In order to improve the precision of attitude calculation of strapdown inertial navigation system based on MEMS inertial sensor, an adaptive cubature Kalman filter (CKF) data fusion algorithm is proposed. The data fusion algorithm takes the attitude quaternion as the system state, uses the accelerometer information and the magnetometer information as the system observation, and performs real-time adaptive estimation on the system process noise matrix and the observed noise matrix to solve the problem of rapid descent of attitude calculation accuracy caused by the sudden change of system noise. The experimental results show that the adaptive CKF data fusion algorithm is significantly better on calculation accuracy than the gyroscope-based strapdown attitude calculation. The roll and pitch angle errors measured in the carrier dynamics are within 1° and the heading angle error is within 2°.