Golay filter (SG), first and second derivative and mean centering (MC) ( Çataltaş & Tütüncü, 2021 ; Rinnan, Berg & Engelsen, 2009 ). However, the preprocessing method is determined by trial and error since the usefulness of the preprocessing method varies according to each spectrum.…”