In order to solve the problem that the Beidou multipath error caused by environmental complexity cannot be effectively eliminated, an adaptive unscented Kalman filter (UKF) method based on a new error model is proposed. Firstly, the method uses the measurement state amplification method to solve the problem that the measurement noise is colored noise, and then uses the improved Sage-Husa adaptive UKF to dynamically estimate the system noise and measurement noise, so as to solve the situation that the error offset effect is not obviously caused by the undetermined statistical characteristics of noise. Comparing this method with the traditional UKF, the experimental results show that the improved Sage-Husa adaptive UKF algorithm under colored observation noise can reduce the multipath error bynearly 60%. This method has strong applicability for the case where the noise generated by the multipath error is unknown to the Beidou positioning.