基于YOLOv8改进的密集行人检测算法：MER-YOLO

摘要/Abstract

摘要： 在大型人员密集的场所中，人群异常聚集情况时有发生，对自动驾驶和大型公共场所人流量监控系统等应用场景中涉及到的密集行人检测技术带来了一定挑战，新一代的密集行人检测技术要求精确度更高、计算开销更小、检测速度更快以及部署更加方便等。针对上述需求，提出了一种基于YOLOv8改进的轻量级密集行人检测算法MER-YOLO，首先采用MobileViT作为主干网络，提升模型在总体上对行人聚集区域的特征提取能力；引入EMA注意力机制模块，对全局信息进行编码，通过维度交互来进一步聚合像素级特征，并结合160×160尺度的检测头加强小目标检测能力；使用排斥损失(Repulsion Loss)作为边界框损失函数，减少了人群密集情况下小目标行人的漏检误检的情况。实验结果表明，相较于YOLOv8n，MER-YOLO行人检测算法在Crowd Human数据集上mAP@0.5提升了4.5%，在WiderPerson数据集上mAP@0.5提升了2.1%，同时只有3.1×106的参数量和9.8 GFLOPs，满足低算力兼顾高精度的部署需求。

关键词: 目标检测, 行人检测, 轻量化, 注意力机制

Abstract: In large-scale crowded places, abnormal crowd gathering occurs from time to time, which brings certain challenges to the dense pedestrian detection technology involved in application scenarios such as autonomous driving and large-scale public place crowd monitoring systems. The new generation of dense pedestrian detection technology requires higher accuracy, smaller computing overhead, faster detection speed and more convenient deployment. In view of the above requirements, a lightweight dense pedestrian detection algorithm MER-YOLO based on YOLOv8 is proposed, which first uses MobileViT as the backbone network to improve the overall feature extraction ability of the model in pedestrian gathering areas. The EMA attention mechanism module is introduced to encode the global information, further aggregate pixel-level features through dimensional interaction, and strengthen the detection ability of small targets by combining the detection head with 160×160 scale. The use of Repulsion Loss as the bounding box loss function reduces the missed detection and misdetection of small target pedestrians under dense crowds. The experimental results show that compared with YOLOv8n, the mAP@0.5 of the MER-YOLO pedestrian detection algorithm is improved by 4.5% on the Crowd Human dataset and 2.1% on the WiderPerson dataset, while only 3.1×106 parameters and 9.8 GFLOPs, which meet the deployment requirements of low computing power and high precision.

Key words: object detection, pedestrian detection, light weight, attention mechanism

中图分类号:

王泽宇, 徐慧英, 朱信忠, 李琛, 刘子洋, 王子奕. 基于YOLOv8改进的密集行人检测算法：MER-YOLO[J]. 计算机工程与科学, 2024, 46(06): 1050-1062.

WANG Ze-yu, XU Hui-ying, ZHU Xin-zhong, LI Chen, LIU Zi-yang, WANG Zi-yi. An improved dense pedestrian detection algorithm based on YOLOv8: MER-YOLO[J]. Computer Engineering & Science, 2024, 46(06): 1050-1062.

