基于可变形卷积与注意力机制的X光安检违禁品检测  

Prohibited items detection based on deformable convolution and attention mechanism in X-ray security inspection

作　　者：粟兴旺 王晓明 黄金玻 许茹玉 吴琳 Su Xingwang;Wang Xiaoming;Huang Jinbo;Xu Ruyu;Wu Lin(School of Computer and Software Engineer,Xihua University,Chengdu 610039,China)

机构地区：[1]西华大学计算机与软件工程学院,成都610039

出　　处：《电子测量技术》2023年第10期98-108,共11页Electronic Measurement Technology

基　　金：四川省自然科学基金(2022NSFSC0533);西华大学研究生创新基金(ycjj2021030)项目资助。

摘　　要：X光安检违禁品检测被广泛应用于维护公共交通安全和人身安全。针对X光图像中违禁品存在形状尺度多变、重叠遮挡严重等问题,提出一种结合可变形卷积与注意力机制改进的YOLOv5s模型用于违禁品检测。首先在主干网络中引入可变形卷积,通过学习采样偏移量来适应物体的不同形变,增强空间特征信息提取能力;其次利用混合卷积注意力模块加强模型对检测目标的感知能力,抑制无关背景干扰;然后构造通道引导的空洞空间金字塔模块,获取更加准确的全局上下文信息,提高模型对重叠遮挡目标的识别能力;最后采用CARAFE算子代替最近邻插值,在上采样过程中充分利用内容信息,提高模型检测精度。在SIXray_OD和OPIXray数据集上实验结果显示,所提出模型的mAP@0.5相较于原YOLOv5s分别提高了2.1%和1.8%,达到了90.6%和90.0%。与现有诸多先进算法相比,具有较好的检测精度与实时性。Prohibited items detection in X-ray security inspection is widely used to maintain public traffic safety and personal safety.In order to solve the problems of variable shape and scale,severe overlap and occlusion in X-ray images,an improved YOLOv5s model combining deformable convolution and attention mechanism is proposed for prohibited items detection.Firstly,deformable convolution is introduced into the backbone network to enhance spatial feature information extraction by learning sampling offsets to adapt to different deformations of objects.Secondly,the mixed convolution attention module is used to enhance the model’s ability to perceive the detected target and suppress irrelevant background interference.Then a channel-guided atrous space pyramid module is constructed to obtain more accurate global contextual information and improve the model's ability to identify overlapping occlusion targets.Finally,the CARAFE operator is used to replace the nearest neighbor interpolation to make full use of the content information in the upsampling process and improve model’s detection accuracy.The experimental results on the SIXray_OD and OPIXray datasets show that the model’s mAP@0.5 is 2.1%and 1.8%higher than the original YOLOv5s,reaching 90.6%and 90.0%,respectively.Compared with many existing advanced algorithms,it has better detection accuracy and real-time performance.

关 键 词：X光图像 违禁品检测 可变形卷积 注意力机制 上采样 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]