Path planning for automated parking remains challenged by the demand to balance general parking scenarios and computational efficiency. This paper proposes a two-stage rapid-exploring random tree (RRT) algorithm to improve the computational efficiency. At first the proposed algorithm performs space exploration and establishes prior knowledge, represented as waypoints, using cheap computation. Secondly a waypoint-guided RRT algorithm, with a sampling scheme biased by the waypoints, constructs a kinematic tree connecting the initial and goal configurations. Numerical study demonstrates that the two-stage algorithm achieves at least 2X faster than the baseline one-stage algorithm.