基于改进YOLOX的X射线违禁物品检测

武连全; 楚宪腾; 杨海涛; 牛瑾琳; 韩虹; 王华朋

基于改进YOLOX的X射线违禁物品检测

1.
中国刑事警察学院警察技能战术训练部，辽宁沈阳 110854
2.
中国刑事警察学院公安信息技术与情报学院，辽宁沈阳 110854

基金项目:

公共安全风险防控与应急技术装备”国家重点专项2018年度项目 2018YFC0810102

详细信息

作者简介:
武连全（1979-），男，硕士，副教授，硕士生导师，主要从事警务指挥与战术、反恐处置与大数据应用研究。E-mail：wu_lianquan0402@126.com

中图分类号: TP391.4
计量
- 文章访问数: 132
- HTML全文浏览量: 50
- PDF下载量: 19
- 被引次数: 0
出版历程
- 收稿日期: 2022-03-21
- 修回日期: 2022-04-21
- 刊出日期: 2023-04-20

X-ray Detection of Prohibited Items Based on Improved YOLOX

1.
Department of Police Skills and Tactics Training, Criminal Investigation Police University of China, Shenyang 110854, China
2.
Video and Audio Material Examination Department, Criminal Investigation Police University of China, Shenyang 110854, China

摘要

摘要: 在安全检查过程中快速准确地识别违禁物品有利于维护公共安全。针对X射线行李图像中存在的物品堆叠变形、复杂背景干扰、小尺寸违禁物品检测等问题，提出一种改进模型用于违禁物品检测。改进基于YOLOX模型进行，首先在主干网络中引入注意力机制加强神经网络对违禁品的感知能力；其次在Neck部分改进多尺度特征融合方式，在特征金字塔结构后加入Bottom-up结构，增强网络细节表现能力以此提高对小目标的识别率；最后针对损失函数计算的弊端改进IOU损失的计算方式，并根据违禁物品检测任务特点改进各类损失函数的权重，增大对网络误判的惩罚来优化模型。使用该改进模型在SIXray数据集上进行实验，mAP达到89.72%，FPS到达111.7 frame/s具备快速性和有效性，所提模型与阶段主流模型相比准确率和检测速度都有所提升。
- YOLOX /
- X射线图像 /
- 违禁品 /
- 注意力机制
Abstract: In the process of security inspection, rapid and accurate identification of prohibited items is conducive to maintaining public security. To address the problems of stack deformation, complex background interference, and small-sized contraband detection in X-ray luggage images, an improved model for contraband detection is proposed. This improvement is based on the YOLOX model. First, an attention mechanism was introduced into the backbone network to enhance the ability of the neural network to perceive contrabands. Second, in the neck part, the multi-scale feature fusion method was improved upon, and a bottom-up structure was added after the feature pyramid structure to enhance the performance ability of the network for details, thereby improving the recognition rate of small targets. Finally, the calculation method based on IOU loss was upgraded in view of the disadvantages of the loss function calculation. The weights of various loss functions were also increased according to the characteristics of the contraband detection task, and the punishment of network misjudgment was increased to optimize the model. Upon using the improved model on the SiXray dataset, an mAP of 89.72% was attained and a fast and effective FPS arrival rate of 111.7 frame/s was achieved. Compared with mainstream models, the accuracy and detection speed of the proposed model were improved.
- YOLOX /
- X-ray image /
- prohibited items /
- attention mechanism

HTML全文

图 1 数据增强效果

Figure 1. Example of data augment

下载: 全尺寸图片幻灯片

图 2 改进后的网络结构

Figure 2. Improved model structure

下载: 全尺寸图片幻灯片

图 3 CBAM算法流程

Figure 3. The process of CBAM module

下载: 全尺寸图片幻灯片

图 4 改进后的Neck

Figure 4. Improve Neck structure

下载: 全尺寸图片幻灯片

图 5 比对模型对各类违禁物品的P-R曲线

Figure 5. P-R curves of different models for various prohibited items

下载: 全尺寸图片幻灯片

图 6 改进模型检测效果

Figure 6. Experimental renderings

下载: 全尺寸图片幻灯片

表 1 CBAM不同添加位置的结果

Table 1. Results of different add CBAM locations %

Location	Gun	Knife	Pliers	Wrench	Scissor	Map
CSP_1	95.45	86.43	86.89	83.52	81.61	86.78
CSP_2	97.12	85.56	87.47	84.98	82.67	87.56
CSP_3	96.14	85.23	88.45	85.17	83.24	87.65
CSP_4	97.45	87.77	89.49	86.49	83.71	88.98

下载: 导出CSV

表 2 改进策略的消融实验

Table 2. Ablation study Ablation experiments with improved strategies %

CBAM	Bottom-up	Loss	Gun	Knife	Pliers	Wrench	Scissor	Map
-	-	-	97.32	81.07	88.24	87.25	79.72	86.72
√	-	-	97.45	87.77	89.49	86.49	83.71	88.98
-	√	-	97.49	87.82	89.13	86.51	82.91	88.77
-	-	√	97.46	84.08	87.86	86.47	82.18	87.61
√	√	√	97.57	88.74	89.26	88.97	84.05	89.72

