Abstract: The traditional quaternion served as the description parameter of the nonlinear state model of the UAV， and the accuracy of the attitude estimation was presented. A square-root cubature Kalman filter algorithm based on quaternion was proposed. The algorithm took the attitude quaternion error and the gyro drift error as the state quantities， and measured the attitude quaternion of SINS/SLAM navigation. The square root cubature Kalman filter algorithm was used for pose estimation， which not only can solve the normalization issue of traditional quaternion， but also can reduce the state dimension and computational complexity of the square-root UKF algorithm of traditional quaternion and improve the numerical stability. Compared with the quaternion-based SRUKF and quaternion SRCDKF algorithm， the simulation results showed that the new algorithm estimated the error mean values of the roll angle， pitch angle and yaw angle， which were ${\mathrm{2}}^{″}$$,{\mathrm{3}}^{″}$，${\mathrm{1}}^{″}$， respectively. The errors were the smallest， and algorithm’s accuracy was about 30% higher than that of the quaternion SRUKF-SLAM algorithm， and it exhibited high filtering accuracy and numerical stability and the lowest computational cost.