With the development of science and technology, the application of flight manipulators has received extensive attention. The flying manipulator has broad application prospects, such as the maintenance of high-voltage towers, the storage and retrieval of elevated goods in warehouses, and the delivery of express and takeout goods. Before the actual application of the flight manipulator, due to the complex task requirements and nonlinear environment, it is necessary to continuously optimize the Trajectory Planning and Control (hereinafter referred to as TPC) of the flight manipulator. In order to improve the recognition and positioning accuracy of the robotic arm on the surface of the aircraft, and achieve precise control of the autonomous motion and operation of the robotic arm on the surface of the aircraft, this paper studies the TPC of the flight robotic arm based on deep learning, image moment and vector product methods, establishes a bearing return function model based on deep learning, and a Jacobian matrix of the flight robotic arm based on image moment and vector product methods. Through the experimental research on TPC of the flight manipulator, it was proved that the DL trajectory planning method could reduce the collision risk of the flight manipulator by 4.79% compared with the traditional trajectory planning method, and could improve the task completion speed of the flight manipulator by 4.66%. The application of DL to the TPC of the flight manipulator could improve the trajectory planning effect of the flight manipulator.