下载: 导出CSV

表 3 对比实验结果

Table 3. Comparative experimental results

Models	Map/(%)	FPS/(frame/s)
Fast R-CNN	80.23	52.8
RetinaNet	83.94	55.3
YOLOv3	85.93	56.9
YOLOv4	86.12	73.8
YOLOv5s	89.12	98.5
Guo’s^[26]	73.68	55
Mu’s^[27]	80.16	25
Dong’s^[28]	89.60	-
Ours	89.72	111.7

下载: 导出CSV

参考文献(32)

[1]	陈冰. 基于多能X射线成像的违禁物品自动识别[D]. 北京: 北京理工大学, 2018. CHEN B. Automatic Recognition of Prohibited Items Based on Multi-energy X-ray Imaging[D]. Beijing: Beijing Institute of Technology, 2018.
[2]	邰仁忠. X射线物理学[J]. 物理, 2021, 50(8): 501-511. https://www.cnki.com.cn/Article/CJFDTOTAL-WLZZ202108003.htm TAI R Z. X-ray physics[J]. Physics, 2021, 50(8): 501-511. https://www.cnki.com.cn/Article/CJFDTOTAL-WLZZ202108003.htm
[3]	McCarley J S, Kramer A F, Wickens C D, et al. Visual skills in airport-security screening[J]. Psychological Science, 2004, 15(5): 302-306. doi: 10.1111/j.0956-7976.2004.00673.x
[4]	梁添汾, 张南峰, 张艳喜, 等. 违禁品X光图像检测技术应用研究进展综述[J]. 计算机工程与应用, 2021, 57(16): 74-82. doi: 10.3778/j.issn.1002-8331.2103-0476 LIANG T F, ZHANG N F, ZHANG Y X, et al. Summary of research progress on application of prohibited item detection in X-ray images[J]. Computer Engineering and Applications, 2021, 57(16): 74-82. doi: 10.3778/j.issn.1002-8331.2103-0476
[5]	Mery D, Mondragon G, Riffo V, et al. Detection of regular objects in baggage using multiple X-ray views[J]. Insight-Non-Destructive Testing and Condition Monitoring, 2013, 55(1): 16-20. doi: 10.1784/insi.2012.55.1.16
[6]	Michel S, Mendes M, de Ruiter J C, et al. Increasing X-ray image interpretation competency of cargo security screeners[J]. International Journal of Industrial Ergonomics, 2014, 44(4): 551-560. doi: 10.1016/j.ergon.2014.03.007
[7]	韩萍, 刘则徐, 何炜琨. 一种有效的机场安检X光手提行李图像两级增强方法[J]. 光电工程, 2011, 38(7): 99-105. https://www.cnki.com.cn/Article/CJFDTOTAL-GDGC201107023.htm HAN P, LIU Z X, HE W K, An efficient two-stage enhancement algorithm of X-ray carry-on luggage images[J]. Opto-Electronic Engineering, 2011, 38(7): 99-105. https://www.cnki.com.cn/Article/CJFDTOTAL-GDGC201107023.htm
[8]	Khan S U, Chai W Y, See C S, et al. X-ray image enhancement using a boundary division wiener filter and wavelet-based image fusion approach[J]. Journal of Information Processing Systems, 2016, 12(1): 35-45.
[9]	ZHAO B, Wolter S, Greenberg J A. Application of machine learning to x-ray diffraction-based classification[C]//Anomaly Detection and Imaging with X-Rays(ADIX) Ⅲ. International Society for Optics and Photonics, 2018, 10632: 1063205.
[10]	Gaus Y F A, Bhowmik N, Breckon T P. On the use of deep learning for the detection of firearms in x-ray baggage security imagery[C]//2019 IEEE International Symposium on Technologies for Homeland Security (HST), 2019: 1-7.
[11]	Franzel T, Schmidt U, Roth S. Object detection in multi-view X-ray images[C]//Joint DAGM (German Association for Pattern Recognition) and OAGM Symposium, 2012: 144-154.
[12]	王宇, 邹文辉, 杨晓敏, 等. 基于计算机视觉的X射线图像异物分类研究[J]. 液晶与显示, 2017, 32(4): 287-293. https://www.cnki.com.cn/Article/CJFDTOTAL-YJYS201704008.htm WANG Y, ZOU W H, YANG X M, et al. X-ray image illegal object classification based on computer vision[J]. Chinese Journal of Liquid Crystals and Displays, 2017, 32(4): 287-293. https://www.cnki.com.cn/Article/CJFDTOTAL-YJYS201704008.htm
[13]	Alom M Z, Taha T M, Yakopcic C, et al. The history began from alexnet: a comprehensive survey on deep learning approaches[J/OL]. arXiv preprint arXiv: 1803.01164, 2018.
[14]	WANG L, GUO S, HUANG W, et al. Places205-vggnet models for scene recognition[J/OL]. arXiv preprint arXiv: 1508.01667, 2015.
