Abstract: For electronic component color features in the automatic optical inspection system, a method of solder joint color feature extraction is proposed based on color calibration target and fuzzy membership functions. The proposed approach smooths H and S component of color calibration target in HSV space. Feature functions formed by Gauss model are fitted by using least square fitting. The membership functions are created through a linear combination of feature functions. The approach obtains feature color regions with image segmentation in terms of fuzzy method criteria. Experiments are carried out with 3 sets of PCB images. The result demonstrates that the proposed approach can obtain stable extraction and low differences in intra-class datas with evaluation of Bezdek’s partition coefficient 0.68 and variance 1.0E-4. The approach is also good at applicability, which can be widely applied to special color extraction system.

Key words: automatic optical inspection, color feature segmentation, color calibration target, fuzzy theory

摘要： 针对自动光学检测系统中元件颜色特征，提出了一种依靠颜色标定块和模糊隶属度判别法进行焊点特征颜色提取的方法。该方法对颜色标定块HSV中的H与S分量进行统计直方图分析，以高斯函数为模型拟合出各分量特征函数，建立起了一种以H和S分量为线性特征组合的各个颜色隶属度函数。根据模糊法隶属度判定准则对图片中颜色进行分割，提取出特征颜色。最后将该算法对3组PCB图像进行分割实验。实验结果表明该算法Bezdek分配系数为0.68左右，方差数据为1.0E-4数量级，表明该算法聚类效果稳定，类内数据相比于主色判别式法的一致性高。此外该算法适用性强，可普遍运用于指定颜色的提取系统中。

关键词: 自动光学检测, 颜色特征提取, 颜色标定块, 模糊理论