The invention discloses a spectral image-based meat quality visual non-contact detection method. By means of spatial distribution information and spectral characteristic information which are contained in a spectral image and reflect the characteristics of an object to be detected, multiple indexes of meat quality (such as water content, water activity, volatile basic nitrogen, meat color, microorganism counts, acid value and PH value) are respectively or comprehensively evaluated, and evaluation results are given in an image mode according to spatial distribution conditions of specific indexes in the object to be detected. The method can be used for quick and non-contact inspection of meat production, processing, storage, transportation and marketing links, inspection results are accurate and objective, and the expression mode is intuitive, so that a guarantee means for supervising the production and marketing quality safety of meat is provided.