[15]	Ballester P, Araujo R M. On the performance of GoogLeNet and AlexNet applied to sketches[C]//Thirtieth AAAI Conference on Artificial Intelligence, 2016, 30(1): doi: https://doi.org/10.1609/aaai.v30i1.10171.
[16]	Haque M F, Lim H Y, Kang D S. Object detection based on VGG with ResNet network[C]//2019 International Conference on Electronics, Information, and Communication (ICEIC) of IEEE, 2019: 1-3(doi: 10.23919/ELINFOCOM.2019.8706476).
[17]	ZOU Z, SHI Z, GUO Y, et al. Object detection in 20 years: a survey[J/OL]. arXiv preprint arXiv: 1905.05055, 2019.
[18]	CHEN C, LIU M Y, Tuzel O, et al. R-CNN for small object detection[C]//Asian Conference on Computer Vision, 2016: 214-230.
[19]	Girshick R. Fast r-cnn[C]//Proceedings of the IEEE International Conference on Computer Vision, 2015: 1440-1448.
[20]	R Ren S, He K, Girshick R, et al. Faster R-CNN: Towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2017, 39(6): 1137-1149.
[21]	ZHANG Y, KONG W, LI D, et al. On using XMC R-CNN model for contraband detection within X-ray baggage security images[J]. Mathematical Problems in Engineering, 2020, 2020: 1-14.
[22]	Sigman J B, Spell G P, LIANG K J, et al. Background adaptive faster R-CNN for semi-supervised convolutional object detection of threats in x-ray images[C]//Anomaly Detection and Imaging with X-Rays (ADIX) V, 2020, 11404: 1140404.
[23]	Papageorgiou C P, Oren M, Poggio T. A general framework for object detection[C]//Sixth International Conference on Computer Vision (IEEE Cat. No. 98CH36271) of IEEE, 1998: 555-562.
[24]	LIU Z, LI J, SHU Y, et al. Detection and recognition of security detection object based on YOLO9000[C]//2018 5th International Conference on Systems and Informatics (ICSAI)of IEEE, 2018: 278-282.
[25]	Galvez R L, Dadios E P, Bandala A A, et al. YOLO-based Threat Object Detection in X-ray Images[C]//2019 IEEE 11th International Conference on Humanoid, Nanotechnology, Information Technology, Communication and Control, Environment, and Management (HNICEM), 2019: 1-5.
[26]	郭守向, 张良. Yolo-C: 基于单阶段网络的X光图像违禁品检测[J]. 激光与光电子学进展, 2021, 58(8): 0810003. https://www.cnki.com.cn/Article/CJFDTOTAL-JGDJ202108007.htm GUO S X, ZHANG L. Yolo-C: one-stage network for prohibited items detection within X-ray images[J]. Laser & Optoelectronics Progress, 2021, 58(8): 0810003. https://www.cnki.com.cn/Article/CJFDTOTAL-JGDJ202108007.htm
[27]	穆思奇, 林进健, 汪海泉, 等. 基于改进YOLOv4的X射线图像违禁品检测算法[J]. 兵工学报, 2021, 42(12): 2675-2683. https://www.cnki.com.cn/Article/CJFDTOTAL-BIGO202112015.htm MU S Q, LIN J J, WANG H Q, et al. An algorithm for detection of prohibited items in X-ray images based on improved YOLOv4[J]. Acta Armamentarii, 2021, 42(12): 2675-2683. https://www.cnki.com.cn/Article/CJFDTOTAL-BIGO202112015.htm
[28]	董乙杉, 李兆鑫, 郭靖圆, 等. 一种改进YOLOv5的X光违禁品检测模型[J/OL]. 激光与光电子学进展, [2022-02-21]. http://kns.cnki.net/kcms/detail/31.1690.TN.20220217.1141.008.html. DONG Y S, LI Z X, GU J Y, et al. An improved YOLOv5 model for X-ray prohibited items detection[J]. Laser & Optoelectronics Progress: [2022-02-21]. http://kns.cnki.net/kcms/detail/31.1690.TN.20220217.1141.008.html.
[29]	GE Z, LIU S, WANG F, et al. YOLOX: Exceeding Yolo series in 2021[J/OL]. arXiv preprint arXiv: 2107.08430, 2021.
[30]	WANG C Y, LIAO H, WU Y H, et al. CSPNet: a new backbone that can enhance learning capability of CNN[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2020: 390-391.
[31]	Woo S, Park J, Lee J Y, et al. Cbam: convolutional block attention module[C]//Proceedings of the European Conference on Computer Vision (ECCV), 2018: 3-19.
[32]	MIAO C, XIE L, WAN F, et al. SiXray: a large-scale security inspection x-ray benchmark for prohibited item discovery in overlapping images[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019: 2119-2128.