基于改进YOLOv7-tiny的车辆目标检测算法

doi:10.12382/bgxb.2024.0393

摘要/Abstract

摘要：

为更好地保护人民的生命财产安全,针对目前依靠人力进行交通管理工作时统计不准确、反馈不及时等问题,提出一种适合部署在边缘终端设备上的基于YOLOv7-tiny算法改进的车辆目标检测算法。通过构造深度强力残差卷积块对主干网络的轻量级高效层聚合网络(Efficient Layer Aggregation Network-Tiny,ELAN-T)模块进行轻量化改进;通过削减分支,对特征融合网络的ELAN-T模块进行轻量化改进,降低网络的参数量和计算量,并对特征融合网络的结构进行重新构造;引入高效通道注意力机制和EIOU边界框损失函数提升算法的精度。在预处理后的UA-DETRAC数据集上实验,改进后的算法参数量相比于原始的YOLOv7-tiny算法降低了15.1%,计算量降低了5.3%,mAP@0.5提升了5.3个百分点。实验结果表明,改进后的算法不仅实现了轻量化,而且检测精度有所提升,适合部署在边缘终端设备上,完成对道路中车辆的检测任务。

关键词: 车辆检测, YOLOv7-tiny算法, 深度强力残差卷积块, 轻量级高效层聚合网络模块

Abstract:

At present,the traffic management relying on manpower is characterized by inaccurate statistics and delayed feedback.A vehicle detection algorithm based on the improved YOLOv7-tiny algorithm suitable for deploying on edge terminal devices is proposed to better protect people’s lives and property.A deep powerful residual (DP_Res)convolutional block isconstructed to perform the lightweight improvements on the efficient layer aggregation network-tiny (ELAN-T) module of backbone network.By reducing branches,the lightweight improvement on the ELAN-T module of the feature fusion network is made to reduce the number of parameters and computational load of the network,and the structure of the feature fusion network is reconstructed;The efficient channel attention mechanism and the EIOU bounding box loss function are introduced to improve the accuracy of the algorithm.The experiment is conducted on the preprocessed UA-DETRAC dataset,and the parameters of the improved algorithm are reduced by 15.1% compared to those of the original YOLOv7-tiny,with a reduction in computation of 5.3% and an increase in mAP@0.5 of 5.3 percentage points.The experimental results show that the improved algorithm not only achieves lightweight,but also improves the detection accuracy,making it suitable for deployment on edge terminal devices to complete the task of detecting vehicles on the road.

Key words: vehicle detection, YOLOv7-tiny algorithm, deep powerful residual convolutional block, efficient layer aggregation network-tiny module

中图分类号:

TP391.4

赵海丽, 许修常, 潘宇航. 基于改进YOLOv7-tiny的车辆目标检测算法[J]. 兵工学报, 2025, 46(4): 240393-.

ZHAO Haili, XU Xiuchang, PAN Yuhang. Vehicle Target Detection Algorithm Based on Improved YOLOv7-tiny[J]. Acta Armamentarii, 2025, 46(4): 240393-.

图/表 13

图1 YOLOv7-tiny算法网络结构

Fig.1 Network structure of YOLOv7-tiny algorithm

图2 改进的YOLOv7-tiny算法网络结构

Fig.2 Network structure of improved YOLOv7-tiny algorithm

图3 DP_Res卷积块结构

Fig.3 Structure of DP_Res convolutional block

图4 QLA-ELAN模块结构

Fig.4 Structure of QLA-ELAN module

图5 QLB-ELAN模块结构

Fig.5 Structure of QLB-ELAN module

图6 改进的特征融合网络简要流程

Fig.6 Brief process of improved feature fusion network

图7 ECA模块结构

Fig.7 Structure of ECA module

图8 EIOU示意图

Fig.8 Schematic diagram of EIOU

表1 消融实验结果对比

Table 1 Comparison of ablation experimental results

序号	输入图片尺寸	改进主干网络	改进特征融合网络	ECA	EIOU	参数量/10⁶	计算量/10⁹	mAP@0.5/%
1	3×640×640					6.02	13.2	72.1
2	3×640×640	√				5.26	11.8	71.3
3	3×640×640	√	√			5.11	12.5	73.3
4	3×640×640	√	√	√		5.11	12.5	75.5
5	3×640×640	√	√	√	√	5.11	12.5	77.4

图9 损失值曲线对比

Fig.9 Comparison of loss curves

表2 不同算法实验结果对比

Table 2 Comparison of experimental results of different algorithms

算法	参数量/10⁶	计算量/10⁹	mAP@ 0.5/%	帧率
YOLOv7-tiny算法	6.02	13.2	72.1	149.3
YOLOv3-tiny算法	8.68	13.0	65.7	294.1
YOLOv5s算法	7.03	16.0	74.1	166.7
YOLOv7-MobileNetv3算法	4.17	7.0	72.3	82.6
YOLOv7-GhostNet算法	6.35	9.9	73.4	75.2
本文算法	5.11	12.5	77.4	144.9

