目的 针对铝塑泡罩药品人工检测时存在的包装缺陷，如效率低、成本高、稳定性差等，采用机器视觉技术对铝塑泡罩药品包装进行缺陷检测。方法 采用快速鲁棒特征SURF提取算法、BOW算法和单分类支持向量机组成的缺陷检测算法框架，并完成铝塑泡罩药品包装缺陷检测系统的开发。通过搭建的实验平台获取280幅铝塑泡罩药品图像，并采用文中所提方法对180幅图像实施缺陷检测。结果 实验结果显示，在阈值为1900、视觉单词数量为120、惩罚因子为0.9时，文中方法的准确率为99.4%。结论 文中方法提高了铝塑泡罩药品包装缺陷检测的准确率和稳定性。

The work aims to detect the defects of drug packaging with aluminum plastic bubble cap by means of machine vision technology, regarding such packaging defects as low efficiency, high cost and poor stability in the manual detection of drug with aluminum plastic bubble cap. A defect detection algorithm framework consisting of a fast robust feature SURF extraction algorithm, a BOW algorithm and a one-class support vector machine were adopted and the defect detection system for the drug packaging with aluminum plastic bubble cap was developed. Based on the experimental platform, 280 images of drug with aluminum plastic bubble cap were obtained, and the defects of 180 images were detected in the proposed method. The experimental results showed that, when the threshold was 1,900, the number of visual words was 120 and the penalty factor was 0.9, the accuracy of the proposed method was 99.4%. The proposed method improves the accuracy and stability of the defect detection of drug packaging with aluminum plastic bubble cap.