Vis/NIR spectroscopy was used to detect TVC and color of chilled pork rapidly and non-invasively. Vis/NIR scattering image in the range of 400nm to 1100nm were collected from 54 chilled pork samples. The scattering profiles were fitted accurately by four-parameter Gompertz function. The TVC and color L* prediction models were built with support vector machines (SVM) regression. The regression coefficient (Rv) of combination α, β, θ, δ and combination α, β, δ were 0.937 and 0.935, and the standard error of prediction (SEP) were 0.600\lg CFU/g and 0.702lg CFU/g, respectively. For color L*, the prediction model based on combination β and δ could give satisfactory results with Rv of 0.930 and SEP of 1.515. The results demonstrated that Vis/NIR spectroscopy combined with SVM was precise and potential for valid, rapid, non-invasive detection of chilled pork.