中文摘要：

为了实现面向芯片封装的视觉精确定位技术,提高芯片引线键合的加工精度和效率,提出了一套结合快速傅里叶变换互相关和不变矩的视觉定位算法,构建了视觉系统实验平台,并进行了实验研究.针对芯片和引线框架图像的特点,分别提出了适合于芯片和引线框架的定位算法.对于芯片图像,采用快速傅里叶变换互相关算法进行匹配定位;对于引线框架图像,采用基于特征的不变矩方法进行匹配定位.使用显微视觉系统和二维运动平台构建了实验平台,进行了视觉定位实验.经过实验验证,定位精度均在亚像素级,芯片图像的定位精度小于0.4,像素（2,μm）,引线框架的位置精度小于0.5像素（2.3,μm）,转角的定位精度小于0.1°,能够满足芯片封装的要求,从而实现引线键合的视觉精确定位.

英文摘要：

In order to realize the high-precision vision localization technology and improve the precision and effi- ciency of integrated circuit （IC） wire bonding, a set of vision localization algorithm using fast Fourier transformation （FFT） and invariant moments was presented. The vision system for experiment was established, and experimental research was carried out. The FFT correlation method was applied for locating chip images, and invariant moments method based on features was proposed for locating leadframe images. The experimental table was established using micro-vision system and 2-DOF motion stage. The results of experiments showed that the precision of localization was under sub-pixel level. The positional accuracy of chip was less than 0.4 pixel （2 μm）, the positional accuracy of lead- frame was less than 0.5 pixel （2.3 μm）, and the accuracy of rotation is less than 0.1°. The precision was able to meet the requirements oflC wire bonding, and then the high-precision vision localization could be achieved.