Aiming at the multi-threat coupling problem of radar network detection, missile interception and complex terrain avoidance faced by fighter aircraft in low-altitude penetration under modern integrated air defense system, a trajectory planning method based on multi-strategy improved dung beetle optimizer (MSDBO) is proposed. Firstly, a radar threat model based on Swerling I target detection probability is established, considering the dynamic influence of aircraft attitude angle on radar cross section. Secondly, combining the kinematic hard constraints such as maximum turning angle, climbing angle and minimum turning radius, a multi-objective fitness function is designed using dynamic normalization method. Thirdly, four improvement strategies including Tent-Logistic composite chaotic initialization, nonlinear adaptive weight, Cauchy-Gaussian hybrid mutation and refraction opposition-based learning are introduced to address the shortcomings of standard dung beetle optimizer. Finally, B-spline curve is used to smooth the planned trajectory. Simulation results show that compared with standard dung beetle optimizer, the trajectory length planned by MSDBO is reduced by 8.6% and the threat cost is reduced by 15.2%; compared with particle swarm optimization, the trajectory length is reduced by 18.5% and the threat cost is reduced by 37.1%; the cumulative detection probability throughout the flight is reduced by 31.2%, and the penetration survival probability is increased to 97.8%.