NIR calibration model updating based on slope/intercept analysis: A case study on firmness determination of pears harvested in different years

Because of the uncertain changes (such as sample properties, instrumental response function, the experiment environment, and so on), near infrared (NIR) prediction models for fruit quality determination using spectra acquired from fruit harvested in the last year cannot be directly applied to fruit harvested in this year or next. In order to avoid repetition of the whole calibration procedure and solve the model updating problem in NIR spectroscopy, a method based on slope/intercept analysis for calibration model modification is used in this study. As a result, calibration models using spectra acquired from pears harvested in the year of 2005 has been successfully updated for firmness determination of pears harvested in 2006 and 2007. The standardization procedure was performed on a data set of 1200 pears of three varieties (400 samples of each variety). A comparative study between the results with and without model updating was conducted. The results showed that the standard errors of prediction (SEP) obtained with modified models were much better than those obtained with non-modified models. It can be concluded that slope/intercept analysis is a useful method for NIR model updating.