[1]江新玲,杨 乐,朱家辉,等.面向复杂场景的基于改进YOLOX_s的安全帽检测算法[J].南京师大学报(自然科学版),2023,46(02):107-114.[doi:10.3969/j.issn.1001-4616.2023.02.014]
　Jiang Xinling,Yang Le,Zhu Jiahui,et al.Safety Helmet Detection Algorithm Based on Improved YOLOX_s for Complex Scenes[J].Journal of Nanjing Normal University(Natural Science Edition),2023,46(02):107-114.[doi:10.3969/j.issn.1001-4616.2023.02.014]
点击复制

面向复杂场景的基于改进YOLOX_s的安全帽检测算法()
分享到：

《南京师大学报（自然科学版）》[ISSN:1001-4616/CN:32-1239/N]

卷:
第46卷
期数:
2023年02期
页码:
107-114
栏目:
计算机科学与技术
出版日期:
2023-06-15

文章信息/Info

Title:
Safety Helmet Detection Algorithm Based on Improved YOLOX_s for Complex Scenes
文章编号:
1001-4616(2023)02-0107-08
作者:
江新玲1; 杨 乐1; 朱家辉1; 陶 磊1; 刘 峰1; 2; 段倩倩1
(1.太原理工大学信息与计算机学院,山西 晋中 030600)
(2.中国煤炭工业协会,北京 100010)
Author(s):
Jiang Xinling1; Yang Le1; Zhu Jiahui1; Tao Lei1; Liu Feng1; 2; Duan Qianqian1
(1.College of Information and Computer,Taiyuan University of Technology,Jinzhong 030600,China)
(2.China National Coal Association,Beijing 100010,China)
关键词:
安全帽检测; YOLOX_s; 遮挡目标; 小目标; 密集目标
Keywords:
safety helmet detection;  improved YOLOX_s;  occlusion targets;  small targets;  dense targets
分类号:
TP391
DOI:
10.3969/j.issn.1001-4616.2023.02.014
文献标志码:
A
摘要:
在工业生产过程中,安全帽是生产工人重要的安全保护工具. 针对现有安全帽检测算法在复杂应用场景下对小目标、密集目标以及遮挡目标存在漏检、检测精度较低等问题,提出了一种基于YOLOX_s的改进算法. 首先,通过改进YOLOX_s算法的模型结构,在原有网络结构的基础上新设立了一个预测特征层,其尺寸为160×160,该预测特征层通过将高层语义信息和低层传递的位置信息进行有效融合来预测小目标; 其次,针对复杂的安全帽检测环境,将obj_loss的BCE_Loss改为Focal_Loss,即用Focal_Loss来训练obj分支来降低漏检; 最后,将CSP1_X中的残差块改为shuffleNet基本单元以缩减参数量. 改进后的算法mAP和recall分别提高了1.25%和2.32%,参数量缩减为3.61MB. 改进后的算法有效降低了复杂环境下安全帽的漏检率和提高了检测精度,对实际生产过程中保障企业和工人的生命财产安全起到了一定的促进作用.
Abstract:
In the process of industrial production, safety helmet is an important safety protection tool for production workers. Aiming at the problems of missing detection and low detection accuracy of the existing safety helmet detection algorithm for small targets, dense targets and occlusion targets in complex application scenarios, an improved algorithm based on YOLOX_s is proposed. Firstly, by improving the model structure of YOLOX_s algorithm, a new prediction feature layer with a size of 160×160 is established on the basis of the original network structure, the prediction feature layer can predict small targets by effectively fusing high-level semantic information and low-level location information. Secondly, for the complex helmet detection environment, BCE_Loss of obj_loss is changed to Focal_Loss, that is, Focal_Loss is used to train obj branch to reduce missed detection. Finally, residual block is changed in CSP1_X to the shuffleNet base unit to reduce parameters number, mAP and recall of the improved algorithm are increased by 1.25% and 2.32% respectively, and parameters number is reduced to 3.61MB. The improved algorithm effectively reduces the missed detection rate and improves the detection accuracy of safety helmet in complex environment, and plays a certain role in the actual production process to ensure the property and life safety of enterprises and workers.

