To better address the path planning problem of unmanned aerial vehicles (UAVs) in complex mountainous environments, a dynamic path planning method integrating an enhanced artificial lemming algorithm (EALA) with the dynamic window approach (DWA) is proposed. In this method, the waypoints generated by EALA through global path planning are used as local target points for DWA. Meanwhile, the velocity search space of DWA is extended from two dimensions to three dimensions, and a target-guidance function is introduced to enable real-time avoidance of both static and dynamic obstacles. In EALA, a chaotic mapping strategy combined with opposition-based learning is employed to initialize the lemming population, thereby improving the distribution uniformity of individuals in the solution space. An elite-assisted correction strategy is introduced to dynamically adjust the current optimal individual, enhancing the ability of the algorithm to escape from local optima. An adaptive nonlinear escape coefficient is adopted to improve the predator-avoidance behavior and balance global exploration and local exploitation. Inspired by differential evolution, a hybrid mutation strategy is designed to further improve the global search capability and convergence efficiency of the population. Finally, cubic spline interpolation is employed to smooth the planned path, thereby improving its continuity and flyability. Simulation results show that, compared with the benchmark algorithms, the optimal fitness, average fitness, and standard deviation obtained by EALA are reduced by an average of 3.28%, 7.01%, and 45.19%, respectively, in the simple scenario, and by 1.13%, 6.77%, and 48.95%, respectively, in the complex scenario, demonstrating a significant improvement in global path planning performance. Furthermore, the effectiveness and practicality of the proposed dynamic path planning method in complex dynamic environments are verified through dynamic obstacle-avoidance experiments.