Abstract: In order to get the relationship between quality and external color characteristic of delinted cottonseeds, firstly, the images of cottonseed samples were acquired by using digital camera and color features of cottonseeds were extracted based on image processing technology. Then, the germinating experiment were performed strictly according to the national standard (GB/T3543.4-1995). Finally, a prediction model between delinted cottonseeds germination and color characteristic was developed based on the linear regression and the stepwise regression analysis methods, and the precision of model also were estimated. The experimental results showed that stepwise regression models achieved quality predictions with determination coefficients (R2)of 0.6949 for germination trend and 0.7148 for germination rate based on four color characteristic parameters (R/(R+G+B), G/(R+G+B), (R–G–B)/(R+G) and R/G), and the two determination coefficients passed 0.01 test level.These results provide a theoretic basis for rapid and effective sorting of delinted cottonseeds under the RGB color model.