In this paper, radar extrapolation and high-resolution numerical weather prediction (NWP) are blended to get a 6 h quantitative precipitation forecast for the Yangtze River Delta Region of China. The method used here is based on a mainstream blending framework, and modifications and calibrations are conducted done to both the extrapolation and NWP to improve the result. The traditional COTREC extrapolation is extended by using a mosaic system of 11 radars within the region and NWP data. Intensity and position calibration for the NWP are conducted based on Weibull function fitting and object recognition. A weighted blending of extrapolation and NWP is then carried out based on scale and forecast time. Finally, real-time Z-R relation conversion is performed. The results show that the blending method can extend forecast time of the extrapolation forecast, and calibrate the intensity and position bias of NWP, thus making a better result as a whole.