YOLOv8-DEL：基于改进YOLOv8n的实时车辆检测算法研究

doi:10.3778/j.issn.1002-8331.2407-0125

摘要/Abstract

摘要： 车辆检测是智能交通系统和自动驾驶的重要组成部分。然而，实际交通场景中存在许多不确定因素，导致车辆检测模型的准确率低实时性差。为了解决这个问题，提出了一种快速准确的车辆检测算法——YOLOv8-DEL。使用DGCST（dynamic group convolution shuffle transformer）模块代替C2f模块来重构主干网络，以增强特征提取能力并使网络更轻量；添加的P2检测层能使模型更敏锐地定位和检测小目标，同时采用Efficient RepGFPN进行多尺度特征融合，以丰富特征信息并提高模型的特征表达能力；通过结合GroupNorm和共享卷积的优点，设计了一种轻量型共享卷积检测头，在保持精度的前提下，有效减少参数量并提升检测速度。与YOLOv8相比，提出的YOLOv8-DEL在BDD100K数据集和KITTI数据集上，mAP@0.5分别提高了4.8个百分点和1.2个百分点，具有实时检测速度（208.6?FPS和216.4?FPS），在检测精度和速度方面实现了更有利的折中。

关键词: 车辆检测, YOLOv8, DGCST, Efficient RepGFPN, 轻量级检测头

Abstract: Vehicle detection is of great significance for intelligent transportation system and automatic driving technology. However, actual traffic scene suffers from many uncertain factors, leading to the low accuracy of vehicle detection and poor real-time performance. To solve this problem, this paper proposes a fast and accurate vehicle detection algorithm, named YOLOv8-DEL. First, dynamic group convolution shuffle transformer (DGCST) is used to replace several C2f module to enhance the ability of network feature extraction and make the backbone network lighter. Then a P2 detection layer is introduced to make the model more sensitive to locate and detect small targets, while Efficient RepGFPN is adopted for multi-dimensional feature fusion to strengthen the feature expression ability. Finally, a lightweight shared convolution detection head is proposed, which can reduce the model parameters and improve the detection speed while lossless accuracy. The experimental results demonstrate that compared with the YOLOv8, YOLOv8-DEL improves mAP@0.5 by 4.8?percentage points and 1.2?percentage points on BDD100K dataset and KITTI dataset with real-time detection speed (208.6?FPS and 216.4?FPS). YOLOv8-DEL achieves a better trade-off between detection accuracy and speed.

Key words: vehicle detection, YOLOv8, dynamic group convolution shuffle transformer( DGCST), Efficient RepGFPN, lightweight detection head

古佳欣, 陈高华, 张春美. YOLOv8-DEL：基于改进YOLOv8n的实时车辆检测算法研究[J]. 计算机工程与应用, 2025, 61(1): 142-152.

GU Jiaxin, CHEN Gaohua, ZHANG Chunmei. YOLOv8-DEL: Research on Real-Time Vehicle Detection Algorithm Based on Improved YOLOv8n[J]. Computer Engineering and Applications, 2025, 61(1): 142-152.

