基于改进型B样条分层插值的汽车零件检测研究

Research on auto parts inspection based on improved B-spline hierarchical interpolation

在汽车零件的在线装配过程中,为了减小由理论数模与实际位置之间的偏差导致的装配误差,设计了改进型B样条分层插值算法。首先,根据点云重建曲面的曲率分布对初始数据进行分层,然后再通过三次B样条插值算法获得实际零件点云与数模点云的偏差分布,最后由绝对距离偏差值推导得到路径补偿归一化系数,从而实现对汽车零件自动检测的扫描路径优化。仿真对包含直角、棱边及孔洞特征的零件进行分析,结果显示点云去噪效果良好,重建曲面形貌符合实体零件特征。实验对比了直接重建与采用三次B样条分层插值算法优化后的重建效果,通过对300个测试点的数据统计,三个坐标轴方向的位置优化前后平均误差分别为3.345 mm和0.599 mm。可见,优化后位置预测精度得到了较好的提升,为提高扫描路径精度奠定了基础。

In the online assembly process of auto parts,in order to reduce the assembly error caused by the deviation between the theoretical digital model and the actual position,an improved B-spline hierarchical interpolation algorithm is designed.First,the initial data is layered according to the curvature distribution of the reconstructed surface of the point cloud,and then the deviation distribution of the actual part point cloud and the digital-analog point cloud is obtained through the cubic B-spline interpolation algorithm.Finally,the path compensation is derived from the absolute distance deviation value normalization coefficient,so as to realize the optimization of scanning path for auto parts detection.The parts containing right angles,edges and holes are analyzed through the simulation,and the results show that the denoising effect of the point cloud is good,and the surface shape of the reconstructed surface conforms to the characteristics of the solid part.In the experiment,the reconstruction effect of direct reconstruction and the optimization of the cubic B-spline hierarchical interpolation are compared.Through the statistics of 300 test points,the average errors of the three coordinate axis directions before and after optimization are 3.345 mm and 0.599 mm,respectively.It is demonstrated that the position prediction accuracy has been improved after optimization,laying a foundation for improving the accuracy of the scanning path.