参考文献/References:

[1]WEN C Y. The safety helmet detection technology and its application to the surveillance system[J]. Journal of forensic sciences,2004,49(4):770-780.
[2]DU S,SHEHATA M,BADAWY W. Hard hat detection in video sequences based on face features,motion and color information[C]//Proceedings of International Conference on Computer Research & Development,Shanghai,2011:25-29.
[3]WARANUSAST R,BUNDON N,TIMTONG V,et al. Machine vision techniques for motorcycle safety helmet detection[C]//Proceedings of 2013 28th International Conference on Image and Vision Computing New Zealand,Wellington,New Zealand,2013:35-40.
[4]JIAO L,ZHANG F,LIU F,et al. A survey of deep learning-based object detection[J]. IEEE Access,2019,7:128837-128868.
[5]REN S,HE K,GIRSHICK R B,et al. Faster R-CNN:towards real-time object detection with region proposal networks[J]. IEEE transactions on pattern analysis and machine intelligence,2017,39(6):1137-1149.
[6]GIRSHICK R B,DONAHUE J,DARRELL T,et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]//Proceedings of 2014 IEEE Conference on Computer Vision and Pattern Recognition,Columbus,USA,2014:580-587.
[7]GIRSHICK R B. Fast R-CNN[C]//Proceedings of 2015 IEEE International Conference on Computer Vision(ICCV),Santiago,Chile,2015:1440-1448.
[8]HU X,LI H,LI X,et al. MobileNet-SSD MicroScope using adaptive error correction algorithm:real-time detection of license plates on mobile devices[J]. Iet intelligent transport systems,2020,14(2):110-118.
[9]CHEN Z,WU K,LI Y,et al. SSD-MSN:an improved multi-scale object detection network based on SSD[J]. IEEE access,2019,7:80622-80632.
[10]REDMON J,DIVVALA S K,GIRSHICK R B,et al. You only look once:unified,real-time object detection[C]//Proceedings of 2016 IEEE Conference on Computer Vision and Pattern Recognition(CVPR),Las Vegas,NV,USA,2016:779-788.
[11]YIN Y,LI H,FU W. Faster-YOLO:an accurate and faster object detection method[J]. Digit signal process,2020,102:102756.
[12]SADYKOVA D,PERNEBAYEVA D,BAGHERI M,et al. IN-YOLO:real-time detection of outdoor high voltage insulators using UAV imaging[J]. IEEE transactions on power delivery,2020,35:1599-1601.
[13]NGUYEN D T,NGUYEN T N,KIM H,et al. A high-throughput and power-efficient FPGA implementation of YOLO CNN for object detection[J]. IEEE transactions on very large scale integration(VLSI)systems,2019,27(8):1861-1873.
[14]HUANG Z,WANG J. DC-SPP-YOLO:dense connection and spatial pyramid pooling based YOLO for object detection[J]. Information sciences,2020,522:241-258.
[15]REDMON J,FARHADI A. YOLO9000:better,faster,stronger[C]//Proceedings of 2017 IEEE Conference on Computer Vision and Pattern Recognition(CVPR),Honolulu,HI,USA,2017:6517-6525.
[16]REDMON J,FARHADI A. YOLOv3:an incremental improvement[EB/OL].(2018-04-08). https://doi.org/10.48550/arXiv.1804.02767.
[17]BOCHKOVSKIY A,WANG C Y,LIAO H Y M. YOLO v4:optimal speed and accuracy of object detection[EB/OL].(2020-04-23). https://doi.org/10.48550/arXiv.2004.10934.
[18]FELZENSZWALB P F,MCALLESTER D A,RAMANAN D. A discriminatively trained,multiscale,deformable part model[C]//Proceedings of 2008 IEEE Conference on Computer Vision and Pattern Recognition,Anchorage,AK,USA,2008:1-8.
[19]GE Z,LIU S,WANG F,et al. YOLOX:exceeding YOLO series in 2021[EB/OL].(2021-08-06). https://doi.org/10.48550/arXiv.2107.08430.
[20]GENG R,MA Y,HUANG W. An improved helmet detection method for YOLOv3 on an unbalanced dataset[C]//Proceedings of 2021 3rd International Conference on Advances in Computer Technology,Information Science and Communication(CTISC),Shanghai,China,2021:328-332.
[21]DENG B,LEI X,YE M. Safety helmet detection method based on YOLOv4[C]//Proceedings of 2020 16th International Conference on Computational Intelligence and Security(CIS),Nanning,China,2020:155-158.
[22]TAN S,LU G,JIANG Z,et al. Improved YOLOv5 network model and application in safety helmet detection[C]//Proceedings of 2021 IEEE International Conference on Intelligence and Safety for Robotics(ISR),Tokoname,Japan,2021:330-333.

备注/Memo

备注/Memo:
收稿日期:2022-10-16.
基金项目:国家重点研发计划项目(2020YFB1314001)、山西省基础研究计划项目(20210302124029)、山西省重点研发计划项目(202102030201012)、山西省重点研发计划项目(202102100401015).
通讯作者:杨乐,博士,副教授,研究方向:计算机视觉、图像处理和三维显示. E-mail:yangle@bupt.cn
更新日期/Last Update: 2023-06-15