This article tackles the finite-time global trajectory tracking control problem of the autonomous underwater vehicle (AUV) in presence of input saturation constraints, actuator faults, unknown dynamics, and external disturbances. First, we describe the orientation of the AUV by rotation matrix instead of classical Euler angle or unit quaternion such that the AUV's dynamics could be globally formulated without singularity and unwinding phenomenon. After that, a smooth dead zone-based model is introduced here to linearize the actuator model, leaving that the adaptive laws could be suitable for the solution of input saturation and actuator faults. Considering that the difficulty of model dynamic acquirement, together with the complicity of rotation-matrix-based representation, would trouble deployment of the controller. The minimum learning parameter technology is thereby utilized to approximated the dynamic nonlinearity of the AUV. On the basis of these, a rotation-matrix-based sliding mode control scheme is technically proposed. It is proved that the tracking errors can be stabilized to a small neighborhood of origin within a settling time. Finally, several set numerical experiments are conducted to assess the effectiveness and show the advantages of the proposed control scheme.