基于机器视觉的光纤头表面缺陷检测应用研究-模式识别与智能系统专业论文.docxVIP

万方数据 万方数据 摘要 摘 要 光纤头作为连接两根光纤的专用设备，其表面缺陷会使光能在光纤内传播过 程中经过光纤头时产生高损耗，从而影响光纤网络的性能。传统的由人工负责完 成的光纤头表面缺陷检测已经不能满足社会需求。近年来，基于机器视觉的产品 表面缺陷检测系统因其检测速度快、精度高、无损伤等诸多优点在现代工业的自 动化生产中起着越来越重要的作用。 本文主要研究基于机器视觉的光纤头表面缺陷检测方法，包括光纤头表面的 同心度检测和划痕检测两个方面。在光纤头表面同心度检测方面，本文通过旋转 光纤头时由显微镜获取的多幅光纤头表面图像中内圆圆心的位置拟合出内圆圆心 的轨迹圆的半径，即光纤头表面同心度。为了能更精确地求出每幅光纤头表面图 像中内圆圆心的位置，本文首先通过二值化、内圆轮廓提取和形态学处理得到内 圆的边缘掩膜，再结合 Canny 边缘检测获取内圆的待选边缘像素点，并经随机采 样一致选出最能表示内圆边缘的像素点，然后通过三次 Facet 模型拟合出内圆的边 缘亚像素位置，进而拟合出内圆的圆心位置。 在光纤头表面划痕检测方面，本文首先通过基于梯度特征的区域生长得到多 个线支撑区域，然后每个线支撑区域用一个适当的矩形来逼近。为了能尽量减少 误检并准确定位划痕，本文还对逼近每个线支撑区域的矩形进行验证，最后经过 划痕合并得到真正的划痕。 最后，本文根据对光纤头表面缺陷检测中的同心度检测和划痕检测的研究， 分别做了光纤头表面同心度检测实验和光纤头表面划痕检测实验，并认真细致地 分析了实验结果。实验结果表明，本文研究的基于机器视觉的光纤头表面缺陷检 测方法能够快速准确地检测出光纤头表面的同心度和划痕，可应用于实际的光纤 头表面缺陷检测项目中，从而达到了预期目的。 关键字：机器视觉，表面缺陷检测，同心度检测，划痕检测 I ABSTRACT ABSTRACT Optical fiber connector, as the necessary equipment which connects two optical fibers, its surface defects can make a head high loss when light signal that transmit in optical fiber through optical fiber connector and affect the performance of the optical fiber network. The traditional method of optical fiber connector surface defects detection, mainly completed by artificially responsible for, has been unable to meet demand. In recent years, the system of product surface defects detection based on machine vision technology by its fast detection speed, high precision, no damage, and many other advantages plays a more and more important role in the modern industrial automation production. This paper mainly studies the surface defects detection method of optical fiber connector, which based on machine vision technology, including concentricity detection and scratches detection two aspects. In surface concentricity detection of optical fiber connector, the paper fit the surface concentricity of optical fiber connector by position of the center of internal circle in many surface images of optical fiber connector which are obtained through the microscope when the optical fiber connector is rotated. I