图10 算法检测效果对比

Fig.10 Comparison of detection effects of the algorithms

图11 改进后算法连板推理的结果

Fig.11 The results of the improved algorithm’s board connected inference

参考文献 25

[1]	KRIZHEVSKY A, SUTSKEVER I, HINTON G E. ImageNet classification with deep convolutional neural networks[J]. Communications of the ACM, 2017, 60(6):84-90.
[2]	GIRSHICK R, DONAHUE J, DARRELL T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Columbus,OH, US: IEEE, 2014:580-587.
[3]	GIRSHICK R. Fast R-CNN[C]// Proceedings of the IEEE International Conference on Computer Vision.Santiago, Chile: IEEE, 2015:1440-1448.
[4]	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.
[5]	REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once:Unified,real-time object detection[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Las Vegas,NV, US: IEEE, 2016:779-788.
[6]	REDMON J, FARHADI A. YOLO9000:better,faster,stronger[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Honolulu,HI, US: IEEE, 2017:7263-7271.
[7]	REDMON J, FARHADI A. YOLOv3:an incremental improvement:arXiv:1804.02767[R/OL]. Ithaca,NY, US: Cornell University, 2018(2018-04-08). https://arxiv.org/abs/1804.02767
[8]	BOCHKOVSKIY A, WANG C Y, LIAO H Y M. Yolov4:optimal speed and accuracy of object detection:arXiv:2004.10934[R/OL]. Ithaca,NY, US: Cornell University, 2020(2020-04-23). https://arxiv.org/abs/2004.10934.
[9]	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]// Proceedings of the 2023 IEEE/CVF conference on computer vision and pattern recognition.Vancouver,BC, Canada: IEEE, 2023:7464-7475.
[10]	张博尧, 冷雁冰. 基于YOLOv4网络模型的金属表面划痕检测[J]. 兵工学报, 2022, 43(增刊1):214-221.
	ZHANG B Y, LENG Y B. Metal surface scratch detection based on YOLOv4 network model[J]. Acta Armamentarii, 2022, 43(S1):214-221. (in Chinese)
[11]	惠康华, 杨卫, 刘浩翰, 等. 基于YOLOv5的增强多尺度目标检测方法[J]. 兵工学报, 2023, 44(9):2600-2610. doi: 10.12382/bgxb.2022.1147
	HUI K H, YANG W, LIU H H, et al. Enhanced multi-scale object detection method based on YOLOv5[J]. Acta Armamentarii, 2023, 44(9):2600-2610. (in Chinese)
[12]	LIU W, ANGUELOV D, ERHAN D, et al. SSD:single shot multibox detector[C]// Computer Vision-ECCV 2016.Amsterdam, The Netherlands: Springer, 2016:21-37.
[13]	LIU S T, HUANG D, WANG Y H. Learning spatial fusion for single-shot object detection:arXiv:1911.09516[J]. Ithaca,NY, US: Cornell University, 2019(2019-11-25). https://arxiv.org/abs/1911.09516.
[14]	WANG Q F, XU N, HUANG B J, et al. Part-aware refinement network for occlusion vehicle detection[J]. Electronics, 2022, 11(9):1375.
[15]	ZHANG Y, GUO Z Y, WU J Q, et al. Real-time vehicle detection based on improved YOLOV5[J]. Sustainability, 2022, 14(19):12274.
[16]	WANG J D, DONG Y H, ZHAO S R, et al. A high-precision vehicle detection and tracking method based on the attention mechanism[J]. Sensors, 2023, 23(2):724.
[17]	XU Y Q, CHU K B, ZHANG J. Nighttime vehicle detection algorithm based on improved faster-RCNN[J]. IEEE Access, 2023,12:19299-19306.
[18]	WANG Q L, WU B G, ZHU P F, et al. ECA-Net:efficient channel attention for deep convolutional neural networks[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.Seattle,WA, US: IEEE, 2020:11534-11542.
[19]	YANG Z M, WANG X L, LI J G. EIoU:an improved vehicle detection algorithm based on vehiclenet neural network[C]// Journal of Physics:Conference Series,2021, 1924(1):012001.
[20]	WEN L Y, DU D W, CAI Z W, et al. UA-DETRAC:a new benchmark and protocol for multi-object detection and tracking[J]. Computer Vision and Image Understanding, 2020,193:102907.
[21]	WANG C Y, LIAO H Y M, YEH I H. Designing network design strategies through gradient path analysis:arXiv:2211.04800[R/OL]. Ithaca,NY, US: Cornell University, 2022(2022-11-09). https://arxiv.org/abs/2211.04800.
[22]	HAN K, WANG Y H, TIAN Q, et al. Ghostnet:more features from cheap operations[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.Seattle,WA, US: IEEE, 2020:1580-1589.
[23]	BENJUMEA A, TEETI I, CUZZOLIN F, et al. YOLO-Z:improving small object detection in YOLOv5 for autonomous vehicles:arXiv:2112.11798[R/OL]. Ithaca,NY, US: Cornell University, 2021(2021-12-22).
[24]	HU J, SHEN L, SUN G. Squeeze-and-excitation networks[C]// Proceedings of 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Salt Lake City,UT, US: IEEE, 2018:7132-7141.
[25]	HOWARD A, SANDLER M, CHU G, et al. Searching for MobileNetv3[C]// Proceedings of the IEEE/CVF International Conference on Computer Vision.Seoul, Korea (South): IEEE, 2019:1314-1324.