[1]周力,代亮,杨杰,等.基于CRD-YOLOv5s的交通场景目标检测[J].长安大学学报(自然科学版),2025,45(5):186-199.[doi:10.19721/j.cnki.1671-8879.2025.05.016]
　ZHOU Li,DAI Liang,YANG Jie,et al.CRD-YOLOv5s-based object detection in traffic scenario[J].Journal of Chang’an University (Natural Science Edition),2025,45(5):186-199.[doi:10.19721/j.cnki.1671-8879.2025.05.016]
点击复制

基于CRD-YOLOv5s的交通场景目标检测()
分享到：

长安大学学报（自然科学版）[ISSN:1006-6977/CN:61-1281/TN]

卷:
第45卷
期数:
2025年5期
页码:
186-199
栏目:
交通工程
出版日期:
2025-09-30

文章信息/Info

Title:
CRD-YOLOv5s-based object detection in traffic scenario
文章编号:
1671-8879(2025)05-0186-14
作者:
周力; 代亮; 杨杰; 凌之楷
(长安大学 电子与控制工程学院,陕西 西安 710064)
Author(s):
ZHOU Li;  DAI Liang;  YANG Jie;  LING Zhi-kai
(School of Electronics and Control Engineering, Chang'an University, Xi'an 710064, Shaanxi, China)
关键词:
交通场景目标检测;  YOLOv5s;  卷积注意力模块;  感受野模块;  解耦结构;  平滑交并比
Keywords:
object detection in traffic scenario;  YOLOv5s;  convolutional block attention module;  receptive field block;  decoupled structure;  smooth intersection over union
分类号:
U495
DOI:
10.19721/j.cnki.1671-8879.2025.05.016
文献标志码:
A
摘要:
为解决复杂道路环境交通场景中多尺度目标漏检与误检的问题,在对YOLOv5s模型改进的基础上提出了CRD-YOLOv5s模型; 针对小目标检测中易漏检的问题,在骨干网络中引入卷积注意力模块(CBAM),以强化小目标的特征检测,并采用感受野模块(RFB)替换空间金字塔池化模块,以提升其多目标检测能力; 用解耦结构(DS)替代原有头部网络耦合结构,以提高特征表示的充分性和信息传递效率; 引入平滑交并比(SIoU)损失函数来提升网络结构训练收敛速度和检测准确率,采用准确率、召回率和平均精度等参数评估模型性能,并将试验数据库分为检测行人、汽车、交通灯和交通标志4类。研究结果表明:在汽车类别试验中,CRD-YOLOv5s的准确率比YOLOv5s高1.1%,平均精度比YOLOv5s高0.7%,在多类检测中,CRD-YOLOv5s的平均精度为53.3%,比YOLOv5s高1.3%; CRD-YOLOv5s在多目标检测、边缘目标检测、复杂环境下误检等方面均优于YOLOv5s,在驾驶人视角下的检测精度显著提高,可有效减少漏检与误检,可广泛应用于交通场景目标检测,在保证单一检测竞争性的同时在多目标和复杂环境下具有优势。研究成果可为实际交通检测项目提供有效数学模型,为智慧交通检测和交通安全提供可靠的技术支撑。
Abstract:
To address the issues of missed and false detections for multi-scale targets in traffic scenarios of complex road environments, an improved model based on YOLOv5s, named CRD-YOLOv5s was proposed. To tackle the problem of missed detections for small targets, a convolutional block attention module(CBAM)was introduced into the backbone network to enhance the feature detection for small objects, and the spatial pyramid pooling module was replaced with a receptive field block(RFB)to strengthen the capability to detect multi objects. The coupled structure of the original head network was substituted with a decoupled structure(DS)to enhance the sufficiency of feature representation and the efficiency of information propagation. The smooth intersection over union(SIoU)loss function was adopted to improve the convergence speed and detection accuracy of the network structure during training. The precision, recall and mean average precision were used to evaluate the model performance, and the test dataset was categorized into pedestrians, cars, traffic lights, and traffic signs. The research results demonstrate that in the car category test, the accuracy of CRD-YOLOv5s is 1.1% higher than YOLOv5s, and the average precision is 0.7% higher than YOLOv5s. For the multi-class detection, the mean average precision of CRD-YOLOv5s reaches 53.3%, exceeding YOLOv5s by 1.3%. CRD-YOLOv5s outperforms the YOLOv5s in multi-object detection, edge object detection and false detection reduction in complex environments. It can significantly improve the detection accuracy from the driver's perspective, effectively reduce the missed and false detections, and is well-suited for object detections in traffic scenarios. While maintaining the competitive performance in single-object detection, it exhibits superior capability in multi-object and complex environment detection. The research results provide an effective mathematical model for practical traffic detection applications and offer a reliable technical support for intelligent traffic detection and traffic safety.7 tabs, 14 figs, 33 refs.

