Abstract: A simultaneous localization and mapping(SLAM) algorithm based on the combined EKF(extended Kalman filter) of Sage-Husa adaptive EKF and strong tracking EKF is presented to solve the decrease of filtering accuracy of standard EKF when the statistical characteristics of noise are not accurate and the model builded can not match with the actual one completely.Firstly,the dynamic model,feature model and sensor measurement model of AUV(autonomous underwater vehicle) are set up.Then,feature extraction is implemented through Hough transform,and SLAM of AUV is realized with the combined EKF eventually.Simulation with trial data shows that the described method reduces the influence of both the time-variance of statistical characteristics of noise and the inaccuracy of model,and enhances the accuracy and robustness of SLAM system.