基于深度学习的偏光片检测技术研究

Research on Polarizing Film Detection Technology Based on Deep Learning

偏光片外观缺陷的在线检测一直以来就是业界难题,深度学习工具的出现有助于改善这一现状。实验中,对含有缺陷的偏光片进行图像采集,并将采集到的图像分成训练集和验证集。在训练集中利用深度学习工具学习到了缺陷的特征阈值,将阈值应用到验证集中进行缺陷检测,得到很好的检测效果。然而,打痕缺陷由于图像采集的原因并不能完全检出。此外,偏光片自身的翘曲对检测也有一定程度的影响。

The on-line inspection of the appearance defects of polarizing film has always been a difficult problem in the industry,and the appearance of deep learning tools can help to improve this situation.In the experiment,the polarizer with defects is collected and divided into training set and verification set.In the training concentration,the defect feature threshold is learned by means of deep learning tools,and the threshold value is applied to the verification set for defect detection,and a good detection effect is obtained.However,the imperfection can not be detected completely because of the image acquisition.In addition,the warpage of the polarizer itself also has a certain degree of influence on the detection.