显微视觉自动聚焦研究

提出一种评估空域聚焦函数的方法。从计算时间与聚焦曲线变化率两个方面，提出四个参量作为评估依据．可从多个函数中筛选出最佳聚焦测度。针对研制的微对准装配自动聚焦系统的评估结果表明：灰度熵函数是适合于该系统的最佳聚焦测度。为克服聚焦搜索运动中的机械传动回差，提出不同于传统聚焦策略的“定位法”。该方法的核心是建立聚焦函数值与离焦量之间的函数关系，提出采用多项式拟合、高斯函数拟合以及样条插值拟合的方法可以构建这种函数关系；并给出了计算结果。最后综合灰度熵函数与定位法策略进行自动聚焦实验．得到自动聚焦精度为3．4μm；完成一次自动聚焦平均时间为3．3s。实验结果证明所述方法可以实现准确、实时地自动聚焦。

Research on Autofocus Technique in Micro-Vision

An evaluation method of autofocus function based on spatial domain is presented. Four features are proposed to evaluate the function on computation time and sharpness of the focus curve. An optimum can be filtrated from many classic functions. Entropy function is the evaluating result through experiments on the developed autofocus system of micro assembly workcell. A novel focus searching algorithm——positioning algorithm (PA) which is different from hill climbing searching alorithm (HCSA) and Fibonacci Searching Algorithm (FSA) is put forward. Setting up the relation between defocus value and focus function value is key factor of the algorithm. Three methods: polynomial fitting, Gaussian function fitting and cubic spline interpolation fitting are presented to realize PA. Then an autofocus system comprising entropy function and PA is set up. Experimental results indicate that focusing precision can reach 6.992 μm and one focusing processing can be done within 3.4 s. It proves that the evaluation method and the PA can realize autofocus accurately and quickly.