In view of the problem that the attitude estimation of UAV navigation system in dynamic environment is easily interfered by sensor noise and motion acceleration, a new attitude estimation algorithm of UAV considering motion acceleration interference was proposed. First, an acceleration estimation model was established. The acceleration error model based on Kalman filter and the velocity information provided by the external sensor were combined to accurately estimate the motion acceleration. The estimated motion acceleration was used to correct the original value of accelerometer, so as to reduce the interference of motion acceleration in the attitude estimation of navigation system in dynamic environment. Then, an attitude estimation model based on complementary filter was built. The gyroscope correction value was obtained by using magnetometer information and modified acceleration information, and the original gyroscope value was corrected. The complementary filter was designed to filter the high-frequency noise from accelerometer and magnetometer and the low-frequency noise from gyroscope, so as to avoid the interference of sensor noise signal in attitude estimation. Finally, the sensor information collected during flight test was used to simulate and verify the proposed algorithm. Experimental results show that the algorithm could accurately estimate the motion acceleration, reduce the interference of sensor noise and motion acceleration in attitude estimation, and effectively improve the accuracy and anti-interference ability of UAV navigation system in dynamic environment.