Rapid Assessment of Chilled Chicken Spoilage Based on Hyperspectral Imaging Technology and AdaBoost-RT

The current method for detection of Pseudomonas spp. count (PC) in chilled chicken involves cumbersome operations and a large amount of destruction, resulting in lower efficiency. Our study presents a fast non-destructive detection approach of PC in chilled chicken using hyperspectral imaging (HSI) system (329~1113 nm). The hyperspectral information of chilled chicken samples was collected, combined with the derivative method for spectral preprocessing, and then, a continuous projection algorithm was used to extract the characteristic wavelength. The partial least squares and AdaBoost-RT prediction regression models based on the characteristic and full-spectrum data were established respectively to study the quantitative relationship between the spectral information and chicken PC. The results showed that the second derivative pretreatment effect is relatively better. The best quantitative prediction model of PC in chicken samples was the AdaBoost-RT model based on the characteristic band ( R P = 0.98, RMSE P = 0.27 (lg (CFU/g))). The overall study indicated that HSI technology (329~1113 nm) combined with the AdaBoost-RT algorithm could be used to detect the PC of chilled chicken in a rapid and non-destructive way.