中文摘要：

针对ITO导电薄膜缺陷人工分类效率较低的问题,提出了一种基于灰度差分统计法、灰度共生矩阵法和矩描述法的缺陷特征分析方法。首先,通过灰度差分统计对比得出各缺陷的熵值,利用熵值进行一次分类;其次,通过灰度共生矩阵,提取了二阶矩、对比度和相关性的数值,通过这三个数值分别设置阈值并对图表进行分析完成对5种缺陷的分类;最后,根据矩描述的特性并通过实验对矩方法的可行性进行了分析。实验中,取熵值的阈值为0.5,二阶矩的阈值为230,对比度的阈值为140,相关性的阈值为22,通过这四个参数的阈值可以达到将5种缺陷分类的效果。实验结果表明,基于灰度差分统计法和灰度共生矩阵法的分类方法可以更有效地完成缺陷分类。

英文摘要：

Since the artificial classification of Indium Tin Oxide（ ITO） conductive film is inefficient, a method based on gray differential statistical method, grey-level co-occurrence matrix and moment was proposed. Firstly, the entropy value was figured out using gray differential statistical method to make the first classification; Secondly, the second moment value,contrast value and correlation value were extracted and the classification of defects was finished by threshold values of the parameters; Finally, the feasibility of moment method was analyzed by experiments. In the comparison experiments, the threshold value of entropy was 0. 5, the threshold value of second moment was 230, the threshold value of contrast was 140,and the threshold value of correlation was 22. The classification result of five kinds of defects was achieved using the four threshold values. The experimental results show that the method based on gray differential statistical method and grey-level cooccurrence matrix method is more efficient.