当前随着无人机应用的日益广泛，在复杂环境下的自主飞行能力成为影响无人机的安全性和执行任务可靠性的重要因素，其中，无人机在未知场景下对复杂环境进行快速感知和进行全局规划是其自主飞行能力的重要组成部分。针对当前无人机在复杂场景下难以高效构建环境地图和实现远距离全局规划的难题，本文提出了一种基于概率更新的剪枝可视性地图构建方法和分层规划策略。首先，基于激光雷达或深度相机的原始数据，采用概率更新方式生成栅格地图，然后通过分层映射提取障碍物边界，结合碰撞检测从而构建出可视性地图，为加快路径搜索速度，本文提出了一种可视性地图剪枝策略以减小搜索空间。接着，构建基于搜索和优化的分层规划算法框架，外层规划采用基于探索度的改进A*算法，通过在代价函数中引入路径探索度，显著提升了无人机在复杂场景下的全局规划表现。内层规划则采用基于MINCO轨迹表示的轨迹优化方法，以生成满足无人机速度、加速度约束的平滑飞行轨迹。最后，通过实验仿真和实飞验证，结果显示，基于剪枝可视性地图的改进A*算法相较传统A*算法，飞行距离减少了12.92%，飞行时间缩短了16.43%，证明了本文方法在复杂场景下的有效性，未来将进一步应用到无人机集群中实现复杂环境下的协同规划。

With the increasing application of UAVs, autonomous flight capabilities in complex environments have become crucial factors affecting the safety and reliability of UAV operations. Among these capabilities, the ability for UAVs to rapidly perceive complex environments and implement global planning algorithms in unknown scenarios is a key component of autonomous flight. To address the challenges of efficiently constructing environmental maps and achieving long-distance planning in complex scenarios, this paper proposes a probability-update-based method for constructing pruned visibility maps and a hierarchical planning strategy.First, a grid map is generated from raw data collected by LiDAR or depth cameras using a probability update approach. Then, obstacle boundaries are extracted through layered mapping combined with collision detection to construct a visibility map. To reduce the search space for paths, we introduce a visibility map pruning strategy. Next, a hierarchical planning algorithm framework based on search and optimization is developed, where the outer layer planning employs an improved A* algorithm based on exploration degree. By incorporating path exploration degree into the cost function, the global planning performance of UAVs in complex scenarios is significantly enhanced. The inner layer planning employs a trajectory optimization method based on MINCO trajectory representation to generate smooth flight trajectories that satisfy UAV speed and acceleration constraints. Finally, through experimental simulations and real-flight validations, results show that the improved A* algorithm based on pruned visibility maps reduces flight distance by 12.92% and flight time by 16.43% compared to the traditional A* algorithm, demonstrating the effectiveness of the proposed method in complex scenarios.