Abstract: Aiming at the problem of high positioning cost,low efficiency and difficult to guarantee accuracy in the processing of aircraft blades,a 3D point cloud boundary point feature registration algorithm was proposed to identify the position and pose of aviation blades,which could quickly and accurately locate the specific position of aircraft blades in space through a single scan.The point cloud data of the workpiece was obtained by the structured light camera and preprocessed.Then,the boundary points of the point cloud were extracted as the key points for registration based on the projection vector angle of neighboring points on the cutting plane.The Fast Point Feature Histogram(FPFH) descriptor of the key points was calculated,and the rough registration matrix was solved by using the Sample Consensus Initial Alignment(SAC-IA) algorithm.The point-to-plane Iterative Closest Point(ICP) algorithm was used for accurate registration.Robot vision localization experiments were carried out to verify the method,and the results showed that the algorithm significantly improved the registration accuracy for aircraft blades with a mean square error of 0.6540mm2.The recognition error for the position of aircraft blades in space was within 1mm,which met the requirements of positioning.

Key words: aircraft blades, machining positioning, point cloud registration, feature of boundary

摘要： 针对航空叶片在加工过程中定位成本高、效率低且精度难以保证的问题,提出一种基于3D点云边界点特征配准算法的航空叶片位姿识别方法,通过一次扫描即可快速、精确定位航空叶片在空间中的具体位置。首先,通过结构光相机获取工件的点云数据并对其进行预处理;随后,根据邻域点在切平面的投影向量夹角提取点云边界点作为配准的关键点;计算关键点的快速点特征直方图(FPFH)描述子,采用采样一致性初始配准(SAC-IA)算法求解粗配准矩阵;然后使用点到面的最近点迭代(ICP)算法进行精配准。最后,进行机器人视觉定位实验验证,结果表明针对航空叶片,所提算法显著提高了配准精度,均方误差为0.654 0 mm2;对于航空叶片在空间中位置的识别误差在1 mm以内,能够满足定位需求。

关键词: 航空叶片, 加工定位, 点云配准, 边界特征