无人机多视点自动巡检人造目标的方法

摘要:

针对建筑物、输电塔、桥梁等人造目标巡检过程中存在的视点规划依赖人工经验、覆盖效率低及成像质量不稳定等问题，对无人机多视点自动规划方法进行了研究。采用以人造目标三维模型为输入的方式，通过体素化与面片近似实现了目标表面建模。在综合考虑安全距离、成像分辨率和姿态约束的基础上，生成了多视点候选集，并构建了面片与视点的覆盖关系模型。在此基础上，引入了加权Next-Best-View（NBV）策略，实现了满足覆盖率要求条件下的视点优选。经实验测试，与随机采样和规则化环绕飞行方法相比，所提方法在不同复杂程度的目标巡检任务中均表现出更优的覆盖性能与效率。在相同视点数量条件下，表面覆盖率由17.09%–72.34%提升至86.92%–95.74%，加权覆盖率由11.73%–75.52%提升至86.79%–99.30%。同时，在达到相同覆盖率指标的前提下，所需视点数量显著减少。经实际应用验证，该方法能够满足复杂人造目标巡检任务的工程适用性要求。

Abstract:

To address the problems of manual experience dependent viewpoint planning, low coverage efficiency, and unstable imaging quality in the inspection of man made structures such as buildings, transmission towers, and bridges, a study was conducted on an automated multi viewpoint planning method for UAV based inspection. A target 3D model was adopted as input, and surface modeling was achieved through voxelization and facet approximation. A candidate viewpoint set was generated by integrating constraints including safety distance, imaging resolution, and attitude, and a facet viewpoint coverage relationship model was subsequently constructed. On this basis, a weighted Next Best View (NBV) strategy was introduced to optimize viewpoint selection while meeting coverage requirements. Experimental results show that, compared with random sampling and regular orbiting flight methods, the proposed method exhibits superior coverage performance and efficiency in inspection tasks of varying complexity. Under the same number of viewpoints, surface coverage increases from 17.09%–72.34% to 86.92%–95.74%, and weighted coverage improves from 11.73%–75.52% to 86.79%–99.30%. Furthermore, the number of viewpoints required to achieve the same coverage target is significantly reduced. Practical application demonstrates that the proposed method satisfies the engineering applicability requirements for complex man made target inspection tasks.