Model Establishment for Fast Predicting Centrifugal Loss Rate of Fresh Chicken Based on Hyperspectral Information

The present study aimed to establish a rapid model for predicting centrifugal loss rate of fresh chicken by mining hyperspectral information in the wavelength range of 900~1700 nm. Hyperspectral images of chicken samples were acquired and the spectral information within the region of interest of images were extracted. Partial least squares(PLS)models were established based on the full range wavelengths pretreated by baseline correction(BC),Gaussian filter smoothing(GFS),multiplicative scatter correction(MSC),moving average smoothing(MAS)and median filtering smoothing(MFS),respectively. Through regression coefficient(RC),successive projections algorithm(SPA)and stepwise algorithms,optimal wavelengths were respectively selected to optimize the full wavelength PLS model. The results showed that the PLS model based on the 16 optimal wavelengths(900.6,915.4,1024.0,1089.8,1111.2,1155.6,1165.5,1288.9,1305.4,1433.9,1442.1,1486.7,1493.3,1541.1,1690.1 and 1693.4 nm)selected from raw spectra by stepwise method had better performance,with r_C of 0.94,root mean square error of calibration(RMSEC)of 1.43%,r_P of 0.94 and root mean square error of prediction(RMSEP)of 1.60%. The experiment concluded that PLS model based on hyperspectral information could be used for the rapid prediction of centrifugal loss rate in raw fresh chicken.