Abstract: In view of the large error existed in registration between defective model and original model by using classic Interactive Closest Point (ICP) algorithm,an improved ICP algorithm was proposed to find the corresponding points between two models to achieve an accurate registration.Based on the change of defective parts in size and appearance,the algorithm combined curvature constraints with distance constraints of corresponding points to set curvature and distance thresholds,which could automatically eliminate defective points cloud.The reliable corresponding points were obtained which could ensure the accuracy of registration.The algorithm was programed by MATLAB and its effectiveness was verified by the registration of worn model.

Key words: remanufacturing, registration, interactive closest point algorithm, adaptive threshold

摘要： 针对损伤零部件点云模型与原始模型利用传统迭代最近点算法配准出现较大误差的问题,提出一种基于该算法的改进算法,以实现两模型间对应点的准确获取,从而实现损伤零部件点云与原始模型的准确配准。考虑到损伤零部件表面尺寸和形貌发生变化,该算法将对应点曲率约束与对应点间的距离约束结合,并设定曲率和距离阈值实现损伤点云的自动剔除,保证了配准点云对应点的准确性及配准的快速性。最后,运用MATLAB实现了算法编写,并通过损伤模具的配准验证了该算法的有效性。

关键词: 再制造, 配准, 迭代最近点算法, 自适应阈值