参考文献［36］

［1］	Ge Z,Wang J F,Huang X,et al.LLa:Loss-aware label assignment for dense pedestrian detection［J］.Neurocomputing,2021,462:272-281.
［2］	Zhang Y, He H Y,Li J G,et al.Variational pedestrian detection ［C］∥Proc of 2021 IEEE Conference on Computer Vision and Pattern Recognition,2021:11622-11631.
［3］	魏润辰,何宁,尹晓杰.YOLO-Person：道路区域行人检测［J］.计算机工程与应用,2020,56(19):197-204.
	Wei Run-chen,He Ning,Yin Xiao-jie.YOLO-Person:Pedestrian detection in road areas［J］.Computer Engineering and Applications,2020,56(19):197-204.
［4］	邓杰,万旺根.基于改进YOLOv3的密集行人检测［J］.电子测量技术,2021,44(11):90-95.
	Deng Jie,Wan Wang-gen.Dense pedestrian detection based on improved YOLOv3［J］.Electronic Measurement Technol- ogy,2021,44(11):90-95.
［5］	张忠民,吴泽.基于改进YOLOv5的密集行人检测方法［J］.应用科技,2023,50(1):33-39.
	Zhang Zhong-min,Wu Ze.Dense pedestrian detection method based on improved YOLOv5［J］.Applied Science and Technology,2023,50(1):33-39.
［6］	Ding X H, Zhang X Y, Ma N N.RepVGG:Making VGG-style ConvNets great again［C］∥Proc of 2021 IEEE Confe- rence on Computer Vision and Pattern Recognition,2021:13728-13737.
［7］	Wang Q L, Wu B G, Zhu P F.ECA-Net:Efficient channel attention for deep convolutional neural networks［C］∥Proc of 2020 IEEE Conference on Computer Vision and Pattern Recognition,2020:11531-11539.
［8］	陈一潇,阿里甫·库尔班,林文龙，等.面向拥挤行人检测的CA-YOLOv5［J］.计算机工程与应用,2022,58(9):238-245.
	Chen Yi-xiao,Kurban A,Lin Wen-long,et al.CA-YOLOv5 for crowded pedestrian detection［J］.Computer Engineering and Applications,2022,58(9):238-245.
［9］	Hou Q B,Zhou D Q,Feng J S.Coordinate attention for efficient mobile network design［C］∥Proc of 2021 IEEE Confe- rence on Computer Vision and Pattern Recognition,2021:13713-13722.
［10］	Lin T Y,Dollár P,Girshick R,et al.Feature pyramid networks for object detection［C］∥Proc of 2017 IEEE Confe- rence on Computer Vision and Pattern Recognition,2017:936-944.
［11］	Zhang K,Xiong F,Sun P,et al.Double anchor R-CNN for human detection in a crowd［J］.arXiv:1909.09998,2019.
［12］	Girshick R,Donahue J,Darrell T,et al.Rich feature hierarchies for accurate object detection and semantic segmentation［C］∥Proc of 2014 IEEE Conference on Computer Vision and Pattern Recognition,2014:580-587.
［13］	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.
［14］	Mehta S, Rastegari M.MobileViT:Light-weight,general-purpose,and mobile-friendly vision transformer［J］.arXiv:2110.02178,2022.
［15］	Ouyang D L,He S,Zhang G Z,et al.Efficient multi-scale attention module with cross-spatial learning［C］∥Proc of 2023 IEEE International Conference on Acoustics,Speech and Signal Processing,2023:1-5.
［16］	Wang X L,Xiao T T,Jiang Y N,et al.Reploss loss:Detecting pedestrians in a crowd［C］∥Proc of 2018 IEEE Confe- rence on Computer Vision and Pattern Recognition,2018:7774-7783.
［17］	Shuai S, Zhao Z J,Li B X.CrowdHuman:A benchmark for detecting human in a crowd［J］ arXiv:1805.00123,2018.
［18］	Zhang S S,Benenson R,Schiele B.CityPersons:A diverse dataset for pedestrian detection［C］∥Proc of 2017 IEEE Conference on Computer Vision and Pattern Recognition,2017:3213-3221.
［19］	Zhang S F,Xie Y L,Wan J.WiderPerson:A diverse dataset for dense pedestrian detection in the wild ［J］. IEEE Tran- sactions on Multimedia,2019,22(2):380-393.
［20］	Hu J, Shen L, Sun G.Squeeze-and-excitation networks［C］∥Proc of 2018 IEEE Conference on Computer Vision and Pattern Recognition,2018:7132-7141.
［21］	Woo S,Park J,Lee J-Y,et al. CBAM:Convolutional block attention module［C］∥Proc of the 15th European Conference on Computer Vision,2018:3-19.
［22］	Li X, Hu X L,Yang J. Spatial group-wise enhance:Improving semantic feature learning in convolutional networks［J］.arXiv:1905.09646,2019.
［23］	Zhu X K，Lyu S C，Wang X，et al.TPH-YOLOv5：Improved YOLOv5 based on transformer prediction head for object detection on drone［C］∥Proc of 2021 IEEE Conference on Computer Vision and Pattern Recognition,2021：2778-2788.
［24］	Chen J R, Kao S H,Hao H, et al. Run, don’t walk:Chasing higher FLOPS for faster neural networks［C］∥Proc of 2023 IEEE Conference on Computer Vision and Pattern Recognition,2023:12021-12031.
［25］	Tang Y H, Han K, Guo J Y.GhostNetV2:Enhance cheap operation with long-range attention［C］∥Proc of 2020 IEEE Conference on Computer Vision and Pattern Recognition,2020:1577-1586.
［26］	Tan M,Le Q V.EfficientNet:Rethinking model scaling for convolutional neural networks［C］∥Proc of the 36th International Conference on Machine Learning,2019:2-16.
［27］	Liu Y,Shao Z,Teng Y,et al.NAM:Normalization-based attention module［J］. arXiv:2111.12419,2021.
［28］	Li Y,Yao T,Pan Y,et al.Contextual transformer networks for visual recognition［J］.IEEE Transactions on Pattern Analysis and Machine Intelligence,2022,45(2):1489-1500.
［29］	Yang L X, Zhang R Y, Li L D, et al.SimAM:A simple, parameter-free attention module for convolutional neural networks［C］∥Proc of the 38th International Conference on Machine Learning,2021:11863-11874.
［30］	Liu Y C,Shao Z R,Hoffmann N.Global attention mechanism:Retain information to enhance channel-spatial interactions［J］.arXiv:2112.05561,2021.
［31］	Liu W,Anguelov D,Erhan D,et al.SSD:Single shot multibox detector［C］∥Proc of the 14th European Conference on Computer Vision,2016:21-37.
［32］	Girshick R.Fast R-CNN［C］∥Proc of the 2015 IEEE International Conference on Computer Vision,2015:1440-1448.
［33］	Lü W Y，Zhao Y，Xu S L.DETRs beat YOLOs on real-time object detection［J］. arXiv:2304.08069,2023.
［34］	Li C Y,Li L L,Jiang H L,et al.YOLOv6:A single-stage object detection framework for industrial applications［J］. arXiv:2209.02976,2022.
［35］	Wang C Y,Bochkovskiy A,Liao H Y M.YOLOv7:Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors［C］∥Proc of 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition,2023:7464-7475.
［36］	Selvaraju R R,Cogswell M,Das A.Grad-CAM:Visual explanations from deep networks via gradient-based localization ［C］∥Proc of 2017 IEEE Conference on Computer Vision and Pattern Recognition,2017:618-626.
	作者简介:

