基于改进显著性模型的TFT-LCD面板缺陷检测

Surface defect inspection of TFT-LCD panels based on improved saliency model

在薄膜晶体管液晶显示器(TFT-LCD)面板缺陷检测中,周期性的背景纹理和缺陷可用基于谱残差的显著性模型进行分离,但该模型对缺陷大小敏感。本研究对该模型进行改进,将其用于检测由面板图像二维离散傅里叶变换(DFT)获得的能量谱中的高能量成分,并将该部分消除,再经过二维离散傅里叶逆变换(IDFT)对图像重构,达到去除空域图像周期性纹理和保留缺陷的目的。其中,模型的中均值滤波器窗口的大小、邻域的大小和能量谱中心高能量保护区域的大小可由处理后图像灰度共生矩阵的逆差矩确定。实验结果表明,在固定参数的条件下,改进后的模型对包含纤维、污渍和划痕的TFT-LCD面板缺陷均能正确检测,结果不受缺陷大小和灰度值以及面板周期性纹理方向的影响。

In the surface defect inspection of thin film transistor-liquid crystal display (TFT-LCD) panels,the periodic textured background and defects can be separated by the saliency detection model based on spectral residual,but the parameters in the proposed model are sensitive to the defect sizes. The model was improved and applied to the detection of high-energy frequency components in the spectrum which was obtained from the original image by two-dimensional discrete Fourier transform (DFT). The neighborhood of those areas associated with high-energy frequency components in the spectrum that represented background texture was set to zero,and then a spatial domain image was reconstructed by the two-dimensional inverse discrete Fourier transform (IDFT) in order to remove the background texture and retain the defects. The gray level co-occurrence matrix (GLCM) of the output image was calculated,and after that the size of mean filter window,the radius of neighborhood and the parameter of center protected energy spectrum could be obtained by the homogeneity property of the GLCM. The experiments on a variety of surface defects such as fibers,stains and scratches in TFTLCD panel testified the effectiveness of the improved method under fixed parameters.