Abstract:To investigate the potential of near infrared reflectance spectroscopy (NIRS) and chemometrics methods to quickly predict the freshness of Megalobrama amblycephala,NIR spectra of 150 samples from different seasons,different origins,different specifications and different storage durations were recorded in the range of 1 0001 799 nm. The spectra were preprocessed and then calculated using CARS for wavelength variable selection for establishing quantitative models to predict pH,TVBN,TBA and K values with partial least squares regression (PLSR) method. The results showed that correlation coefficients of the models were 0.961,0.881,0.955 and 0.946,and root mean square errors of cross validation (RMSECV) were 0.049,1.659,0.047 and 2.558,respectively. The models have good prediction ability,which will provide an effective method for predicting the freshness of freshwater fish quickly and nondestructively.