With the advancement of small-scale computing platforms, unmanned aerial vehicles (UAVs) are increasingly used in target tracking, environmental reconnaissance, battlefield operations, disaster rescue, and airspace monitoring. In these tasks, route planning algorithms are crucial. This paper focuses on developing efficient and high-quality online path planning algorithms for UAVs, addressing the limitations of existing methods. Traditional algorithms, such as Breadth-First Search (BFS), Depth-First Search (DFS), and A*, struggle with UAVs’ dynamic requirements. Similarly, sampling-based methods like Rapidly-exploring Random Tree (RRT) and its variants, while effective, often produce suboptimal paths. This paper makes three main contributions: First, it proposes using cubic polynomial spline curves to smooth UAV flight paths, minimizing turning energy and avoiding complex iterative computations. Second, it introduces an optimized sparse trajectory smoothing algorithm for waypoints, optimizing only the sparse key waypoints and ensuring safety constraints like obstacle avoidance and curvature are met. Third, it designs a specialized algorithm to handle non-convex and non-smooth optimization terms, improving convergence efficiency. Simulation experiments demonstrate the proposed algorithms’ effectiveness, showing they can generate smooth trajectories that meet dynamic constraints and optimize path quality under various conditions.