基于改进豪斯多夫距离的非参数轮廓变点识别

A Change Point Detection Method Based on Modified Hausdorff Distance in Nonparametric Profiles

轮廓监控中的变点识别问题是统计过程控制的重要研究内容。以非参数轮廓数据为研究对象,运用图像特征识别的豪斯多夫距离测量了样本轮廓之间的特征差异,设计了改进方法,提出了基于二维空间的改进豪斯多夫距离算法,以识别非参数轮廓变点。大量的仿真与论证表明,改进方法在识别变点位置和稳定性方面具有优异的性能。

Change point identification in profile monitoring is an important research topic in statis- tical process control. Herein,the profile had nonparametric characteristics. The proposed method was based on Hausdorff distance, and could be used to measure difference between profiles. A modified Hausdorff distance algorithm was proposed to identify nonparametric profile change point. The com- parison results of simulation study show that when there exists local changes in nonparametric pro- file,the modified algorithem has advantages in locating change points and performance stability.