| 367 | 3 | 128 |
| 下载次数 | 被引频次 | 阅读次数 |
为提高电动自行车骑乘人员安全头盔佩戴水平，提出一种基于深度学习的安全头盔佩戴行为检测方法。基于EfficientNet目标检测框架，重新设计了主干网中特征图的特征选取层，并提出像素级缩放（pixel-level scaling,PLS）模块，构建了一种新的用于电动自行车骑乘人员安全头盔佩戴行为检测的PLS-Det模型。该模型解决了深度卷积神经网络执行检测任务时视频检测图像中的小目标（远处的电动自行车）、被遮挡的车辆和骑乘人员等容易导致目标丢失的问题，并能适应复杂的电动自行车交通流场景。根据江苏省南通市区典型城市道路，选取不同视角、时间段、天气状况下的电动自行车交通流视频图像数据（包含正样本5 408个和负样本7 156个）训练优化检测模型。通过消融实验和人工检测结果，对比分析了EfficientDet-d0、EfficientDet-Optimize和PLS-Det模型的检测性能。实验结果表明：提出的PLS-Det检测模型通过重新选择特征图层和引入PLS模块，在保证计算效率稳定的同时，能显著丰富小目标及被遮挡目标的特征，检测精确度达95.8%，可以满足电动自行车骑乘人员安全头盔佩戴行为的检测精度要求。
Abstract：To improve the safety helmet wearing rate of electrical bike(e-bike) riders, a safety helmet wearing behavior detection method based on deep learning was proposed. In order to solve the problem of small targets(distant ebikes), covered e-bikes and riders in video detection image which can easily lead to target loss when deep convolutional neural network performs detection task, the feature selection layer of feature graph in the backbone network was redesigned based on EfficientNet target detection framework. Pixel-level scaling(PLS) module was proposed to construct a new PLS-Det model for safety helmet wearing behavior detection, which can adapt to complex e-bike flow scenarios. Take typical urban roads in Nantong city, Jiangsu Province for example, the detection model was trained and optimized using the video image data of e-bike flow in different viewing angle, time periods and under different weather conditions(including 5 408 positive samples and 7 156 negative samples). The detection performances of EfficientDet-d0, EfficientDet-Optimize and PLS-Det models were compared with the results of ablation experiment and artificial detection. The experimental results show that the proposed PLS-Det model can significantly enrich the features of small targets and blocked targets by re-selecting feature layers and adding PLS module, while ensuring the stability of computational efficiency. The detection accuracy of this detection model is 95.8%, which can meet the accuracy requirements of safety helmet wearing behavior of e-bike riders.
[1] GIRSHIC K R, DONAHUE J, DARRELL T, et al. Region-based convolutional networks for accurate object detection and segmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2016, 38(1):142-158.
[2] GIRSHICK R. Fast R-CNN[C]//Proceedings of the 2015IEEE International Conference on Computer Vision(ICCV), December 7-13, 2015, Santiago, Chile. New York:IEEE Xplore, 2016:1440-1448.
[3] REN S Q, HE K M, GIRSHICK R, 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.
[4] REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once:unified, real-time object detection[C]//Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition(CVPR), June 27-30, 2016, Las Vegas, NV, USA. New York:IEEE Xplore, 2016:779-788.
[5]王思雨.基于YOLO的交通目标检测方法[D].北京：华北电力大学，2019.WANG S Y. Traffic target detection method based on YOLO[D]. Beijing:North China Electric Power University,2019.(in Chinese)
[6]邵叶秦，周昆阳，郑泽斌，等.基于改进的轻量级YOLOv3的交通信号灯检测与识别[J].南通大学学报（自然科学版），2021, 20(3):34-40.SHAO Y Q, ZHOU K Y, ZHENG Z B, et al. Traffic light detection and recognition based on improved lightweight YOLOv3[J]. Journal of Nantong U niversity(Natural Science Edition), 2021, 20(3):34-40.(in Chinese)
[7] REDMON J, FARHADI A. YOLO9000:better, faster,stronger[C]//Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition(CVPR), July 21-26, 2017, Honolulu, HI, USA. New York:IEEE Xplore,2017:6517-6525.
