The Effect of Different Chemometric Methods on Determination of Moisture Content in Maize using Near Infrared Reflectance (NIR) Spectroscopy

Moisture content determination is essential analysis whether practice or laboratory studies. Moisture determination can be performed both gravimetrically and chemically. However, these methods have disadvantages in terms of labor and time consumed. Therefore, it needs to develop the alternative methods and Near Infrared Reflectance (NIR) Spectroscopy is an alternative technique that can be used for this purpose. Some special statistical methods are used for analyzing of the target substance in this device. On the other hand, it is known that spectral preprocessing have an effect on the prediction models developed. In this study, ground samples of 250 different maize genotypes were used as material. In this study, spectral data taken from these samples (between 1200 and 2400 nm) and the moisture values obtained from reference method were associated with different chemometric techniques. First derivative (FD) and standard normal variate (SNV) transformations were applied to spectral data as pretreatment methods. Successive Projections Algorithm–Multiple Linear Regression (SPA–MLR) was used as wavelength selection method technique. Result of the study showed that wavelength selection is an effective way for elimination of uninformative wavelengths. It was determined that high correlation (r>0.95) was observed between the results of reference analysis and predictions of NIR models. It was modeled by using SPA–MLR method which gives the most accurate prediction result among the generated models and transformed data.