Effective observation is a key factor in improving cognition and prediction of sea fog. Satellite data is the most feasible observational data source, but requires high quality and fine detection technology. In order to improve the sea fog detection accuracy under current conditions of FY-2 data resources, a new method of sea fog detection for Yellow Sea and Bohai Sea is put forward in this paper. The new method is based on dynamic threshold technics, and includes temperature, texture, noise detection steps. The test results of day-time sea fog detection experiments show that though it still can be improved during autumn and winter, the detection accuracy has been close to the international advanced level in spring, which is a high-occurrence season of sea fog. To further improve the new method, more sea fog events need to be collected to optimize the threshold acquisition scheme.