The rainfall estimation technique based on the artificial neural network model is developed. Using the hourly GMS four channel′s data and rainfall records from hydrolgraphical station,the study of a heavy rain case which occurred in Huai River and Yangtze River basins (east of 108°E,from 24°to 36°N) shows that the average correlative coefficient between the quantitative precipitation estimation and the observation rainfall is 0.57,which is far higher than that in operation.