参考文献

[1] ZHAO H, WANG C, LIN Y, et al. On-road vehicle trajectory collection and scene-based lane change analysis: part I[J]. IEEE Transactions on Intelligent Transportation Systems, 2017, 18(1): 192-205.
[2] SONATA I, HERYADI Y, LUKAS L, et al. Autonomous car using CNN deep learning algorithm[J]. Journal of Physics: Conference Series, 2021: 012071.
[3] GIRSHICK R, DONAHUE J, DARRELL T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]//Proceedings of the 2014 IEEE Conference on Computer Vision and Pattern Recognition, 2014: 580-587.
[4] GIRSHICK R. Fast R-CNN[C]//Proceedings of the 2015 IEEE International Conference on Computer Vision, 2015: 1440-1448.
[5] 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 and Machine Intelligence, ?2017, 39(6): 1137-1149.
[6] DJENOURI Y, BELHADI A, SRIVASTAVA G, et al. Vehicle detection using improved region convolution neural network for accident prevention in smart roads[J]. Pattern Recognition Letters, 2022, 158: 42-47.
[7] WANG B, JIANG P, GAO J, et al. A lightweight few-shot marine object detection network for unmanned surface vehicles[J]. Ocean Engineering, 2023, 277: 114329.
[8] LI Y, CHEN Y, WANG N, et al. Scale-aware trident networks for object detection[C]//Proceedings of 2019 IEEE/CVF International Conference on Computer Vision, Seoul, Korea (South), 2019: 6054-6063.
[9] HU M, WU Y, YANG Y, et al. DAGL-Faster: domain adaptive Faster R-CNN for vehicle object detection in rainy and foggy weather conditions[J]. Displays, 2023, 79: 102484.
[10] REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: unified, real-time object detection[C]//Proceedings of the 2016 Conference on Computer Vision and Pattern Recognition, 2016: 779-788.
[11] LIU W, ANGUELOV D, ERHAN D, et al. SSD: single shot multibox detector[C]//Proceedings of the 14th European Conference on Computer Vision, Amsterdam, The Netherlands, Oct 11-14, 2016. Cham: Springer International Publishing, 2016: 21-37.
[12] BENJDIRA B, KHURSHEED T, KOUBAA A, et al. Car detection using unmanned aerial vehicles: comparison between Faster R-CNN and YOLOv3[C]//Proceedings of the 2019 1st International Conference on Unmanned Vehicle Systems, Oman, 2019: 1-6.
[13] AHMADI M, XU Z, WANG X, et al. Fast multi object detection and counting by YOLO V3[C]//Proceedings of the 2021 China Automation Congress, 2021: 7401-7404.
[14] BOCHKOVSKIY A, WANG C Y, LIAO H Y. YOLOv4: optimal speed and accuracy of object detection[J]. arXiv: 2004.10934, 2020.
[15] HE K, ZHANG X, REN S, 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.
[16] LIU S, QI L, QIN H, et al. Path aggregation network for instance segmentation[C]//Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2018: 8759-8768.
[17] JOCHE G, CHAURASIA A. Ultralytics/YOLOv5 in PyTorch[EB/OL].[2024-05-10]. https://github.com/ultralytics/yolov5.
[18] LI C, LI L, GENG Y, et al. YOLOv6 v3.0: a full-scale reloading[J]. arXiv:2301.05586, 2023.
[19]?XU S, WANG X, LV W, et al. PP-YOLOE: an evolved version of YOLO[J]. arXiv:2203.16250, 2022.
[20] DING X, ZHANG X, MA N, et al. RepVGG: making VGG-style ConvNets great again[C]//Proceedings of 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021: 13733-13742.
[21] WANG C Y, BOCHKOVSKIY A, LIAO H Y. YOLOv7: trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern, 2023: 7464-7475.
[22] HUANG S, HE Y, CHEN X. M-YOLO: a nighttime vehicle detection method combining MobileNet v2 and YOLO v3[J]. Journal of Physics: Conference Series, 2021: 012094.
[23] LI X, QIN Y, WANG F, et al. Pitaya detection in orchards using the MobileNet-YOLO model[C]//Proceedings of the 2020 39th Chinese Control Conference, 2020: 6274-6278.
[24] DONG X, YAN S, DUAN C. A lightweight vehicles detection network model based on YOLOv5[J]. Engineering Applications of Artificial Intelligence, 2022, 113: 104914.
[25] HAN K, WANG Y, TIAN Q, et al. GhostNet: more features from cheap operations[C]//Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Seattle, WA, USA, 2020.
[26] WOO S, PARK J, LEE J Y, et al, CBAM: convolutional block attention module[C]//Proceedings of the European Conference on Computer Vision, 2018: 3-19.
[27] BIE M, LIU Y, LI G, et al. Real-time vehicle detection algorithm based on a lightweight you-only-look-once (YOLOv5n-L) approach[J].?Expert Systems with Applications, 2023, 213: 119108.
[28] HU J, SHEN L, SUN G. Squeeze-and-excitation networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018: 7132-7141.
[29] JOCHE G, CHAURASIA A. Ultralytics/YOLOv8 in PyTorch[EB/OL].[2024-05-10].https://github.com/ultralytics/ultralytic.
[30] GONG W. Lightweight object detection: a study based on YOLOv7 integrated with ShuffleNetv2 and vision Transformer[J]. arXiv:2403.01736, 2024.
[31] XU X, JIANG Y, CHEN W, et al. DAMO-YOLO: a report on real-time object detection design[J]. arXiv:2211.15444, 2022.
[32] TIAN Z, SHEN C, CHEN H, et al. FCOS: fully convolutional one-stage object detection[C]//Proceedings of the 2019 IEEE/CVF International Conference on Computer Vision, Seoul, Korea (South), 2019.