参考文献/References:

[1] 肖雨晴,杨慧敏.目标检测算法在交通场景中应用综述[J].计算机工程与应用,2021,57(6):30-41.
XIAO Yu-qing, YANG Hui-min. Research on application of object detection algorithm in traffic scene[J]. Computer Engineering and Applications, 2021, 57(6): 30-41.
[2]张凯祥,朱 明.基于YOLOv5的多任务自动驾驶环境感知算法[J].计算机系统应用,2022,31(9):226-232.
ZHANG Kai-xiang, ZHU Ming. Environmental perception algorithm for multi-task autonomous driving based on YOLOv5[J]. Computer Systems and Applications, 2022, 31(9): 226-232.
[3]赵 萍,李 欣,朱少武.基于时空图注意力神经网络的交通道路拥塞和异常预测[J].科学技术与工程,2022,22(3):1271-1278.
ZHAO Ping, LI Xin, ZHU Shao-wu. Traffic road congestion and anomaly prediction based on spatio-temporal graph attention neural networks[J]. Science Technology and Engineering, 2022, 22(3): 1271-1278.
[4]何朋朋.基于深度学习的交通场景多目标检测与分类研究[D].西安:长安大学,2018.
HE Peng-peng. Multi-object detection and classification based on deep learning in traffic scene[D]. Xi'an: Chang'an University, 2018.
[5]杨 康.基于视频的车辆检测理论与方法研究[D].西安:长安大学,2013.
YANG Kang. Research on theory and method of vehicle detection based on images[D]. Xi'an: Chang'an University, 2013.
[6]李永豪.基于YOLOv5s的车辆检测改进算法[D].西安:长安大学,2022.
LI Yong-hao. Improved vehicle detection algorithm based on YOLOv5s[D]. Xi'an: Chang'an University, 2022.
[7]LIU W, ANGUELOV D, ERHAN D, et al. SSD: Single shot multibox detector[C]//LEIBE B, MATAS J, SEBE N, et al. Computer Vision-ECCV 2016, Part Ⅰ. Berlin: Springer, 2016: 21-37.
[8]REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: unified, real-time object detection[C]//IEEE. 2016 IEEE Conference on Computer Vision and Pattern Recognition(CVPR). New York: IEEE, 2016: 779-788.
[9]GIRSHICK R, DONAHUE J, DARRELL T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]//IEEE. 2014 IEEE Conference on Computer Vision and Pattern Recognition(CVPR). New York: IEEE, 2014: 580-587.
[10]HE K M, ZHANG X Y, REN S Q, et al. Spatial pyramid pooling in deep convolutional networks for visual recognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 37(9): 1904-1916.
[11]GIRSHICK R. Fast R-CNN[C]//IEEE. 2015 IEEE International Conference on Computer Vision(ICCV). New York: IEEE, 2015: 1440-1448.
[12]REN S Q, HE K M, GIRSHICK R, et al. Faster R-CNN: Towards real-time object detection with region proposal networks[C]//IEEE. IEEE Transactions on Pattern Analysis and Machine Intelligence. New York: IEEE, 2017: 1137-1149.
[13]REDMON J, FARHADI A. YOLO9000: Better, faster, stronger[C]//IEEE. 2017 IEEE Conference on Computer Vision and Pattern Recognition(CVPR). New York: IEEE, 2017: 6517-6525.
[14]REDMON J, FARHADI A. YOLOv3: An incremental improvement[J]. arXiv, 2018, https://arxiv.org/abs/1804.02767.
[15]BOCHKOVSKIY A, WANG C Y, LIAO H Y M. YOLOv4: Optimal speed and accuracy of object detection[J]. arXiv, 2020, https://arxiv.org/abs/2004.10934.
[16]WANG A, CHEN H, LIU L, et al. YOLOv10: Real-time end-to-end object detection[J]. Advances in Neural Information Processing Systems, 2024, 37: 107984-108011.
[17]KHANAM R, HUSSAIN M. YOLOv11: An overview of the key architectural enhancements[J]. arXiv, 2024, https://arxiv.org/abs/2410.17725.
[18]肖雨晴,杨慧敏.基于改进YOLOv3算法的交通场景目标检测[J].森林工程,2022,38(6):164-171.
XIAO Yu-qing, YANG Hui-min. Object detection based on improved YOLOv3 algorithm in traffic scenes[J]. Forest Engineering, 2022, 38(6): 164-171.
[19]LI Y, LYU C. SS-YOLO: An object detection algorithm based on YOLOv3 and ShuffleNet[C]//XU B, MOU K.Proceedings of 2020 IEEE 4th Information Technology, Networking, Electronic and Automation Control Conference(ITNEC 2020). New York: IEEE, 2020: 769-772.
[20]张丽莹,庞春江,王新颖,等.基于改进YOLOv3的多尺度目标检测算法[J].计算机应用,2022,42(8):2423-2431.
