RGB三通道缺陷检测投票分类器方法

RGB three-channel defect detection voting classifier method

针对低温多晶硅/金属氧化物(LTPS/IGZO)玻璃基板缺陷检测图像具有周期性和固定性的特性,为提高缺陷的识别性能,提出一种投票分类器。将待检测的LTPS/IGZO玻璃基板的RGB (红绿蓝三通道色彩模式)缺陷图像分成R、G、B这3种颜色的合成图像,设计一个投票分类器,对缺陷图片和正常图片的R、G、B颜色所占比例进行对比,使用投票的方式,多数服从少数原则,检测待检测图像是否含有缺陷,判断缺陷类型。实验结果表明,所提方法在加快缺陷检测速度的基础上,保证了产品的识别准确率。

Aiming at the periodicity and fixity characteristics of low temperature polycrystalline silicon/indium gallium zinc oxide(LTPS/IGZO)glass substrate defect detection images,to improve the recognition performance of the defect,a voting classifier was proposed.The RGB defect detection images of the detected LTPS/IGZO glass substrate were divided into synthetic images of three colors of R,G and B,and a voting classifier was designed,and the R,G,B color proportion of image defects and normal images were compared using ballot method,most of the minority obey principle was used to determine whether the image contained defects,and the defect types were determined.Experimental results show that using the proposed method speeds up defect detection process,and generating good product detection rate is ensured.