Abstract:     GPS signals are more likely to be blocked or reflected by tall buildings in urban canyons, resulting in poor positioning accuracy even positioning failure to smart terminal. Based on the analysis of the GPS location error in urban canyons, horizontal velocity of INS and Azimuth of magnetometer data are fused by particle filter to positioning based on smart terminal. Experimental results show that the average error of proposed algorithm is 3.19m,which is reduced by 76.9％ compared to 13.81 m of GPS, and also reduced 34.1％ and 33.8％ compared to 4.84 m of EKF and 4.82m of UKF fusion algorithm.