In order to overcome the disadvantages of poor repeatability and strong subjectivity brought by traditional artificial experience classification and realize the quantitative intelligent classification of turquoise, RGB model needs to be fitted. By fitting the RGB model of turquoise classification with linear function and polynomial function, it is proved that after fitting the model with polynomial function, the Adjusted R-Square (R2) reach more than 0.99. Using this method, the RGB model of optimal turquoise classification can be obtained, and improve the measurement accuracy of the color parameters of turquoise more than doubled.