Partial Least Squares Regression (PLSR) Applied to NIR and HSI Spectral Data Modeling to Predict Chemical Properties of Fish Muscle

Partial least squares regression (PLSR) is a classical and widely used linear method for modeling of spectral data. Measurement of fish chemical properties has been playing an important role in providing superior quality products for human health and international trade. This review focuses on the PLSR applied to near-infrared (NIR) and hyperspectral imaging (HSI) spectral data for rapid and chemical-free modeling and predicting chemical properties of fish muscle, including moisture content, lipid content, protein content, pH, and freshness indicators, such as total volatile basic nitrogen, thiobarbituric acid reactive substances, and K index value. Furthermore, the commonly used spectral preprocessing methods and variable selection algorithms are mentioned and discussed for the enhancement of PLSR analysis. The limitations and future trends of NIR and HSI techniques with PLSR analysis are also presented. In a word, NIR and HSI technique in tandem with PLSR method have been developed to be suitable and trustworthy alternatives to the traditional chemical analytical methods such as Kjeldahl, Soxhlet, and chromatography methods for detecting chemical information of fish muscle in an objective, rapid, noninvasive, and chemical-free manner.