Abstract: A method for detecting multi-pads in PCB assembly line based on machine vision in real-time is proposed.Multi-resolution strategy is adopted to accelerate the matching speed.Circular projection matching method is used to get the initial candidate matching sub-images based on the gray properties of the template image and the searched image.And then，the mismatch sub-images are removed through the numbers of matching points between matching sub-image and template using SIFT feature matching method.A rough rotated angle is calculated through point pattern matching and the precise angle is gained through re-sampling and interpolation method.Experimental results show the robustness and validity of the proposed method，and can track objects in real-time.

Key words: circular projection matching, Scale Invariance Feature Transform（SIFT）, multi-resolution, point pattern matching

摘要： 探讨了一种基于机器视觉的PCB自动装配线多焊盘实时定位方法。采用多分辨率图像金字塔匹配策略，利用模板图像与待搜索图像的灰度特性，使用圆投影匹配进行初始候选匹配点的选取，得到一系列的候选匹配子图；使用SIFT算法对候选匹配子图和模板图像进行特征匹配，确定对应匹配点，消除误匹配的候选子图；根据点的模式匹配，确定大致的旋转角度，使用重采样和插值的方法计算精确的旋转角度。实验表明，该方法可以准确、实时地实现目标定位。

关键词: 圆投影匹配, 尺度不变特征变换（SIFT）, 多分辨率, 点模式匹配