自动焊机中定位技术的研究

【作者】 钟雪灵；

【导师】 鲍苏苏；

【作者基本信息】 华南师范大学 ， 计算机软件与理论， 2005， 硕士

【摘要】 超声铝丝焊接机和超声金丝球焊接机主要运用于半导体生产后工序中芯片焊盘与外框架间引线的焊接。本课题是广东省教育厅科学基金项目,对中小企业目前现有的手动焊接设备进行自动化改造。项目所完成的自动设备,采用机器视觉自动识别技术,不需要每片都人工干预,同类工件只需要进行一次检测识别,以后无需再人工操作焊接,人工只需要上下工件即可。可以做到一人多机,在生产效率上有较大幅度的提高。可以说在今后相当一段时间内,是中小企业所希望装备的产品。本论文完成的主要研究如下:1、 提出手动焊机改造的方案,具体设计了自动焊机的硬件系统和软件系统。同时重点分析图像识别系统可以利用的两种方法,即硬件型图像处理系统和软件型图像处理系统,并结合课题的实际情况选择采用后一种方法。2、 针对噪声污染,系统采用先平滑后增强的预处理。预处理过程为首先对图像进行中值滤波,在保持边缘的同时有效地滤除噪声干扰,然后通过直方图均衡提高图像的对比度,忽略了一些不稳定的细节。为加快速度,对图像的预处理只针对模板图像,而不针对场景图像。3、 讨论分析目前常用的基于灰度的模板匹配技术,即FFT相关算法、序贯相似检测算法(SSDA)、幅度排序相关算法、多辨率塔型结构算法。在本课题中,精度与速度要求极高。经实验对比发现,序贯相似检测算法具有较高的计算速度、定位精度,比较适合于高要求的工业应用中。为进一步满足系统的要求,本文改进了自适应门限序列的SSDA算法,将匹配过程分为两步进行,第一步为粗匹配,第二为精匹配,大大提高定位的精度与速度。4、 自动焊机的最终目的是精确定位焊点。焊点定位是指精确快速找到每对焊点的位置,即每对焊点的坐标。对处于相同的外部环境,同一种芯片型号,摄像头位置不改变的条件下,焊点到芯片中心的距离大致一样。但由于在手工贴芯片时,肯定存在芯片的旋转与偏移误差,这个距离还是会有所偏差。基于此,项目采用类比学习来解决这个问题。其过程为:人工引导测定各个焊点在芯片坐标(以芯片中心为原点)中的坐标,并保存下来,然后利用模板匹更多还原

【Abstract】 Ultrasonic aluminum wire bonder and ultrasonic gold wire bonder are mainly applied to joint the chip pad with the lead-frame in the later process of semiconductor producing. This project aims at the achieving automatic reconstruction of these manual equipments using in the small and medium-sized enterprises, which is funded by Department of Education of Guangdong Province.The automatic equipment after reconstructing adopts the machine vision automatic recognition technology. If the same type of chips needs to be jointed, the workers just need to inspect and identify once, and then the system will run automatically, no need interfere any more. The rest task for workers is to convey the chips. So one worker can operate several automatic bonders, and naturally the productivity improves considerably. This automatic bonder will be welcomed by the small and medium-sized enterprises in the future. In order to achieve the target, our project is to complete this automatic reconstruction. This dissertation introduces and summarizes the main parts of this project.The primary research is as follows:(1) The reconstruction proposal is brought forward, and the hardware system and software system of automatic bonder are specifically designed. At the same time, two schemes that can be used are analyzed with emphasis, that is, the hardware image processing system and the software image processing system. Compared with these two schemes, we decide to adopt the latter method combining with the actual condition.(2) The quality of image declines a bit because of the noise pollution. To counter this problem, we determine to smooth the image firstly, and then enhance the image. The course of pre-processing is as follow: first, the image is smoothed by the median filter. The image is filtered the disturbance of noise while the edge is retained. Then the image is enhanced by gray histogram trimming and improved its contrast, neglecting some unsteady details. Consequently the pre-processing paves the way forlatter image matching.(3) Some commonly used fast image matching algorithm base on gray matching are discussed, including FFT algorithm, Multitude Sorting Relational algorithm, Sequential Similarity Detection algorithm, Multi-resolution Pagoda Structure algorithm and Genetic algorithm. In this project, precision and velocity is highly demand. By comparison, Sequential Similarity Detection algorithm bears the characteristics of speedy calculation, accurate orientation and anti-jamming, which is fit for the high requirement in industry. In this project, we improve the arithmetic of SSDA and divide the matching process into two steps: the first one is the rough matching and the second one is the precise matching. This greatly improves the precision and pace in orientation.(4) The automatic jointing machine is to locate bonding point preciously, which means locating bonding point with the place of every couple of bonding points in a precise way. In detail, the places means coordinates of the bonding points. In the condition of the same outer environment and type of chips and the place of camera remain unaltered, the distance between bonding point and chip center is approximately the same. If the chips are mounted on frame manually, there must be the rotation and excursion of chips existing , so there will be some deviations in the distance.In order to solve this problem, we decide to apply analogy study. The course is as follow: First the distance between welding points of chip center is measured by manual work. And then the position of chip is confirmed by template matching. Last, through modifying, the precise coordinates is calculated. This method of locating bonding point is precious and quick.The productions of this project improve the manufacturing status of our nation’s small and medium-sized enterprises. It’s of significant for raising the large economic and society benefit. The outlook of industrialization is optimistic. Besides that, this research points out a new approach for further developmen更多还原