Addressing the issue of the traditional A* algorithm encountering “dead zones” in dense and complex urban environment due to its failure to timely identify local optimal paths, resulting in flight hazards, an optimal backtracking model within the local field of view was established by designing a route cost function from the current waypoint to the UAV's visual range. A comprehensive route planning method, combining local backtracking with breadth-first principles, was proposed. Simulation comparative experiments have demonstrated that this method can plan safe and feasible flight paths for UAVs in dense and complex urban environment, offering advantages over other algorithms. By applying Bézier curves to smooth the planned routes, the safety of UAV flights in three-dimensional complex environment is enhanced.