基于AOI的PCB检测系统中，样本PCB图像与待测PCB图像之间存在错位偏差，需要对准后才能进行下一步的检测工作。首先，依据PCB线路特征，定义了样本图像边缘点的概念，制订了边缘点抽样规则。针对待测PCB图像，提出了基于样本边缘点对准强度评价函数，建立了对准评价模型及对准快速算法。最后，通过PCB对准实验以及大量的实际测试应用，证明对准强度评价模型能准确地描述待测PCB图像与样本图像间的对准程度，对准模型为后期的缺陷搜索提供了可靠的偏移参数。

In the printed circuit board defect inspection based on AOI, there are some dislocate deviation between sample PCB image and the PCB image, which make the need of the image registration to be processed. First of all, define the sample image edge point concept according to the PCB line features, formulate the edge points sampling rules. Then, in the view of the PCB image, propose the registration intensity evaluation based on the sample edge points, and establish of evaluation model and fast algorithm for registration. Finally, through the PCB registration experiment and plenty of application testing, prove that the alignment intensity evaluation model can accurately describe the alignment level for the sample PCB image and the tested PCB image, which can provide reliable migration parameters for the defects search for later period.

郭峰林（1969 — ），男，湖北武汉市人，副教授，研究方向为图像处理和智能检测