一种改进YOLOv5的X光违禁品检测模型  

Improved YOLOv5 Model for X-Ray Prohibited Item Detection

作　　者：董乙杉 李兆鑫 郭靖圆 陈天宇 卢树华[1,2] Dong Yishan;Li Zhaoxin;Guo Jingyuan;Chen Tianyu;Lu Shuhua(College of Information and Cyber Security,People’s Public Security University of China,Beijing 102600,China;Key Laboratory of Security Technology and Risk Assessment Ministry of Public Security,Beijing 102600,China)

机构地区：[1]中国人民公安大学信息网络安全学院,北京102600 [2]公安部安全防范技术与风险评估重点实验室,北京102600

出　　处：《激光与光电子学进展》2023年第4期349-356,共8页Laser & Optoelectronics Progress

基　　金：中央高校基本科研业务经费重大项目(2021JKF102);公安学科基础理论研究专项(2021XKZX08)。

摘　　要：针对X光行李图像安全检测过程中存在物品高度重叠遮挡及复杂背景干扰等问题,提出了一种融合注意力机制、数据增强策略与加权边框融合算法的改进YOLOv5网络模型用于X光违禁品检测。模型在Neck部分引入卷积注意力模块加强网络对违禁品深层重要特征的提取,抑制背景干扰;训练阶段采用Mixup数据增强策略模拟带有高度重叠及遮挡物品的检测场景,加强模型复杂样本的学习能力;测试阶段采用加权边框融合算法对冗余预测框进行优化,提高模型精准预测能力。所提模型在3个大型复杂数据集SIXray、HiXray、OPIXray进行了测试,平均精度均值分别达到了89.6%、83.1%和91.6%。结果表明:所提模型能够有效提高YOLOv5检测复杂违禁品的能力,与现有诸多先进算法相比,具有较高的准确率和稳健性。An improved YOLOv5 network model is proposed to resolve high overlap,heavy occlusion,and complex background interference in Xray luggage image security detection by introducing the convolutional block attention module,data enhancement strategy,and the weighted boxes fusion algorithm for Xray prohibited item detection.The convolutional block attention module is introduced in the Neck to enhance the extraction of deep important features and suppress background interference of Xray prohibited items features.The Mixup data augmentation strategy is employed during the training process to simulate the detection scene with high overlap and heavy occlusion items to strengthen the learning ability of the model for complex samples.During the testing process,the weighted boxes fusion algorithm is used to optimize the redundant prediction boxes to enhance its prediction accuracy.The proposed model is tested on three largesize complex datasets(SIXray,HiXray,and OPIXray),resulting in mean average precision values of 89.6%,83.1%,and 91.6%,respectively.The results show that the proposed model can effectively improve the ability of YOLOv5 in detecting complex contrabands.The proposed model performs better than many current advanced algorithms,indicating its high accuracy and robustness.

关 键 词：X光图像 违禁品检测 注意力机制 加权边框融合 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]