目的 针对试剂卡生产企业采用人工分选印刷缺陷的试剂卡存在效率低、成本高、易漏检的问题，提出一种基于深度神经网络YOLOv5s的改进试剂卡印刷缺陷检测算法YOLOv5s-EF。方法 通过图像预处理算法获得高质量的缺陷图像数据集，在YOLOv5s的主干特征提取网络中添加高效通道注意力(Efficient Channel Attention, ECA)机制，增强特征图中重要特征的表示能力;引入焦点损失函数(Focal Loss)来缓解正负样本不均衡的影响;结合印刷区域的定位结果，二次精确定位并构建方位特征向量，提出一种特征向量相似度匹配方法。结果 实验结果表明，本文提出的试剂卡印刷缺陷检测算法在测试集上的检测平均准确度可以达到97.3%，速度为22.6帧/s。结论 相较于其他网络模型，本文提出的方法可以实现对多种印刷缺陷的识别与定位，模型具有较好的检测速度和鲁棒性，有利于提高企业生产的智能化水平。

The work aims to propose an improved reagent card printing defect detection algorithm YOLOv5s-EF based on deep neural network YOLOv5s to solve the problems of low efficiency, high cost and easy to miss detection in manual sorting of reagent cards with printing defects in reagent card manufacturers. High-quality defect image data sets were obtained by image preprocessing algorithm. Efficient Channel Attention (ECA) mechanism was added to the backbone feature extraction network of YOLOv5s to enhance the representation ability of important features in feature maps. Focal loss function was introduced to alleviate the influence of imbalance between positive and negative samples. Combined with the positioning results of the printing area, a method of similarity matching of feature vectors was proposed, which was based on the quadratic accurate positioning and the construction of azimuth feature vectors. The experimental results showed that the average detection accuracy of the reagent card printing defect detection algorithm proposed in this paper could reach 97.3% and the speed was 22.6 FPS on the test set. Compared with other network models, it can identify and locate various printing defects. The model has good detection speed and robustness, which is beneficial to improve the intelligent level of enterprise production.