目的 研究印品图像的各类形状缺陷， 建立基于支持向量机（Support vector machine， SVM）的印品形状缺陷分类模型。方法 对印品进行符合人眼视觉特性的缺陷识别， 并对提取缺陷进行特征分析。将特征数据导入支持向量机进行训练学习， SVM分类器对缺陷图像进行测试。结果 分类器对点缺陷和面缺陷的识别率为100%， 对线缺陷的分类准确率达93.94%。结论 基于SVM的缺陷分类方法能较好地满足印品质量检测的需求。

Objective To get a classifier model based on support vector machine (SVM) for printing defects detection. Methods The characteristics of printing defects were extracted according to human vision characteristics, and which were then analyzed. After that these feature data were imported into SVM for learning and training. The SVM classifier was then used to test the defect images. Results The classification accuracy rate of the classifier for point defects and planar defects reached 100%, and the classification accuracy rate for linear defects was 93.94%. Conclusion The classification method based on SVM could meet the demand of printing defect detection.