【摘要】 液晶显示界面缺陷检测是过程装备领域的重要研究课题之一。针对传统检测方法效率低、误检率高等问题，提出了一种基于结构相似度的缺陷检测算法，并建立了高精度室内视觉检测系统。该方法通过计算标准图像与被检图像的局部结构相似度差异定位缺陷区域，结合热力图对缺陷特征进行可视化表征，实现了缺陷位置与形态的直观显示。试验以分辨率480×800像素的TFT液晶屏为对象，结果表明，在窗口大小30×30像素、相似度阈值0.1、缺陷分辨率10×10像素的条件下，系统单次检测耗时小于1 s，误检率低至4.4%，检测效率与精度均满足生产的要求。更多还原

【Abstract】 Defect detection in liquid crystal display interfaces is one of the critical research topics in the field of process equipment engineering. To address the issues of low efficiency and high false detection rates in traditional methods,this study proposes a defect detection algorithm based on structural similarity and establishes a high-precision indoor visual inspection system. The method locates defect regions by calculating the local structural similarity differences between standard images and test images,and visually characterizes defect features through heatmaps,enabling intuitive visualization of defect locations and morphologies. Experiments on TFT liquid crystal screens( resolution of 480 × 800 pixels) show that under the conditions of a 30 × 30 pixels window size,a similarity threshold of 0. 1,and a defect resolution of 10 × 10 pixels,the system achieves a detection time of less than 1 second per test with a false detection rate as low as 4. 4%,meeting both efficiency and accuracy requirements for industrial applications.更多还原