[8] REDMON J, FARHADI A. YOLOv3:an in cremental improvement[EB/OL].(2018-04-08)[2021-03-13]. https：//arxiv.org/abs/1804.02767.
[9] LIU W, ANGUELOV D, ERHAN D, et al. SSD:single shot MultiBox detector[C]//Proceedings of the 14th European Conference on Computer Vision(ECCV), October 11-14, 2016, Amsterdam, The Netherlands. Berlin:Springer,2016:21-37.
[10]谢娟英，鲁银圆，孔维轩，等.基于改进RetinaNet的自然环境中蝴蝶种类识别[J].计算机研究与发展，2021,58(8):1686-1704.XIE J Y, LU Y Y, KONG W X, et al. Butterfly species identification from natural environment based on improved RetinaNet[J]. Journal of Computer Research and Development, 2021, 58(8):1686-1704.(in Chinese)
[11] DALAL N, TRIGGS B. Histograms of ori ented gradients for human detection[C]//Proceedings of the 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition(CVPR), June 20-25, 2005, San Diego,CA, USA. New York:IEEE Xplore, 2005:886-893.
[12]赵才荣，齐鼎，窦曙光，等.智能视频监控关键技术：行人再识别研究综述[J].中国科学：信息科学，2021, 51(12):1979-2015.ZHAO C R, QI D, DOU S G, et al. Key technology for intelligent video surveillance:a review of person re-identification[J]. Scientia Sinica(Informationis), 2021, 51(12):1979-2015.(in Chinese)
[13]杨永胜，邓淼磊，李磊，等.基于深度学习的行人重识别综述[J].计算机工程与应用，2022, 58(9):51-66.YANG Y S, DENG M L, LI L, et al. Overview of pedestrian re-identification based on deep learning[J]. Computer Engineering and Applications, 2022, 58(9):51-66.(in Chinese)
[14]王梦缘.基于深度学习的行人及非机动车违规检测系统[D].大连：大连理工大学，2021.WANG M Y. Pedestrian and non-motor vehicle violation detection system based on deep learning[D]. Dalian:Dalian University of Technology, 2021.(in Chinese)
[15]方建成.智能交通系统中交通流量控制及小目标检测识别[D].成都：电子科技大学，2021.FANG J C. Traffic flow control and small target detection and recognition in intelligent transportation system[D]. Chengdu:University of Electronic Science and Technology of China, 2021.(in Chinese)
[16]胡恬，王新刚.基于小波变换和神经网络的安全帽识别系统分析与设计[J].软件导刊，2006, 5(23):37-39.HU T, WANG X G. Analysis and design of helmet recognition system based on wavelet transform and neural network[J]. Software Guide, 2006, 5(23):37-39.(in Chinese)
[17] FANG Q, LI H, LUO X C, et al. Detecting non-hardhatuse by a deep learning method from far-field surveillance videos[J]. Automation in Construction, 2018, 85:1-9.