ZHANG Li-ying, PANG Chun-jiang, WANG Xin-ying, et al. Multi-scale object detection algorithm based on improved YOLOv3[J]. Journal of Computer Applications, 2022, 42(8): 2423-2431.
[21]宦 海,陈逸飞,张 琳,等.一种改进的BR-YOLOv3目标检测网络[J].计算机工程,2021,47(10):186-193.
HUAN Hai, CHEN Yi-fei, ZHANG Lin, et al. An improved BR-YOLOv3 object detection network[J]. Computer Engineering, 2021, 47(10): 186-193.
[22]朱程铮.高密度交通场景下智能汽车多目标检测与跟踪算法研究[D].镇江:江苏大学,2022.
ZHU Cheng-zheng. Research on multi-target detection and tracking algorithm of intelligent vehicle in high-density traffic scenario[D]. Zhenjiang: Jiangsu University,
2022.
[23]董 想.复杂场景下基于多尺度特征的目标检测算法研究[D].北京:北京交通大学,2022.
DONG Xiang. Research on object detection algorithm based on multi-scale feature in complex scene[D]. Beijing: Beijing Jiaotong University, 2022.
[24]GE Z, LIU S, Wang F, et al. YOLOX: Exceeding YOLO series in 2021[J]. arXiv, 2021, https://arxiv.org/abs/2107.08430.
[25]李明芳. DMP-YOLO:面向自动驾驶的多尺度目标检测算法[J/OL]. 无线电工程,(2025-08-18)[2025-09-15].https://link.cnki.net/urlid/13.1097.TN.20250818.
1352.002.
LI Ming-fang. DMP-YOLO: A multi-scale object detection algorithm for autonomous driving[J]. Radio Engineering,(2025-08-18)[2025-09-15]. https://link.cnki.net/urlid/13.1097.TN.20250818.1352.002.
[26]赵树恩,龚道元,田卓帅.基于改进YOLOv8模型的复杂交通场景目标检测算法[J].重庆交通大学学报(自然科学版),(2025-05-27)[2025-09-15].https://link.cnki.net/urlid/50.1190.U.20250527.1407.004.
ZHAO Shu-en, GONG Dao-yuan, TIAN Zhuo-shuai. Algorithm for object detection in complex traffic scenes based on improved YOLOv8 model[J]. Journal of Chongqing Jiaotong University(Natural Science Edition),(2025-05-27)[2025-09-15]. https://link.cnki.net/urlid/50.1190.U.20250527.1407.004.
[27]黄崇庆,徐慧英,张晓雷,等.BGR-YOLO:基于YOLOv8改进的交通场景下目标检测算法[J].计算机工程与科学,(2025-05-27)[2025-09-15].https://link.cnki.net/urlid/43.1258.TP.20250408.1455.002.
HUANG Chong-qing, XU Hui-ying, ZHANG Xiao-lei, et al. BGR-YOLO: An improved object detection algorithm under traffic scenarios based on YOLOv8[J]. Computer Engineering and Science,(2025-04-08)[2025-09-15]. https://link.cnki.net/urlid/43.1258.TP.20250408.1455.002.
[28]WOO S, PARK J, LEE J Y, et al. CBAM: Convolutional block attention module[M]//FERRARI V, HEBERT M, SMINCHISESCU C, et al. Computer Vision—ECCV 2018, Part Ⅶ. Berlin: Springer, 2018: 3-19.
[29]LIU S T, HUANG D, WANG Y H. Receptive field block net for accurate and fast object detection[C]//FERRARI V, HEBERT M, SMINCHISESCU C, et al. Computer Vision—ECCV 2018, Part Ⅺ. Berlin: Springer, 2018: 404-419.
[30]周 力,惠 飞,张嘉洋,等.基于RDB-YOLOv5的遥感图像车辆检测[J].长安大学学报(自然科学版),2024,44(3):149-160.
ZHOU Li, HUI Fei, ZHANG Jia-yang, et al. Remote sensing images vehicle detection based on RDB-YOLOv5[J]. Journal of Chang'an University(Natural Science Edition), 2024, 44(3): 149-160.
[31]乔 朋,袁 彪,申迎港,等.基于YOLOv5 DeepSORT和虚拟检测区的车轴时空定位方法[J].长安大学学报(自然科学版),2023,43(3):34-44.
QIAO Peng, YUAN Biao, SHEN Ying-gang, et al. Spatio-temporal axle localization method based on YOLOv5 DeepSORT and virtual detection area[J]. Journal of Chang'an University(Natural Science Edition), 2023, 43(3): 34-44.
[32]SZEGEDY C, LIU W, JIA Y Q, et al. Going deeper with convolutions[C]//IEEE. 2015 IEEE Conference on Computer Vision and Pattern Recognition(CVPR). New York: IEEE, 2015: 1-9.
[33]GEVORGYAN Z. SIoU loss: More powerful learning for bounding box regression[J]. arXiv, 2022, https://arxiv.org/abs/2205.12740.

备注/Memo

备注/Memo:
收稿日期:2025-01-24
基金项目:国家重点研发计划项目(2021YFB2601400)
作者简介:周 力(1987-),男,陕西省西安人,讲师,工学博士,E-mail:zhouli180172@chd.edu.cn。
通信作者:代 亮(1981-),男,陕西省西安人,教授,工学博士,E-mail:ldai@chd.edu.cn。
更新日期/Last Update: 2025-09-30