Procedure for calibration transfer between seasons for on-line NIR evaluation of SSC in onion breeding lines.

The present work attempts to progress in the definition of a procedure for maintain the prediction model performance of a NIR based onion grading line operating at the industry. This task is quite difficult to achieve as the selection is made in an industrial environment and the onions for classification are highly heterogeneous and evolves during seasons because the breeding process for soluble solids content (SSC). During operation at the industry (seasons 2004-2008), significant deviations from the initial behaviour and a reduction of prediction performance were observed. A non supervised analysis of the spectral database (more than1,000,000 spectra from onion bulbs) was made in order to evaluate the global performance during seasons (OrtizCanavate, Moya-Gonzalez et al. 2010). The application of the multivariate control statistics defined in this previous study to the calibration data sets will determine whether the performance will be improved by its use or additional calibration transfer procedures are needed.