复杂场景下的改进YOLOv8-n密集行人检测模型

doi:10.19678/j.issn.1000-3428.0070531

摘要/Abstract

摘要： 密集行人检测是大型公共场所人流监控系统发展的关键问题之一。针对密集行人检测场景中由于人群遮挡导致的小目标检测困难以及模型轻量化的部署需求,提出一种改进YOLOv8-n的密集行人检测模型CAD-YOLO。嵌入CGDown下采样模块,通过高效的上下文信息提取机制,有效缓解传统目标检测器在处理密集场景时上下文特征易丢失的问题,显著增强模型对密集行人特征的捕获能力以及对小目标的聚焦性能。设计一种BiFPN-Adaptive结构并重构颈部网络,通过自适应融合不同尺度的特征信息,使模型在提取被遮挡行人及中小型目标行人特征时更加精准,同时大幅减少模型的参数量与计算成本。引入动态检测头Dyhead,结合新增的160×160尺度的小目标检测层,使模型能够更加精确地捕获密集小目标区域的细微特征,从而有效缓解遮挡场景中的漏检问题。实验结果显示,相较于YOLOv8-n,CAD-YOLO在Crowd Human和WiderPerson数据集上检测精度分别提升5.1和2.1百分点。同时,CAD-YOLO的参数量为2.9×10⁶,模型计算量为12.3×10⁹,能够满足在边缘设备或移动设备上部署时对低功耗、高精度的要求。

关键词: 密集行人检测, YOLOv8, CGDown下采样模块, BiFPN, 小目标检测头

Abstract: Improving dense pedestrian detection performance is a key issue in the development of crowd monitoring systems for large public places. Existing systems present difficulties in detecting small targets due to crowd occlusion in dense pedestrian detection scenarios and do not adequately meet the deployment requirements for lightweight models. To address these issues, this paper proposes an improved dense pedestrian detection model, CAD-YOLO, based on YOLOv8-n. Introducing a CGDown downsampling module and an efficient context information extraction mechanism helps in effectively mitigating the problem of context feature loss, which is encountered by traditional object detectors when dealing with dense scenes. These inclusions significantly enhances the model's ability to capture dense pedestrian features and focus on small targets. A BiFPN-Adaptive structure is designed, and the neck network is reconstructed, allowing the model to more accurately extract the features of occluded pedestrians as well as small- and medium-sized pedestrians by adaptively fusing feature information at different scales. Further, the introduction of the dynamic detection head Dyhead, combined with a newly added small-target detection layer with a size of 160×160, enables the model to more precisely capture the subtle features of dense small target areas, thereby effectively alleviating the problem of missed detections in occlusion scenarios. Experimental results show that, compared to YOLOv8-n, CAD-YOLO achieves a detection accuracy improvement of 5.1 and 2.1 percentage points on the Crowd Human and WiderPerson datasets, respectively. Additionally, CAD-YOLO has a parameter count of 2.9×10⁶ and model computation of 12.3×10⁹; thus, it satisfies the requirements of low power consumption and high accuracy for deployment on edge devices or mobile devices.

Key words: dense pedestrian detection, YOLOv8, CGDown subsampling module, BiFPN, small target detection head

中图分类号:

TP391.4

陈海秀, 陈子昂, 房威志, 卢海涛, 黄仔洁, 成荣. 复杂场景下的改进YOLOv8-n密集行人检测模型[J]. 计算机工程, 2026, 52(7): 220-232.

CHEN Haixiu, CHEN Ziang, FANG Weizhi, LU Haitao, HUANG Zijie, CHENG Rong. Improved Model Based on YOLOv8-n for Dense Pedestrian Detection in Complex Scenes[J]. Computer Engineering, 2026, 52(7): 220-232.

