Abstract: In order to improve efficiency of digital aircraft skin detection, aiming at the alignment problems of local aircraft skins and the digital model, a local scanning point cloud localization algorithm was proposed based on multi-descriptor voting. The localization region on the digital model was determined by various 3D point cloud descriptors, since they had discrepancy in recognition ability. A weighted voting point method was proposed to decrease deviation. A well-designed comparision experiment and a concrete example on real data were conducted. The results are as follows: compared with the methods based on single descriptor, the localization accuracy is improved by 37.5% on average. The algorithm is feasible in practical engineering.

Key words: aircraft skin, digital inspection, point cloud data, local point cloud localization, 3D point cloud descriptor

摘要： 为了提高数字化飞机蒙皮检测技术的效率，针对其中局部蒙皮和数字模型的对齐问题，提出了基于多描述子投票的飞机蒙皮局部扫描点云定位算法，采用基于测地距离的蒙皮面片选取方法，综合利用多种三维点云描述子的识别能力差异来确定数字模型上的定位区域，同时提出了投票点加权定位法来提高定位位置精度。根据对比实验结果和工程实例上的实际应用，得出结论如下：所提算法相比于基于单一描述子的定位方法定位准确率平均提高约37.5%；所提算法在实际工程中具有可行性。

关键词: 飞机蒙皮, 数字化测量, 点云数据, 局部点云定位, 三维点云描述子