【作者】 韩瑞雨；

【导师】 王晋疆；

【作者基本信息】 天津大学 ， 光学工程， 2012， 硕士

【摘要】 基于计算机视觉的精密测量技术具有操作简易、非接触、精度高等特点,广泛应用于工程领域,也适用于微零件装配的高精度在线测量。光学系统对所成图像的细节分辨能力是影响图像的视觉测量精度的主要因素:当图像对焦时才最清楚,具有更多的细节,便于后续的精密测量工作;而当图像离焦时,图像模糊,细节丢失,就无法对微零件进行形状或者位置的检测。因此如何获取对焦面就是基于计算机视觉精密测量的重要前提。本论文叙述了自动对焦的历史背景,分析了图像离焦的具体原因,详细介绍了自动对焦的各种传统方法,并且对基于图像的自动对焦技术进行了深入的研究,这包括了图像的预处理技术、对焦区域的设计技术、清晰度评价函数的技术以及聚焦点搜索算法技术。1)对图像的几种滤波去噪的方法进行了实验验证,分析了其对清晰度评价函数曲线的影响,综合考虑了滤波效果和算法速度,最终选择中值滤波作为本课题滤波方法;2)对几种对焦区域的设计算法进行了实验验证,并且结合本课题,对柱腔的待测对象采用手动对焦区域设计算法、对低对比度下的微零件小球的待测对象采用环状区域设计算法;3)研究了各种聚焦点搜索算法,对两种新型的搜索算法,通过仿真实验分析了其优缺点,并且将基于最优化理论的抛物线法设计为聚焦点的搜索算法,与其它几种常用的搜索算法进行了实验比较,说明了其耗时较少,对焦较准确的优点。4)研究了各种常用的清晰度评价函数,对其进行了实验与分析,比较了各种函数的优劣性;针对低对比度环境下大多数清晰度评价函数失去理想特性的问题,提出了一种基于Zernike正交矩的清晰度评价函数,通过调节各阶矩的权重系数,可以适应不同对比度下的自动对焦。5)搭建了自动对焦实验平台,对本文研究的自动对焦算法进行了系统实现。更多还原

【Abstract】 Computer vision-based precision measurement technology with easy operation, strong anti-jamming, high precision, has been widely used in the engineering fields, applicable to micro-parts assembly process of high-precision measurement. the ability of imaging systems for optical images to distinguish details that affect the accuracy of digital image optical measurement: When the image is in focus, it is most clear with more detail, the precise measurement is easier; and when the image deviate from the focus, the image blur, detail is lost, detect the shape or location of micro-parts becoming more difficult. So how to get the focus area exactly has become a urgent problem to solve.This paper studies the historical development of auto-focus, analyzed the specific reasons of out-of-focus image, described the various traditional methods of auto-focus in detail, and in-depth studied image-based autofocus technology, which including the image preprocessing technology ,focus area designing, autofocusing function and the focal point of the search algorithm technology.1) Verified several image filtering denoising method through experiments , analyzed their impact on AF function curve, considering the filtering results and algorithms speed, finally chose median filtering as the subject filtering method;2) Verified several focus area designing algorithms, combined with the issue, designed the algorithm with manual focus area for the column cavity, designed the algorithm with circular focus area for the micro-ball under low contrast;3) Studied a variety of focal point searching algorithms, through simulation experiments, analyzed two new search algorithms’advantages and disadvantages, and designed a new focal point searching algorithm with parabolic method based on based on optimization theory, compared with several common searching algorithms, demonstrated its advantages of less time-consuming, focusing accurately;4) Studied a variety of commonly used AF function, compared the advantages and disadvantages of various functions through analysis, and for the problem that the most AF functions losing their ideal characteristics under low-contrast environment , proposed an AF function based on Zernike orthogonal moments, by adjusting the weight coefficients of order moments, can adapt to the autofocus under different contrast;5) Built an auto-focus experimental platform, achieved the AF algorithm of the paper studied through the system.更多还原