https://www.ecice06.com/CN/Y2026/V52/I7/220

参考文献

[1] LI N F, BAI X L, SHEN X F, et al. Dense pedestrian detection based on GR-YOLO[J]. Sensors, 2024, 24(14): 4747.
[2] WANG J M, HUANG K, PI J Y. RUP2S-YOLO: an improved YOLOv8-based algorithm for dense pedestrian detection[C]//Proceedings of the 5th International Seminar on Artificial Intelligence, Networking and Information Technology (AINIT). Washington D.C.,USA:IEEE Press,2024: 667-671.
[3] BHAGYA C, SHYNA A. An overview of deep learning based object detection techniques[C]//Proceedings of the 1st International Conference on Innovations in Information and Communication Technology (ICIICT). Washington D.C.,USA:IEEE Press,2019: 1-6.
[4] LIANG Y J, CUI X P, XU X H, et al. A review on deep learning techniques applied to object detection[C]//Proceedings of the 7th International Conference on Information Science and Control Engineering (ICISCE). Washington D.C.,USA:IEEE Press,2021: 120-124.
[5] GE N M, YONG Y. A survey of vision-based object detection[C]//Proceedings of the International Conference on Image Processing, Computer Vision and Machine Learning (ICICML). Washington D.C.,USA:IEEE Press, 2023: 240-244.
[6] DWIVEDI U, JOSHI K, SHUKLA S K, et al. An overview of moving object detection using YOLO deep learning models[C]//Proceedings of the 2nd International Conference on Disruptive Technologies (ICDT). Washington D.C.,USA:IEEE Press,2024: 1014-1020.
[7] 张阳婷, 黄德启, 王东伟, 等. 基于深度学习的目标检测算法研究与应用综述[J]. 计算机工程与应用, 2023, 59(18): 1-13. ZHANG Y T, HUANG D Q, WANG D W, et al. Review on research and application of deep learning-based target detection algorithms[J]. Computer Engineering and Applications, 2023, 59(18): 1-13. (in Chinese)
[8] 徐彦威, 李军, 董元方, 等. YOLO系列目标检测算法综述[J]. 计算机科学与探索, 2024, 18(9): 2221-2238. XU Y W, LI J, DONG Y F, et al. Survey of development of YOLO object detection algorithms[J]. Journal of Frontiers of Computer Science and Technology, 2024, 18(9): 2221-2238. (in Chinese)
[9] ZHANG W C, FU C, XIE H Y, et al. Global context aware RCNN for object detection[J]. Neural Computing and Applications, 2021, 33(18): 11627-11639.
[10] ARORA N, KUMAR Y, KARKRA R, et al. Automatic vehicle detection system in different environment conditions using Fast R-CNN[J]. Multimedia Tools and Applications, 2022, 81(13): 18715-18735.
[11] LI X M, XIE Z J, DENG X, et al. Traffic sign detection based on improved Faster R-CNN for autonomous driving[J]. The Journal of Supercomputing, 2022, 78(6): 7982-8002.
[12] GAWANDE U, HAJARI K, GOLHAR Y. SIRA: scale illumination rotation affine invariant Mask R-CNN for pedestrian detection[J]. Applied Intelligence, 2022, 52(9): 10398-10416.
[13] REN S Q, HE K M, GIRSHICK R, et al. Faster R-CNN: towards real-time object detection with region proposal networks[C]//Proceedings of the IEEE Transactions on Pattern Analysis and Machine Intelligence. Washington D.C.,USA:IEEE Press,2016: 1137-1149.
[14] 王泽宇, 徐慧英, 朱信忠, 等. 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(6): 1050-1062. WANG Z Y, XU H Y, ZHU X Z, et al. An improved dense pedestrian detection algorithm based on YOLOv8: MER-YOLO[J]. Computer Engineering and Science, 2024, 46(6): 1050-1062. (in Chinese)
[15] HU W C, HU Q, PI J Y, et al. Dense pedestrian detection algorithm based on multiscale feature fusion in YOLOv8[C]//Proceedings of the 5th International Conference on Computer Vision, Image and Deep Learning (CVIDL). Washington D.C.,USA:IEEE Press,2024: 1362-1366.
[16] 谢永明, 王红蕾. 复杂背景下远距离及小尺寸行人检测改进算法[J]. 计算机工程与设计, 2021, 42(5): 1323-1330. XIE Y M, WANG H L. Improved algorithm for long-distance and small-size pedestrian detection in complex backgrounds[J]. Computer Engineering and Design, 2021, 42(5): 1323-1330. (in Chinese)
[17] WANG J P, ZHANG X Y, GAO G H, et al. OP Mask R-CNN: an advanced Mask R-CNN network for cattle individual recognition on large farms[C]//Proceedings of the International Conference on Networking and Network Applications (NaNA). Washington D.C.,USA:IEEE Press,2023: 601-606.
[18] YU Y B, YAN S C, HAO X P. Pedestrian detection based on improved YOLOv3 network[C]//Proceedings of the IEEE International Conference on Control, Electronics and Computer Technology (ICCECT). Washington D.C.,USA:IEEE Press,2023: 297-301.
