【作者】 葛动元；

【导师】 张庆英； 赵天婵；

【作者基本信息】 武汉理工大学 ， 机械设计及理论， 2004， 硕士

【摘要】 计算机视觉作为一种非接触式的先进测量技术，具有精度高、非接触、效率高、成本低等诸多优点，随着计算机技术以及摄像机制造技术工艺的提高，图像处理技术的不断发展，其测量精度在不断的提高，在生产实践中的应用也日益广泛。 测量的精度和效率，决定着科学技术和工业的水平。测量，是利用一定手段采集数据，并通过对这些数据的处理得到科学的结论，从而进一步地指导生产实践。测量本身就是一种实验。产品的生产是一个连续的过程，对精度的控制也应当全面地、系统地进行。精度系数和精度指标，可作为系统精度及质量的综合评判指标。现代的质量控制，强调变事后检验为动态控制，这就需要依赖各种先进的测量手段。计算机视觉测量技术就是随着科学技术的发展和生产实践的需求应运而生的。它的核心及最终目标是利用二维投影图像来重构三维物体的几何信息，其中的关键技术之一就是图像处理。由于各种干扰的存在，以及因逆成像而使得深度信息的丢失，使得计算机视觉研究领域的大部分问题是病态的。这种病态问题的求解已经成为计算机视觉研究中一个非常重要的方面。在本论文所依托的课题研究中，以计算机视觉测量技术在工业检测中的应用为研究对象，针对计算机视觉检测中的一些关键性问题，如摄像机参数的标定、图像处理、工程计算中的数据处理和精度控制、特征匹配等，从理论上进行了深入的研究与探讨，采用较高精度的图像处理技术和能够保持数值稳定性与求解精度的构造性方法，提高了摄像机的标定精度以及测量精度，通过对被测量工件特征的三维重建，较好地满足了测量的精度要求。以圆锥类零件的锥度作为特定对象，进行了较为精确的检测实践，达到了一定的测量精度。 论文的主要内容包括以下几个方面： (1)论述了在经典的边缘算子基础上发展起来的最优边缘检测算子，比如LOG算子、坎尼边缘检测算子等检测定位精度较高的最优算子； (2)针对计算机视觉求解中的病态问题，利用构造性数值求解方法，比如最小二乘法以及具有不改变被变换向量长度的豪斯荷尔德正交变换，保证了求解数值的稳定性； (3)以透视理论为基础，根据计算机图形学的知识，推导并建立了决定空间武汉理工大学硕士学位论文物体某点的几何位置与该点在采集图像上相应位置关系的摄像机模型，即针孔模型; (4).根据曲面左右图像的外轮廓线的点不对应物体同一点的特殊性，阐述了对以二次曲线为基元进行三维重建的方法，以此得到相应的空间三维几何信息，并介绍了该测量系统的一个应用实例。最后在总结全文的基础上，对进一步的研究提出了建议和展望。更多还原

【Abstract】 As an advanced measurement technology, Computer Vision ( CV ) has many merits such as high accuracy, non-touch, high efficiency, low cost, etc. With the improvement of the computer technology and the techniques of camera fabricating and development of image processing technology, CV has been adopted as a practical tool and played a powerful role in industry measurement.Precision and effectiveness of measurement are crucial to the development of science and technology, as well as industry. What we called measurement is the experimental process of collecting data, and dealing with the data in a correct way in order to get a reasonable result, and in turn, to direct the production properly. Since the production is a continuous process, the precision control should be made comprehensively and systematically. Both precision coefficient and its index can be taken as the synthesis index to judge the quality and precision of the system. From the view point of modern quality control, checking-out afterwards should be replaced by the way of dynamic controlling beforehand. Thus various advanced measurements are considered to be the useful and essential ways. With the development of science and technology and need of practice the computer vision emerged as the times require. The core and ultimate goal of CV is to re-construct three dimensional shape by means of two dimensional information. One of the key technologies is image processing. Because of the interference of different factors and lost of deep-information in image inverting, the ill-posed problem becomes common, and its solution decides the reliability of this measurement. In our research project, by taking the industry applying as an object, we did a lot of work to deal with different problems such as camera calibration, image processing, data treating and precision controlling, and, feature matching as well. In the paper, those problem are discussed theoretically and practically. An available way of constructing which ensure the accuracy and reliability the measurement is pointed out, and the three dimensional rebuilding the configuration of the work is satisfying. Conic parts are taken as example to show the result of the method which has enough accuracy and effectiveness.The dissertation mainly includes several parts as follows:1. Discussing of some optimal edge-detected operators, such as LOG operator, Canny edge-detected operator, which are developed on the base of classical ones, and shown higher precision than the formers.2. To solve the ill-posed problem mentioned above, choosing a useful way of constructing mathematics model with Householder orthogonal transformation to keep the method and its result in a good condition;3. On the base of perspective theory and graphic knowledge, inferring and constructing camera model (pin-hole model) to determine geometry local of some feature points in world and its coordinate on the of the camera.4. According to the special nature that the outline of curved surface in left and right images does not refer to the same outline in physical surface, expounding the ways of three dimensional restructure with base element of curved line, obtaining their three dimensions geometry information. An application example is given to show the result of the measurement system.更多还原