基于非线性扩散与图像配准的印刷品表面缺陷检测

Printed Product Surface Defect Detection Based on Nonlinear Diffusion and Image Comparison

针对目前工业印刷品表面缺陷检测时误判率高、检测效率较低以及缺陷区域信息有限等问题,笔者提出一种基于改进Perona-Malik(P-M)非线性扩散模型与图像差分模型配准的印刷品表面缺陷检测方法。该方法首先利用非线性扩散模型,在维持非缺陷区域平滑程度保持不变的情况下,对待测图像进行平滑处理,将原始图像与扩散后图像做差分运算以增强缺陷区域的对比度,从而得到差分图像;然后,将差分图像与标准模板图像进行配准,检测得到缺陷所在的区域的特征值,以便对印刷品印刷质量进行分析;笔者搭建了实验平台,来验证本方法的可行性。实验结果表明:该方法能够快速并且有效地将印刷品表面的缺陷检测出来,准确率高达99.97%。该项研究能够满足印刷品生产质量需求。

Aiming at the problems of high false positive rate,low detection efficiency and limited defect area information in the detection of industrial defects on surface defects,a detection method of surface defects of printed matter based on improving the Perona-Malik(PM)nonlinear diffusion model and image difference model registration was proposed.Firstly,the nonlinear diffusion model was used to smooth the image to be measured while maintaining the smoothness of the non-defective region.The original image and the diffused image were compared to enhance the contrast of the defect region,so as to obtain the difference image.Secondly,the difference image was registered with the standard template image to obtain the eigenvalues of the area where the defect was located.The obtained eigenvalues were used for the analysis of the print quality of the printed matter.An experimental platform was set up to verify the feasibility of the proposed method.The experimental results show that the method can detect the defects on the surface of the printed surface quickly and effectively,and the accuracy is up to 99.97%,which can meet the production quality requirements.