High-rate GNSS positioning technology can measure the crustal displacement caused by earthquakes and can therefore be used for seismic monitoring and early warning. Compared with traditional seismographs, high- rate GNSS positioning technology has no range limitation and can acquire real-time centimeter-scale surface deformation. The application of GNSS to recover instantaneous dynamic deformation information during earthquakes has been mature. Some researchers have successfully captured seismic signal through PPP technology. The precision of the dynamic GNSS position can reach centimeter level currently, but the precision in the vertical direction is 2-3 times lower than horizontal direction, which makes it difficult to capture the weak seismic signal such as P wave. What’s more, the bandwidth of the GNSS receiver is only 20Hz, and accelerometer has wider bandwidth to monitor high frequency seismic signal. Thus data fusion based on accelerometers can help us obtain permanent displacement of the station and clearly show the vibration in all directions during the earthquake. We propose a seismic monitor system based on real-time multi-GNSS (GPS/GLONASS/GAILIEO/BEIDOU/QZSS) PPP (Precise Point positioning) and data fusion. Several stations are installed in the Yunnan province of China for earthquakes monitoring. This system is mainly made up of two parts. One is the service center which includes precise satellite clock estimation, precise satellite orbit process and precise satellite FCB (fractional cycle biases) products. And all products are broadcasted by generating SSR data streams. The other is client application which includes real-time PPP and real-time data fusion. The basic principle of data fusion is that we solve the position and speed of the station based on the common parameter of two observation equations and using the square root information filter. Datafusion combines the advantage of the GNSS and the accelerometer, and it can improve the ability of the instrument to detect the seismic signal. Furthermore, we realize the real-time smoothing for the data fusion process which can inflect the results more clearly while that way will delay for a period of smooth windows.