[18]杨莉琼，蔡利强，古松.基于机器学习方法的安全帽佩戴行为检测[J].中国安全生产科学技术，2019, 15(10):152-157.YANG L Q, CAI L Q, GU S. Detection on wearing behavior of safety helmet based on machine learning method[J].Journal of Safety Science and Technology, 2019, 15(10):152-157.(in Chinese)
[19]张明媛，曹志颖，赵雪峰，等.基于深度学习的建筑工人安全帽佩戴识别研究[J].安全与环境学报，2019, 19(2):535-541.ZHANG M Y, CAO Z Y, ZHAO X F, et al. On the identification of the safety helmet wearing manners for the construction company workers based on the deep learning theory[J]. Journal of Safety and Environment, 2019, 19(2):535-541.(in Chinese)
[20]吴冬梅，王慧，李佳.基于改进Faster RCNN的安全帽检测及身份识别[J].信息技术与信息化，2020(1):17-20.WU D M, WANG H, LI J. Safety helmet detection and identification based on improved faster RCNN[J]. Information Technology and Informatization, 2020(1):17-20.(in Chinese)
[21]王兵，李文璟，唐欢.改进YOLOv3算法及其在安全帽检测中的应用[J].计算机工程与应用，2020, 56(9):33-40.WANG B, LI W J, TANG H. Improved YOLOv3 algorithm and its application in helmet detection[J]. Computer Engineering and Applications, 2020, 56(9):33-40.(in Chinese)
[22]李华，王岩彬，益朋，等.基于深度学习的复杂作业场景下安全帽识别研究[J].中国安全生产科学技术，2021,17(1):175-181.LI H, WANG Y B, YI P, et al. Research on recognition of safety helmets under complex operation scenes based on deep learning[J]. Journal of Safety Science and Technology, 2021, 17(1):175-181.(in Chinese)
[23] DAHIYA K, SINGH D, MOHAN C K. Automatic detection of bike-riders without helmet using surveillance videos in real-time[C]//Proceedings of the 2016 International Joint Conference on Neural Networks(IJCNN), July 24-29, 2016,Vancouver, BC, Canada. New York:IEEE Xplore, 2016:3046-3051.
[24] VISHNU C, SINGH D, MOHAN C K, et al. Detection of motorcyclists without helmet in videos using convolutional neural network[C]//Proceedings of the 2017 International Joint Conference on Neural Networks(IJCNN), May 14-19,2017, Anchorage, AK, USA. New York:IEEE Xplore, 2017:3036-3041.
[25] SIEBERT F W, LIN H H. Detecting motorcycle helmet use with deep learning[J]. Accident Analysis and Prevention,2020, 134:105319.
[26]刘琛，王江涛.基于RFB-Net的摩托车驾驶人头盔检测研究[J].太原科技大学学报，2021, 42(6):496-500.LIU C, WANG J T. Research on motorcycle driver′s helmet detection based on RFB net[J]. Journal of Taiyuan University of Science and Technology, 2021, 42(6):496-500.(in Chinese)
[27]李彦辰，张小俊，张明路，等.基于改进Efficientdet的自动驾驶场景目标检测[J].计算机工程与应用，2022,58(6):183-191.LI Y C, ZHANG X J, ZHANG M L, et al. Object detection in autonomous driving scene based on improved efficientdet[J]. Computer Engineering and Applications, 2022,58(6):183-191.(in Chinese)
[28] HAN K, WANG Y H, TIAN Q, et al. GhostNet:more features from cheap operations[C]//Proceedings of the 2020IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR), June 13-19, 2020, Seattle, WA, USA.New York:IEEE Xplore, 2020:1577-1586.
[29] TAN M, LE Q V. Efficientnet:rethinking model scaling for convolutional neural networks[C]//Proceedings of the 36th International Conference on Machine Learning, June 9-15, 2019, Long Beach, California. Long Beach:MLR Press,2019:6105-6114.
[30]张永强.基于深度卷积神经网络的真实场景物体检测算法研究[D].哈尔滨：哈尔滨工业大学，2020.ZHANG Y Q. Research on object detection in the wild based on deep convolutional neural network[D]. Harbin:Harbin Institute of Technology, 2020.(in Chinese)
基本信息:
DOI：
中图分类号:TP391.41;U492.8
引用信息:
[1]汤天培,龚昊,李洪亮等.电动自行车骑乘人员安全头盔佩戴行为检测[J],2023,22(02):12-19.
基金信息:
国家自然科学基金青年基金项目（72101128）;; 中国博士后科学基金面上项目（2023M730560）;; 江苏省社会科学基金项目（20GLC015）;; 南通市科技计划民生项目（MS22022093）