针对当前TFT-LCD玻璃基板在线检测过程中存在的显微图像对焦模糊问题，对基于图像处理的自动调焦方法在TFT-LCD玻璃基板检测中的应用进行了研究。基于对TFT-LCD液晶面板的光学检测，重点介绍了目前较为常用的、具有代表性的几种基于图像处理的清晰度评价算法,并从算法的单峰性、无偏性、灵敏性以及计算量等方面进行了比较与分析，提出了一种粗细调相结合的复合自动对焦方法，分别选取斜边缘检测算子和梯度滤波器算子作为调焦粗搜索和精搜索阶段的对焦评价算子。实验结果表明：使用调焦行程为15步的全局搜索算法，传统单一算法耗时2.18 s，得到图像的清晰度评价函数值为46.78；粗细调相结合的复合自动对焦方法时间仅为1.41 s,得到图像的清晰度评价函数值为52.49。该方法不仅对焦精度高，有较大的对焦范围，还能保证高的计算效率。

The auto-focusing devices in TFT-LCD glass substrate line detection process have problem of blurred image. In order to solve this problem, the application of image based on auto-focusing method in TFT-LCD glass substrate detection process was presented. Based on the optical detection of the TFT-LCD liquid crystal panel, this paper mainly introduces several typical image processing algorithms based on image processing. The focusing evaluation functions are compared and analyzed quantitatively through experiments in the aspects of unique value, unbiased, sensitivity and calculation speed. An evaluation method combining coarse and fine adjustment is presented. The Lean operator and Brenner operator were respectively chosen as the focus measure method of coarse and fine adjustment. The result show that when a lens of 15 steps focus length for local full search is used, traditional single algorithm takes 2.18 s, the new method takes 1.41 s, and get clarity-evaluation function value is 52.49 higher than 46.78 of traditional single algorithm. This method is not only high focusing accuracy, a greater focus range, but also can ensure high computational efficiency.