基于改进YOLOv4的车辆检测算法

doi:10.16180/j.cnki.issn1007-7820.2025.01.011

电子科技 ›› 2025, Vol. 38 ›› Issue (1): 81-87.doi: 10.16180/j.cnki.issn1007-7820.2025.01.011

基于改进YOLOv4的车辆检测算法

赖颖(), 巨志勇, 叶雨新

上海理工大学光电信息与计算机工程学院,上海 200093

收稿日期:2023-05-24 修回日期:2023-07-04 出版日期:2025-01-15 发布日期:2025-01-06
通讯作者: 赖颖(1996-),男, E-mail:2224735890@qq.com,硕士研究生。研究方向:图像识别、目标检测。
作者简介:巨志勇(1975-),男,博士,讲师。研究方向:模式识别、机器视觉、深度学习。
叶雨新(1999-),男,硕士研究生。研究方向:图像识别、目标检测。
基金资助:
国家自然科学基金(81101116)

A Vehicle Detection Algorithm Based on Improved YOLOv4

LAI Ying(), JU Zhiyong, YE Yuxin

School of Optical-Electrical and Computer Engineering, University of Shanghai for Science and Technology,Shanghai 200093,China

Received:2023-05-24 Revised:2023-07-04 Online:2025-01-15 Published:2025-01-06
Supported by:
National Natural Science Foundation of China(81101116)

摘要/Abstract

摘要：

在交通监控中进行车辆检测过程时,存在车辆互相遮挡和远距离目标尺寸不足的问题,导致在检测中存在漏检和误检情况。针对此问题,文中提出一种基于YOLOv4(You Only Look Once version 4)的多尺度融合与注意力机制的交通车辆检测算法。在YOLOv4的路径聚合网络中增加一个新的特征层进行多尺度特征融合,提升模型对底层纹理特征的提取能力。在YOLO Head检测头前嵌入ECA(Efficient Channel Attention)通道注意力模块,对聚合后的特征进行合理的抑制和增强,将CIoU(Complete Intersection over Union)损失函数替换为Soft-CIoU损失函数,提高小目标车辆对损失函数的贡献度。在公开车辆数据集UA-DETRAC与KITTI中的实验结果表明,相较于原YOLOv4算法,所提算法的平均精度分别提升了2.45百分点和1.14百分点,检测速度达到41.67 frame·s^-1。相较于其他先进算法,所提算法在检测精度上表现良好。

关键词: 车辆检测, 多尺度特征融合, 注意力机制, Soft-CIOU损失函数, YOLOv4, 深度学习, 目标检测, 小目标

Abstract:

In the process of vehicle detection in traffic monitoring, there are some problems such as vehicles shielding each other and insufficient distance target size, which leads to missing detection and false detection. To solve this problem, this study proposes a traffic vehicle detection algorithm based on YOLOv4(You Only Look Once version 4) multi-scale fusion and attention mechanism. A new feature layer is added to YOLOv4's path aggregation network for multi-scale feature fusion to improve the model's ability to extract underlying texture features. The ECA (Efficient Channel Attention) channel attention module is embedded in front of YOLO Head detection head to reasonably suppress and enhance the aggregated features. The CIoU (Complete Intersection over Union) loss function is replaced by the Soft-CIoU loss function to improve the contribution of small target vehicles to the loss function. The experimental results on the publicly available vehicle data sets UA-DETRAC and KITTI show that compared to the original YOLOv4 algorithm, the average accuracy of the proposed algorithm improves by 2.45 percentage points and 1.14 percentage points, respectively, and the detection speed reaches 41.67 frame·s^-1.The proposed algorithm performs well in detection accuracy when compared with other advanced algorithms.

Key words: vehicle detection, multi-scale feature fusion, attention mechanism, Soft-CIOU loss function, YOLOv4, deep learning, target detection, small target

中图分类号:

TP391

赖颖, 巨志勇, 叶雨新. 基于改进YOLOv4的车辆检测算法[J]. 电子科技, 2025, 38(1): 81-87.

LAI Ying, JU Zhiyong, YE Yuxin. A Vehicle Detection Algorithm Based on Improved YOLOv4[J]. Electronic Science and Technology, 2025, 38(1): 81-87.

