Identification of unfertilized duck eggs before hatching using visible/near infrared transmittance spectroscopy

Identification of unfertilized duck eggs before hatching was hardly achieved in the poultry farm by the traditional method. Visible/near infrared (VIS/NIR) transmittance spectroscopy was used as a rapid and non-destructive technique to distinguish the unfertilized duck eggs before hatching. Transmittance spectral data of duck eggs laid by the same flock of ducks was obtained. Prediction models were developed and used to link the non-destructive assessment methods and qualitative observation values of spectral with the spectral data. Established models were validated by the internal leave-one-out cross-validation (LOOCV) and external validation. The prediction model established by standard normal variate (SNV) pre-processing method and naive Bayes (NB) modeling method got the best performance. The overall prediction accuracy of the validation set was 94.54% and the prediction accuracy of internal validation was 95.06%. The prediction accuracy of 667 duck eggs by means of NB modeling method by SNV pre-processing was 94.77%. Results showed that the VIS/NIR transmittance spectroscopy was an effective tool for detecting the unfertilized duck eggs. This provided technical support for identification of the unfertilized duck eggs.