To address the issue of the accumulated heading angle drift of low-cost MEMS inertial sensors and the inconvenience of using traditional magnetic field calibration methods, a nine-axis inertial fusion algorithm based on dynamic magnetic field calibration was proposed. First, the relationship between the gyroscope and magnetometer was established by a rotation matrix, and the magnetometer was dynamically calibrated by the extended Kalman filter. Second, trust functions were defined considering the free acceleration and the sudden change in the magnetic field environment of the complementary filter, which improved the PI controller. Finally, the attitude angle of the inertial sensor with high stability was obtained. The experimental results show that, the dynamic calibration of the magnetic field by a gyroscope can effectively address the shortcomings of the traditional ellipsoid fitting algorithm, which requires high calibration data. This method can essentially be used in real time, and users can complete magnetic field calibration without a specific “8” winding action. The complementary filtering algorithm is improved to eliminate the influence of free acceleration on attitude angle calculation. When the sampling rate of the sensor is 100 Hz, the running time is approximately 11 min, the number of rotations is 117, and the drift of the heading angle is 0.42°, which is 14.9° less than that of the commercial IMU module, indicating an improvement on the order of a magnitude. This algorithm has considerable advantages in controlling and reducing the heading drift and meets the requirements of convenient calibration and various applicable scenarios. It has a broad application prospects in the field of low-cost MEMS inertial navigation.