[19] CHEN J T, WEI Y, ZHOU Y. Dense crowd detection algorithm for YOLOv5 based on coordinate attention mechanism[C]//Proceedings of the 2nd International Conference on Algorithms, High Performance Computing and Artificial Intelligence (AHPCAI). Washington D.C.,USA:IEEE Press,2023: 187-190.
[20] 黄昆, 齐肇建, 王娟敏, 等. 基于改进YOLOv8的密集行人检测模型[J]. 计算机工程, 2025, 51(5): 133-142. HUANG K, QI Z J, WANG J M, et al. Aggregation pedestrian detection model based on improved YOLOv8[J]. Computer Engineering, 2025, 51(5): 133-142. (in Chinese)
[21] VARGHESE R, M S. YOLOv8: a novel object detection algorithm with enhanced performance and robustness[C]//Proceedings of the International Conference on Advances in Data Engineering and Intelligent Computing Systems (ADICS). Washington D.C.,USA:IEEE Press,2024: 1-6.
[22] WU T Y, TANG S, ZHANG R, et al. CGNet: a light-weight context guided network for semantic segmentation[C]//Proceedings of the IEEE Transactions on Image Processing. Washington D.C.,USA:IEEE Press,2020: 1169-1179.
[23] LIU R Y. Pedestrian detection based on SENet with attention mechanism[C]//Proceedings of the 7th IEEE Information Technology and Mechatronics Engineering Conference (ITOEC). Washington D.C.,USA:IEEE Press,2023: 616-619.
[24] ZHANG Z, ZHANG X, TIAN H P, et al. Application of improved YOLOv7 algorithm in pedestrian detection[C]//Proceedings of the 5th International Seminar on Artificial Intelligence, Networking and Information Technology (AINIT). Washington D.C.,USA:IEEE Press,2024: 956-961.
[25] LIU W, ANGUELOV D, ERHAN D, et al. SSD: single shot MultiBox detector[EB/OL].[2024-08-05]. https://arxiv.org/abs/1512.02325.
[26] GIRSHICK R. Fast R-CNN[C]//Proceedings of the IEEE International Conference on Computer Vision (ICCV). Washington D.C.,USA:IEEE Press,2016: 1440-1448.
[27] WEI J G, QU Y, GONG M H, et al. VE-YOLOv6: a lightweight small target detection algorithm[C]//Proceedings of the 4th International Conference on Neural Networks, Information and Communication Engineering (NNICE). Washington D.C.,USA:IEEE Press,2024: 873-876.
[28] YANG X H, ZHOU C G, YU J J, et al. A new strategy for bird detection in coastal wetlands based on improved YOLOv63.0[C]//Proceedings of the 2nd International Conference on Artificial Intelligence and Intelligent Information Processing (AIIIP). Washington D.C.,USA:IEEE Press,2024: 293-296.
[29] YANG Y N, WANG X M. An improved YOLOv7-tiny-based lightweight network for the identification of fish species[C]//Proceedings of the 5th International Conference on Robotics and Computer Vision (ICRCV). Washington D.C.,USA:IEEE Press,2023: 188-192.
[30] CAI H, LI J Y, HU M Y, et al. EfficientViT: lightweight multi-scale attention for high-resolution dense prediction[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV). Washington D.C.,USA:IEEE Press,2024: 17256-17267.
[31] WANG T T, LU X Q. Face forgery detection algorithm based on improved MobileViT network[C]//Proceedings of the 8th International Conference on Intelligent Computing and Signal Processing (ICSP). Washington D.C.,USA:IEEE Press,2023: 1396-1400.
[32] LIU H Y, ZHANG Y Y, LIU S G, et al. UAV wheat rust detection based on FasterNet-YOLOv8[C]//Proceedings of the IEEE International Conference on Robotics and Biomimetics (ROBIO). Washington D.C.,USA:IEEE Press,2023: 1-6.
[33] YANG G Y, LEI J, ZHU Z K, et al. AFPN: asymptotic feature pyramid network for object detection[C]//Proceedings of the IEEE International Conference on Systems, Man, and Cybernetics (SMC). Washington D.C.,USA:IEEE Press,2024: 2184-2189.
[34] ANTONY VIGIL M S, BARHANPURKAR M M, ANAND N R, et al. EYE SPY face detection and identification using YOLO[C]//Proceedings of the International Conference on Smart Systems and Inventive Technology (ICSSIT). Washington D.C.,USA:IEEE Press,2020: 105-110.
[35] SELVARAJU R R, COGSWELL M, DAS A, et al. Grad-CAM: visual explanations from deep networks via gradient-based localization[J]. International Journal of Computer Vision, 2020, 128(2): 336-359.

选择文件类型/文献管理软件名称

选择包含的内容