为解决一元测量系统分析方法无法准确评价机器视觉系统能力的问题,运用多元测量系统分析方法,通过探讨机器视觉系统的特点,建立其测量值的变异源模型,并通过交叉试验和多元方差分析(MANOVA)方法估计各变异源的方差协方差矩阵,据此计算它的重复性和再现性指标,给出对机器视觉系统的评价.结果表明:将这些方法运用到芯片规格测量的具体案例中,并与一元方法进行对比,取得了较好的效果.
Univariate measurement system analysis may not assess the multivariate systems accurately. To address this is- sue, influential factors in the machine vision system is explored and a new sources of variation model for the measured values is established. Crossed design and multivariate analysis of variance (MANOVA) method are used to estimate the covariance matrices, which are the inputs for multivariate reproducibility and repeatability. The aforementioned method is applied in a chip size machine vision system and good results are achieved.