In view of the MEMS sensors' low precision and large errors, three-axis accelerometer, three-axis magnetometer and three-axis gyroscope are combined to measure attitude. And Kalman filter are used for multi-sensor data fusion. The gyroscope's attitude quaternion calculated by rotation vector becomes a predictive vector of Kalman filter. The accelerometer and magnetometer's attitude quaternion calculated by Gauss-Newton method becomes a observation vector of Kalman filter. Kalman propagation equation is established to obtain more accurate attitude quaternion to calculate the attitude angle and correspond the attitude angle in AHRS with the vehicle' attitude angle. Experimental results show that the changed attitude angle of the vehicle coincides with the different states of the vehicle state.