基于机器视觉的树脂镜片水印疵病检测

Automatic detection of water-mask for resin glasses by machine vision

为实现对树脂镜片水印缺陷的自动检测,建立了点光源投影成像光学仿真模型,并对光源与被检镜片相对距离、成像光屏反射率、镜片物理特性以及成像距离等参数进行了探索,以此为据设计了一套基于机器视觉的水印疵病自动检测系统装置,可实现对镜片投影图像的采集与分析.采用区域生长、形态学运算和椭圆拟合等算法将镜片投影区域从背景中分割,结合Sobel边缘检测、阈值分割方法提取投影图像灰度突变区域,并以面积阈值为约束条件实现对水印缺陷的识别与定位.试验结果表明,该方法可对不同类型镜片具有自适应性,对水印缺陷镜片判别率为100%,图像采集与分析平均耗时872 ms.

To realize the automatic detection of watermask for resin glasses, the point light projection illumination model was built to explore the influence factors of relative distance between light source and resin glasses, imaging screen reflectivity, glasses physical properties and imaging distance on the imaging. Based on the simulation result, the automatic detection system for watermask was designed to capture and analyze the projection image. The algorithms of region growing, morphological operation and ellipsefitting were employed to segment the glasses projection region from the background. Sobel edge detection and thresholding were combined to extract the region with sudden change of grayscale in local pixel space. The connected region area was calculated and served as constraint condition for further selection of the defect region. The experimental results show that the digital image process can be applied to different type of resin glasses. The recognition of glasses watermask can reach 100%, and the process of imaging, recognition and location consumes 872 ms.