图/表 9

图1

图2

图3

图4

表1

表2

表3

图5

表4

参考文献 21

[1]	Wu S, Nagahashi H. Parameterized daBoost:Introducing a parameter to speed up the training of real AdaBoost[J]. IEEE Signal Processing Letters, 2014, 21(6):687-691.
[2]	Wang C C R, Lien J J J. Automatic vehicle detection using local features-A statistical approach[J]. IEEE Transactions on Intelligent Transportation Systems, 2008, 9(1):83-96.
[3]	Yang X F, Yang Y. A method of efficient vehicle detection based on HOG-LBP[J]. Computer Engineering, 2014, 40(9):210-214.
[4]	Tu W T, Guo J X. Research on vehicle detection technology based on SIFT feature[C]. Beijing: The Eighth International Conference on Electronics Information and Emergency Communication, 2018:274-278.
[5]	Girshick R, Donahue J, Darrell T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]. Nashville: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2014:580-587.
[6]	Ren S, He K, Girshick R, et al. Faster R-CNN:Towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2017, 39(6):1137-1149.
[7]	Vuola A O, Akram S U, Kannala J. Mask R-CNN and U-Net ensembled for nuclei segmentation[C]. Venice: IEEE the Sixteenth International Symposium on Biomedical Imaging, 2019:208-212.
[8]	Cai Z, Vasconcelos N. Cascade R-CNN:Delving into high quality object detection[C]. Venice: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018:6154-6162.
[9]	段仲静, 李少波, 胡建军, 等. 深度学习目标检测方法及其主流框架综述[J]. 激光与光电子学展, 2020, 57(12):59-74.
	Duan Zhongjing, Li Shaobo, Hu Jianjun, et al. Review of deep learning based object detection methods and their mainstream frameworks[J]. Laser & Optoelectronics Progress, 2020, 57(12):59-74.
[10]	Liu W, Anguelov D, Erhan D, et al. SS:Single shot multibox detector[C]. Amsterdam: Proceedings of the Fourteenth European Conference,Springer International Publishing, 2016:1-37.
[11]	Redmon J, Divvala S, Girshick R, et al. You only look once:Unified,real-time object detection[C]. New Orleans: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016:779-788.
[12]	刘肯, 何姣姣, 张永平, 等. 改进YOLO的车辆检测算法[J]. 现代电子技术, 2019, 42(13):47-50.
	Liu Ken, He Jiaojiao, Zhang Yongping, et al. Improve YOLO vehicle detection algorithm[J]. Modern Electronics Technique, 2019, 42(13):47-50.
[13]	杨帆, 吴韶波. 基于SSD的目标车辆检测算法研究[J]. 物联网技术, 2021, 11(6):19-22.
	Yang Fan, Wu Shaobo. Research on target vehicle detection algorithm based on SSD[J]. Internet of Technologies, 2021, 11(6):19-22.
[14]	Dong W, Yang Z, Ling W, et al. Research on vehicle detection algorithm based on convolutional neural network and combining color and depth images[C]. Munich: The Second International Conference on Information Systems and Computer Aided Education, 2019:274-277.
[15]	Jiang P, Ergu D, Liu F, et al. A review of YOLO algorithm developments[J]. Procedia Computer Science, 2022, 19(9):1066-1073.
[16]	梁鸿, 王庆玮, 张千, 等. 小目标检测技术研究综述[J]. 计算机工程与应用, 2021, 57(1):17-28.
	Liang Hong, Wang Qingwei, Zhang Qian, et al. Small object detection technology:A review[J]. Computer Engineering and Applications, 2021, 57(1):17-28.
[17]	Wang Q, Wu B, Zhu P, et al. ECA-Net:Efficient channel attention for deep convolutional neural networks[C]. Nashville: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020:11534-11542.
[18]	Hu J, Shen L, Sun G. Squeeze-and-excitation networks[C]. Salt Lake City: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018:7132-7141.
[19]	张莹, 刘子龙, 万伟. 基于Faster R-CNN的无人机车辆目标检测[J]. 电子科技, 2021, 34(11):11-20.
	Zhang Ying, Liu Zilong, Wan Wei. UAV vehicle target detection based on Faster R-CNN[J]. Electronic Science and Technology, 2021, 34(11):11-20.
[20]	Han G, Su J, Zhang C. A method based on multi-convolution layers joint and generative adversarial networks for vehicle detection[J]. KSII Transactions on Internet and Information Systems, 2019, 13(4):1795-1811.
[21]	Chu J, Guo Z, Leng L. Object detection based on multi-layer convolution feature fusion and online hard example mining[J]. IEEE Access, 2018, 6(1):19959-19967.

