【作者】 孙双花；

【导师】 叶声华； 曲兴华；

【作者基本信息】 天津大学 ， 测试计量技术及仪器， 2007， 博士

【摘要】 随着计算机视觉技术和光电技术的飞速发展,视觉测量技术得到了迅速发展和广泛应用,其作为一种非接触测量手段已经越来越引起人们的重视。本文深入研究了基于计算机视觉测量技术的系统方案、软硬件设计,对视觉检测的图像获取、图像处理、亚像素定位技术及系统标定、调焦的关键技术进行了深入分析,并实现了视觉测量在喷油器底座工件自动检测系统中的应用。喷油器底座是汽车油路中的重要部件,由金属精密冲压、内圆磨削而成。该类零件主要用于与柱形零件进行精密配合焊接装配,为保证装配精度,需要对零件的内尺寸进行精密测量。喷油器底座呈碗形,内壁表面是强反射表面,碗口处与圆倒角相连,底部与碗底的曲面连接。采用视觉测量的方法,以内壁口处边缘的投影作为被测边缘,建立了工件视觉自动检测系统,提高了工件测量的精度和速度。本文研究的主要内容有:1.基于视觉测量技术建立了喷油器底座内尺寸自动检测系统,完成了视觉系统、软件系统和控制执行系统的软硬件设计。强反射表面的检测是视觉检测的难题,通过对工件反射性能的分析,设计了发光强度连续可调的均匀散射LED背光光源,使得被测边缘在图像中真实体现出来,实现了工件强反射内表面的视觉检测,提高了测量精度。设计了工件的自动上料、自动定位及自动分选机构,真正实现了工件的全自动化测量,提高了检测效率。2.对视觉测量系统的图像处理算法进行了研究。研究了视觉测量系统中图像预处理、阈值分割的算法;对常用边缘检测算法进行分析,根据本系统测量图像边缘简单、对比度高的特点,提出了基于阈值分割及链码跟踪的边缘检测算法;分析了二次曲线拟合法、多项式拟合法、空间矩、灰度矩等亚像素算法,并进行了实验研究。3.对视觉测量系统的标定技术进行了研究。分析了二维视觉测量系统的摄像机成像的数学模型;设计了精密平面网格靶标,提出了基于灰度统计分布的图像分割方法进行特征点提取;提出基于图像的测量平面与光轴垂直度的快速调整方法;对该测量系统用实验分析了现有的标定方法,并提出分区线性标定方法,将测量图像分成多个小区域,对每个区域分别进行线性标定,减小了标定误差。4.对视觉测量系统的调焦技术进行实验研究。对常用的调焦评价函数进行了对比性实验,分析了调焦区域、边缘点数及光源强度对调焦函数的影响。用实验分析了调焦精度对标定结果的影响。提出了一种基于特征边缘跟踪的调焦评价方法,解决了被测工件图像中边缘在高度上不共面时普通调焦函数无法进行正确调焦定位的问题,提高了工件的调焦精度。5.分析了曲线拟合特征参数的计算方法。对工件测量系统进行了实验研究和误差分析。更多还原

【Abstract】 With the rapid development of computer vision technology and photovoltaic technology, computer vision based precision measurement technology has been well developed and widely applied in various fields. As a non-contact measurement approach, computer vision measurement technique is playing a more and more significant role.The thesis has deeply investigated a series of relevant key topics, including system scheme of precision measurement based on the computer vision, design of software and hardware system, acquisition and processing of the measurement images, sub-pixel positioning, calibration of the measurement system, the key technology of focusing. Furthermore, the built computer vision measurement system is successfully applied in the automatic measurement system for injector base.The injector base, fabricated by metal casting and stamping, is an important part in automobile oil circuits. Such parts are mainly used for welding assembly with cylinder parts. In order to ensure the accuracy of assembly, the inside dimension of the parts are required to be measured accurately. The injector base displays as a bowl shape, and the inner surface is a strong reflecting surface. The edge of the base is connected with the circular bevel, and the bottom surface is linked with the curved face of the base. The projection of the inner surface edge is employed as the measurement object. With the built automatic computer vision precise measurement system, the accuracy and speed of the part measurement are greatly improved.The main contributions of the thesis include:1. An automatic computer-vision based system for the precise measurement of the injector base inner dimension is designed and built. The relevant designs of computer-vision system, software system, controlling and actuating system are completed. The detection of strong reflecting surface is always a challenging topic in the field of computer vision based measurement technique. Through the analysis on the reflecting feature of the object part, we design a LED back lighting source with uniform scattering, hence, the edge in the measurement image is greatly enhanced. Consequently, the vision measurement of the inner surface of the part is completed with a higher accuracy. Additionally, the automatic charging of the parts, automatic positioning, and automatic sorting mechanism are design to realize a genuine automatic measurement and further enhance the efficiency of measurement.2. The imaging processing algorithms for the computer vision based measurement system are studied. The preprocessing and threshold segmentation algorithms are investigated. Through the analysis of common edge detection algorithms and the simple edge, high-contrast characteristics of the measurement image, we propose a novel edge detection algorithm based on the edge segmentation and the chain code tracking. In addition, several sub-pixel algorithms, including the conic curve fitting, polynomial fitting, space mement, grey moment, are studied both theoretically and experimentally.3. The calibration technique for the computer vision based measurement system is studied. The mathematic model of camera imaging is analyzed, and the appropriate plane grid target for the experiment is designed. Furthermore, a rapid adjustment method based on the image measurement plane and the verticality of optical axis is proposed. According to the existing calibration approach for the measurement system, we present a novel method using linear calibration for different zones. The method put the measurement image into several zones so that the linear calibration can work for different zones respectively. As a result, the calibration errors are minimized.4. The focusing technique for the measurement system is experimentally studied. The common evaluation functions for focusing are experimentally compared. And the influences resulting from the focusing region, number of edge point, and the intensity of light source are analyzed. And the influence on the calibration result due to the accuracy of focusing is experimentally analyzed. A focusing evaluation method based on the characteristic edge tracking is propose, with the result that the evaluation function can be correctly focusing and positioning even when the height of the edge in the part measurement image is not coplanar. Hence, the focusing accuracy is improved to a higher degree.5. The characteristic parameters of the curve fitting method are analyzed. And the measurement system is experimentally studied and the relevant errors are investigated.更多还原