In this study, physicochem. indicators in chilled beef were first determined, including colony counts for Pseudomonas, Lactobacillus, Enterobacteria, Aeromonas followed by determination of total volatile base nitrogen (TVB-N) and sensory evaluation.Near-IR (NIR) hyperspectral imaging (HSI) combined with entropy weight method (EWM) was developed for the first time to evaluate the degree of spoilage of chilled beef.Comprehensive index of Dominant psychrophilic bacteria consisting of Pseudomonas, Lactobacillus, Enterobacter, Aeromonas was constructed by the EWM.Four methods were used to select the optimal feature wavelength.Compared to all models, the Condor search algorithm optimizes echo state networks (BES-ESN) was firstly used in modeling to predict the content of comprehensive index.The results showed that for the psychrophiles of synthetical index bacteria (POSIB), the prediction effect based on Normalize-LSSVM model was the most ideal (R2c = 0.92, R2p = 0.71, RMSEC = 0.48 lg CFU/g, and RMSEP = 0.82 lg CFU/g).For the total number of colony counts, the prediction effect based on CARS-LSSVM model reached R2c = 0.92, R2p = 0.82, RMSEC = 0.56 lg CFU/g, RMSEP = 0.87 lg CFU/g.Effective prediction of each colony in chilled beef was achieved.