To address the issues of low attitude estimation accuracy and magnetic interference affecting low-cost inertial navigation devices, this paper proposes a multi-sensor fusion algorithm based on Mahony Complementary Filter and Adaptive Cubature Kalman Filter (MACKF). First, the Mahony filter is used to fuse magnetometer and accelerometer data to correct gyroscope output in real-time and actively compensate for magnetically disturbed data through a keyframe mechanism. The corrected attitude quaternion is then used in the Cubature Kalman Filter, with adaptive adjustment of the measurement noise covariance matrix to reduce magnetic interference. Vehicle-mounted experimental results show that this algorithm significantly improves attitude estimation accuracy, with roll, pitch, and yaw angle precision improved by 45.3%, 50.2%, and 32.8%, respectively, compared to traditional methods. Therefore, the proposed algorithm demonstrates excellent performance in mitigating gyroscope drift and resisting magnetic interference.