Detection of the total viable counts in chicken based on visible/near-infrared spectroscopy

The viable counts in chicken have significant effects on food safety. Exceeding standard index can have negative influence to the public. Visible-near infrared spectra have had rapid development in food safety recently. The objective of this study was to detect the total viable counts in chicken breast fillets.36 chicken breast fillets used in the study were stored in a refrigerator at 4°C for 9 days. Each day four samples were taken and Vis/NIR spectra were collected from each sample before detecting their total viable counts by standard method. The original data was processed in four main steps: Savitzky-Golay smoothing method, standard normalized variate (SNV), model calibrating and model validating. Prediction model was established using partial least squares regression (PLSR) method. Several statistical indicators such as root mean squared errors and coefficients were calculated for determination of calibration and validation accuracy respectively. As a result, the Rc, SEC, Rv and SEV, of the best model were obtained to be 0.8854, 0.7455, 0.9070 and 0.6045 respectively, which demonstrate that visible-near infrared spectra is a potential technique to detect the total viable counts(TVC) in chicken and the best wavelengths for the establishment of the calibration model are near 449nm.