An online detection system based on dynamic weight measurement and graphic processing skill was developed. The developed system weighted dynamically the mass of egg sample using strain gauge weight and calculated major-minor-axis of egg through image generation. Comparison of the egg long axis and egg mass by different methods proved that the online system had a strong robust. Sample of 117 eggs were used to check the system, with mass error of ±1 g, minor-axis error of -0.5~0.7 mm and major-axis error of ±0.6 mm. The correct classification rate of XL grade, L grade, M grade and S grade were 100%,100%,93.84% and 100%. The detection velocity was 4 eggs per second. Experiments show that the online detection system meets corresponding standards.