The method of finding an optimal aircraft trajectory to minimize the probability of detection by opponent radar detection systems is proposed. RCS data model, a three dimension (3D) aircraft and detection model, was established first. And then, the problem was described as an optimal control problem. The constraint functions were constructed by attitude control, trajectory constraints and boundary conditions. The cost function was formed to minimize both the probability of detection and the total flight time. After that, the continuous optimal control problem was parameterized to discrete nonlinear programming problem by Gauss pseudospectral method (GPM). Simulation results demonstrate the feasibility of the proposed method in low observable three dimension trajectory optimization problem with monostatic radar and bistatic radar system, and it is also shown that the probability of detection and the exposure time decrease significantly.