TFT-LCD表面缺陷检测中一维DFT方法中邻域r的自动选取

Automatic Neighbor r Selection for One-dimensional DFT Method in the Surface Defect Inspection of TFT-LCD

在TFT-LCD平板表面缺陷检测中,为了自动获取一维DFT方法中的邻域r,首先将无缺陷的TFT-LCD平板表面图像的每一行经过预处理,再计算其灰度共生矩阵,然后提取灰度共生矩阵的均匀度,最后计算不同r时的均匀度差值。最大差值所对应的r就是最佳邻域。搭建了TFT-LCD平板检测实验系统,实验对不同长度的非整周期、整周期、非整周期补整的图像进行处理。从获取的r可知,在相同环境下,非整周期补整后的r和整周期的r相同。将这一结果应用到实际缺陷检测中,系统可以准确地检测出缺陷。

In order to obtain neighbor r automatically,each line of a faultless surface image for TFT-LCD panel was preprocessed firstly in the detection of surface defects on TFT-LCD panel.Then thier gray level co-occurrence matrix（GLCM）was calculated,and the homogeneity property of the GLCM was extracted.Finally,the difference of homogeneity was computed,and the optimal neighbor was that level where the difference showed a maximum value.An experimental system of TFT-LCD panel detection was built.The images with various length including integer-period,non-integer-period and non-integer-period filled into integer-period were processed in the experiments.From the acquisition of r,it may be seen that the rfor non-integer-period fills into integer-period is the same as it for integer-period under the same environments.And by applying the results into actual defect detection,the system may detect the defects accurately.