For the research of marine remote sensing calibration of China, a mathematical model of bulk-skin sea surface temperature is proposed to improve the measurement precision of skin sea surface temperature to 0.3℃. Skin sea surface temperature is measured by the self-designed automatic multi-spectral infrared radiometer system installed on the PY-301 oil platform in the northern part of the South China Sea, the measurement accuracy of the radiometer is better than±0.5 ℃ through standard blackbody calibration, pupil correction and skylight correction. The effect of wind speed on the bulk-skin sea surface temperature is analyzed based on the model, the skin sea surface temperature is corrected by the proposed model and Donlon model, the biases of corrected results are －0.007 6±0.297 1℃ and 0.044 6±0.324 8℃, respectively. Results prove that the proposed model is effective and can improve the measurement accuracy of skin sea surface temperature.