权重因子k	AP/%
权重因子k	UADETRAC	KITTI
1	73.03	95.26
2	73.25	95.31
3	72.94	95.02
4	72.83	93.45
5	70.81	92.12

模型	AP/%	Precision/%	Recall/%
YOLOv4	72.89	83.56	69.53
YOLOv4-1	74.62	84.37	68.72
YOLOv4-2	73.74	84.12	69.01
YOLOv4-3	73.25	83.77	69.43
改进YOLOv4	75.34	84.75	69.89

模型	AP/%	Precision/%	Recall/%
YOLOv4	94.90	96.10	89.73
YOLOv4-1	95.97	96.55	91.07
YOLOv4-2	95.43	96.43	91.02
YOLOv4-3	95.31	96.35	90.51
改进YOLOv4	96.31	96.89	91.01

模型	AP/%		FPS /frame·s^-1
模型	UADETRAC	KITTI	FPS /frame·s^-1
Faster RCNN	71.17	89.02	8.14
SSD	67.75	85.67	30.43
YOLOV3	40.13	72.32	52.06
YOLOV4	72.89	94.90	55.51
改进YOLOv4	75.34	96.31	41.67

基于改进YOLOv4的车辆检测算法

A Vehicle Detection Algorithm Based on Improved YOLOv4

RichHTML

PDF (PC)

赞

可视化

摘要/Abstract

引用本文

使用本文

图/表 9

参考文献 21

相关文章 15

Metrics

本文评价

推荐阅读 0

[1]	王紫祎, 陈世平. 基于图对比学习的自监督网络流量检测模型[J]. 电子科技, 2025, 38(3): 22-31.
[2]	赵云飞, 薛存金. 一种融合空洞卷积与池化模型的遥感影像水体提取方法[J]. 电子科技, 2025, 38(3): 40-46.
[3]	陈宇洋, 李峰. 融合CNN与Transformer的视网膜OCT图像积液分割方法[J]. 电子科技, 2025, 38(3): 47-59.
[4]	郑方亮, 王延年, 廉继红, 阮佩. 基于条件先验Swin Transformer的人脸图像超分辨重建[J]. 电子科技, 2025, 38(2): 35-41.
[5]	蒯新晨, 李烨. 多种及多尺度注意力混合的图像超分辨率重建[J]. 电子科技, 2024, 37(9): 34-42.
[6]	何志强, 孙占全. 基于Swin-Transformer的颈动脉超声图像斑块分割[J]. 电子科技, 2024, 37(9): 48-56.
[7]	朱勇, 黄永明, 何幸. 基于雷达和视频融合的目标检测[J]. 电子科技, 2024, 37(8): 1-7.
[8]	谢熙君, 李菲菲. 基于非局部支持注意力的小样本目标检测算法[J]. 电子科技, 2024, 37(8): 75-83.
[9]	李筠, 刘虎, 杨海马, 王原, 徐文成, 黄宏欣. 融合目标检测与DWA算法的AGV路径[J]. 电子科技, 2024, 37(7): 33-42.
[10]	何幸, 黄永明, 朱勇. 基于改进YOLOv5的路面坑洼检测方法[J]. 电子科技, 2024, 37(7): 53-59.
[11]	王新宇, 赵静文, 刘翔, 石蕴玉, 佘云浪. 融合坐标注意力机制的YOLOv3肺结节检测算法[J]. 电子科技, 2024, 37(6): 1-7.
[12]	叶雨新, 巨志勇, 赖颖. 融合感受野增强和注意力机制的交通标志检测算法[J]. 电子科技, 2024, 37(6): 8-16.
[13]	袁傲, 齐金鹏, 贾灿, 薛宇鑫, 郭阳阳. 一种基于深度学习的时序病变数据段分类方法[J]. 电子科技, 2024, 37(6): 84-91.
[14]	朱子豪, 宋燕. 融合注意力和胶囊池化的轻量型胶囊网络[J]. 电子科技, 2024, 37(5): 1-8.
[15]	夏榕成, 刘德儿. 基于深度学习的城市内涝区域车辆检测与分析[J]. 电子科技, 2024, 37(5): 18-24.