The meandering channel deposit of the upper member of Neogene Guantao Formation in Shengli Chengdao extra-shallow sea oilfield is characterized by rapid change in sedimentary facies. In addition, affected by surface tides and sea water reverberation, the double sensor seismic data processed by conventional methods has low signal-to-noise ratio and low resolution, and thus cannot meet the needs of seismic description and oil-bearing fluid identification of thin reservoirs less than 10 meters thick in this area. The two-step high resolution frequency bandwidth expanding processing technology was used to improve the signal-to-noise ratio and resolution of the seismic data, as a result, the dominant frequency of the seismic data was enhanced from 30 Hz to 50 Hz, and the sand body thickness resolution was enhanced from 10 m to 6 m. On the basis of fine layer control by seismic data, three types of seismic facies models, floodplain, natural levee and point bar, were defined, and the intelligent horizon-facies controlled recognition technology was worked out, which had a prediction error of reservoir thickness of less than 1.5 m. Clearly, the description accuracy of meandering channel sand bodies has been improved. The probability semi-quantitative oiliness identification method of fluid by prestack multi-parameters has been worked out by integrating Poisson's ratio, fluid factor, product of Lame parameter and density, and other prestack elastic parameters, and the method has a coincidence rate of fluid identification of more than 90%, providing solid technical support for the exploration and development of thin reservoirs in Shengli Chengdao extra-shallow sea oilfield, which is expected to provide reference for the exploration and development of similar oilfields in China.