The terminal flight of the gliding missile involves high complexity, strong uncertainty, and many constraints. It is difficult to model and solve the corresponding trajectory planning and guidance problems. To increase the maneuvering range of the gliding missile and reduce the difficulty of trajectory planning, a multi-constrained trajectory intelligent planning method was proposed. This method included waypoint design and maneuverability prediction using the Continuous Deep Belief Network (CDBN). The CDBN was used to predict the maneuvering ability online, and the feasibility of the waypoints state was determined rapidly. With the intelligent design of the waypoints, optimized allocation of energy was realized, which increases the flight envelope. To realize oscillatory maneuvering, the Line of Sight (LOS) relative to the target was designed as a trigonometric function, which was tracked by designing the optimal maneuvering guidance law. Finally, the desired velocity constraint was satisfied by adjusting the frequency of the LOS angle. The simulation results show that the CDBN has higher maneuverability prediction accuracy than the BP network. The proposed method can realize oscillatory maneuvering and achieve a large increase in the flight envelope while satisfying the terminal velocity constraint. Online trajectory planning based on the CDBN can be completed in half a second, which satisfies the rapidity requirements for engineering applications.