【作者】 吕清花；

【导师】 钟毓宁；

【作者基本信息】 湖北工业大学 ， 测试计量技术及仪器， 2006， 硕士

【摘要】 视觉自动对准技术是集成电路后封装设备中的一项重要单元技术。视觉自动对准系统由光学照明系统、光学成像系统、CCD摄像器件、图像处理软件和运动控制系统等部分组成,其中图像处理算法对视觉自动对准系统的精度起着决定性的作用。本文研究了相关的图像处理算法,实现由粗到精的高精度视觉对准定位,全文主要包括以下几个方面的内容:(1)根据机器视觉原理和自动对准技术,建立了自动对准系统的构架,实现图像自动采集,自动定位的目的。(2)根据自动对准系统精度的需要,对CCD摄像机的比例系数、光学中心以及系统的像素转换当量等参数进行了标定。(3)分析了定位标记的作用和特征,将其分为两类,一类是粗定位标记,另一类为精定位标记。同时,运用一系列图像处理技术,例如:灰度直方图分析、图像平滑、域值分割、有效区域提取(ROI)、边缘提取等,将两组定位标记分别提取出来。(4)结合图像定位标记的分类,运用图像匹配算法实现由粗到精的自动对准过程。首先,利用形心与重心匹配算法找到粗定位标记,实现粗定位;然后对粗定位后图像进行ROI提取,在ROI区域采用当前视觉测量的一种较先进的点模式匹配的方法实现精确定位。在实现精定位之前,为了有效突出定位标记的特征,提高整个系统的定位精度和定位效率,运用基于空间矩的亚像素边缘提取方法,实现定位标记边缘的亚像素级跟踪提取,再利用SUSAN角点提取的方法对亚像素级的边缘进行特征点的有效提取。(5)运用仿真分析的方法对边缘提取的精度和匹配的精度进行了实验分析.实验表明,该系统具有较高的精度、较强的鲁棒性和实用性。更多还原

【Abstract】 Vision automatic alignment technique is one of the most important techniques in IC encapsulation. The vision automatic alignment system comprises optical lighting system, optical imaging system, CCD camera device, image processing software, control system and so on, in which the high precision image processing algorithms are the determinant part for the final accuracy of alignment.In this thesis, the correlative image processing algorithms are presented to realize rough pose estimation and exact pose estimation of high precision vision alignment, and the following parts are studied:(1)Based on the machine vision theory and automatic alignment technique,the truss of the automatic alignment system is built to realize the automatic grabbing image and alignment.(2)Some parameters are calibrated, such as, the dimension proportion of x axis and y axis, the optical center of the CCD camera and the pixel transform quantum of the imaging system, etc.(3)The alignment masks are partitioned to two groups, one group is rough pose estimation marks, and the other group is exact pose estimation marks. And then the marks were extracted from the image by a series of image processing algorithms, such as, histogram analyses, image filter, binarization, region of interest, edge detection and so on.(4)There are two steps to finish the alignment and obtain high precision. Firstly, feature pattern matching algorithm is proposed to finish rough pose estimation. And secondly, point pattern matching, whose speed is faster and matching precision is higher, is applied to exact pose estimation. Before exact pose estimation, the sub-pixel edge detection that uses the spatial moments of a gray level to determine the location of edge, and the SUSAN corner detector has been attached importance.(5)With the help of the simulated experiments, the precisions of the edge detector and the matching are studied. The results of experiments indicate that the system is provided with higher precision, stronger reliability and practicability.更多还原