Rock color reflects the geological environment and mineral and element composition of its formation. And it is one of the important basis and indicators for vertical stratigraphic correlation and lateral environmental evolution. Rock color mainly relies on visual identification and subjective description or uses color charts for comparative interpretation. These methods are greatly affected by individual differences and the environment, lack quantitative calculation methods, and cannot meet the needs of accurate color identification in batches. Therefore, it is of great significance for the research and application of geological work to quickly realize the objective identification and numerical quantification of color. This research is based on the principle of colorimetry, using spectral analysis technology combined with the quantitative recognition software of rock color compiled by Python computer language. It can perform numerical quantification and automatic batch conversion of rock color. The method improves color judgment accuracy and recognition efficiency. The comparison of 《Munsell Rock Book》shows that the calculation results of the CIE RGB color system are highly consistent with the color card. In the calculation results of the Munsell system, the consistency of hue value (<3 NBS units) reached 86.7%, the consistency of lightness value and purity value reached 92.2% and 82.2%, and the correlation was 98.83% and 87.50%. Their errors all belong to the small chromatic aberration range. Compared with the calculation results of the Munsell system, the CIE RGB calculation results of the 31 rock samples are more consistent and accurate with the color of the samples. The reasons for the errors are complex and diverse, related to the conversion error between color systems , human subjective comparison and interpretation, and closely related to the particularity of rock samples and the environment and other factors. This study provides a feasible method for the quantitative characterization of rock color. This research has good application value.