The measured patches were partitioned into different parts according to the lightness value. The training samples were selected according to the districts' percentage of total samples. BP neural network model was used to fit X-Rite 530's measured data with the SP60's. The simulation results and the subjective evaluation Z-score showed that BP neural network adaptive correction for the difference between X-Rite 530 and SP60 is better than 3D fitting algorithm. The purpose was to provide theoretical basis for realizing prediction of low-precision to high-precision color measurement instrument chroma values, and improve the accuracy of printing quality detection.