Abstract: In the paper, statistic survey and yield estimation by remote sensing were combined by using two kind of ground survey data provided by Beijing Statistical Bureau, including sample village scale and plot scale to predict winter wheat yield. Sample villages, plots and plots stratification were used to estimate yield, and the correlations between predicted yield and statistic data were analyzed on different scales. The results showed that using ground survey data of sample village scale and plot scale both could get high-precision yield in Beijing area; on district-level, using ground survey data of plot scale could get more accurate yield than sample village scale; on the village level, predicting yield with ground survey data of plot scale could fit statistical yield better than sample village scale, and the yield model was more stable. Therefore, using ground survey data of plot scale to build entirety regression and stratified regression model are feasible and effective, both of them can get estimated yield of high-precision on small region level.