[1]	邓翔宇, 裴浩媛, 盛迎. 基于网络融合的改进MobileViT人脸表情识别[J]. 计算机工程与科学, 2024, 46(06): 1072-1080.
[2]	马长林, 孙状. 基于实体知识的远程监督关系抽取[J]. 计算机工程与科学, 2024, 46(05): 945-950.
[3]	曹浩东, 汪海涛, 贺建峰. 融合序列局部信息的日期感知序列推荐算法[J]. 计算机工程与科学, 2024, 46(04): 734-742.
[4]	黄珍伟, 陈伟, 王文杰, 路锦通. 基于改进 RetinaNet网络的水下机器人目标检测与实验[J]. 计算机工程与科学, 2024, 46(02): 264-271.
[5]	王姗姗, 汪梦竹, 骆志刚. 局部判别损失无监督域适应方法[J]. 计算机工程与科学, 2024, 46(01): 132-141.
[6]	梁秀满, 周佳润, 杨若兰. LPD-YOLO：轻量级遮挡行人检测模型[J]. 计算机工程与科学, 2023, 45(12): 2197-2205.
[7]	焦佳辉, 马思远, 宋玉, 宋伟. 基于卷积注意力机制的双模态音乐流派分类模型MGTN[J]. 计算机工程与科学, 2023, 45(12): 2226-2236.
[8]	贾康, 李晓楠, 李冠宇. 一种基于自适应结构感知池化图匹配的图相似度计算模型[J]. 计算机工程与科学, 2023, 45(11): 1999-2007.
[9]	张千锟, 韩虎, 郝俊. 基于双注意力融合知识的方面级情感分类[J]. 计算机工程与科学, 2023, 45(10): 1866-1873.
[10]	尹春勇, 冯梦雪. 基于注意力机制的半监督日志异常检测方法[J]. 计算机工程与科学, 2023, 45(08): 1405-1415.
[11]	余子丞, 凌捷. 基于Transformer和多特征融合的DGA域名检测方法[J]. 计算机工程与科学, 2023, 45(08): 1416-1423.
[12]	吴栋梁, 刘知贵, . 基于轻量化YOLOX的电子元器件缺陷检测方法研究[J]. 计算机工程与科学, 2023, 45(08): 1463-1471.
[13]	曹玉东, 陈冬昊, 曹睿, 赵朗. 融合Mask R-CNN的在线多目标行人跟踪方法[J]. 计算机工程与科学, 2023, 45(07): 1216-1225.
[14]	王剑, 姜林, 王琳钦, 余正涛, 张松, 高盛祥, . 基于BiLSTM的低资源老挝语文本正则化任务[J]. 计算机工程与科学, 2023, 45(07): 1292-1299.
[15]	李校林, 王复港, 张鹏飞, 张琳玉, . 基于多尺度特征提取的YOLOv5s算法优化[J]. 计算机工程与科学, 2023, 45(06): 1054-1062.