In order to solve the problems of low precision and large error of MEMS sensor components and high noise and low accuracy of heading angle data measured by magnetic sensor, a method of attitude fusion based on complementary particle filter is proposed. Firstly, the attitude angle of gyroscope is corrected by using complementary filtering algorithm combined with accelerometer and magnetometer, and then particle filtering is performed on gyroscope data by using four element methods. The simulation results show that the algorithm can quickly calculate the attitude angle and improve the calculation accuracy.