This paper presents a strong tracking unscented Kalman filter (STUKF) algorithm as an in-flight method to correct the magnetic deviation for a miniature attitude and heading reference system (Mini-AHRS) used in the navigation system of the small autonomous underwater vehicle (SAUV). In our approach, magnetic deviation caused by hard iron effect, soft iron effect and the initial heading error of Mini-AHRS are considered as the states of STUKF algorithm. First the system and measurement model is presented for magnetic deviation calibration and initial heading error estimation. Then STUKF algorithm is derived to overcome the wave and wind disturbance, when SAUV running on the sea surface to calibrate the magnetic deviation. Finally the proposed algorithm is verified by numerical simulation using Matlab. Simulation result demonstrates that STUKF algorithm can effectively reduce the disturbance caused by the sea wave and wind. When the AUV running on comparatively rough sea surface, our magnetic deviation correction algorithm convergent fast and estimate the errors correctly.