针对无人机在复杂公路巡检环境中安全风险大、路径规划效率低等问题，提出一种由底层局部感知驱动上层全局决策的IA*-IACO融合路径规划方法。首先，设计了融合几何安全约束、转弯惩罚与父节点启发的改进A*（Improved A*, IA*）算法，并引入带安全校验的Floyd平滑策略，实现局部路径的避障与平滑优化。其次，通过引入包含动态挥发因子与自适应权重机制的改进蚁群（Improved Ant Colony Optimization, IACO）优化算法，克服了非对称旅行商问题（Traveling Salesman Problem, TSP）的早熟收敛缺陷。最后，利用IA*生成的最短避障路径驱动IACO全局搜索最优遍历路径，从而规划出避障环境下的最优巡检路径。仿真实验结果表明，IA*生成的局部路径冗余拐点相较于传统A*减少了16.7%~25%，计算效率更高，轨迹更加平滑安全；IACO全局寻优各项性能指标均优于ACO，18个目标点的解集极差降幅达41.7%，并能智能生成高效巡检时序。该融合方法实现了从离散多目标到连续无碰撞巡检路径的一体化生成，有效提升了公路自动化巡检的安全性与效率。

To address the problems of high safety risks and low path planning efficiency of Unmanned Aerial Vehicles (UAVs) in complex highway inspection environments, an IA*-IACO fusion path planning method was proposed, in which lower-level local perception drives upper-level global decision-making. Firstly, an improved A* (IA*) algorithm was designed by integrating geometric safety constraints, turning penalties, and a parent-node heuristic, and a Floyd smoothing strategy with safety verification was introduced to realize local obstacle avoidance and path smoothing. Secondly, an improved ant colony optimization (IACO) algorithm with a dynamic evaporation factor and an adaptive weight mechanism was developed to overcome premature convergence in solving asymmetric traveling salesman problems (TSP). Finally, the shortest collision-free paths generated by IA* were used to drive IACO to search the globally optimal visiting sequence, so that optimal inspection routes under obstacle constraints were planned. Simulation results showed that IA* reduced the redundant inflection points in local paths by 16.7%~25% compared with the conventional A* algorithm,achieved higher computational efficiency, and produced smoother and safer trajectories. The IACO algorithm outperformed the conventional ACO algorithm in global optimization performance, and the solution-set range for the case with 18 target points was reduced by 41.7%, enabling intelligent generation of efficient inspection sequences. The proposed fusion method achieves integrated generation of continuous collision-free inspection paths from discrete multi-target sets and effectively improves the safety and efficiency of automated highway inspection.