YOLOv5的改进算法及其在自动驾驶多目标检测的应用研究

doi:10.3778/j.issn.1002-8331.2302-0185

摘要/Abstract

摘要： 车辆、行人等交通参与者的实时检测是自动驾驶汽车与外界环境实现信息交互的重要环节，而在复杂天气条件下多目标检测精度低仍是一个挑战性问题。提出了一种YOLOv5的改进算法及其在自动驾驶多目标检测的应用方法。该方法采用K-means++算法对数据集中的目标样本聚类，以获得更符合不同目标尺度的锚框，提高多目标定位及其实体分割的精度；在原YOLOv5的骨干网络中添加Coordinate Attention（坐标注意力）模块，以提高模型的特征提取能力；将原YOLOv5网络中的PANet（路径聚合网络）结构替换为BiFPN（双向特征金字塔）结构，实现自上而下与自下而上的深浅层特征双向融合，提高模型对不同尺度目标的整体检测精度。对比实验结果表明：改进后的YOLOv5算法获得了更好的性能，目标检测的mAP达到了92.2%，比改进前的YOLOv5算法提升了8.47%。

关键词: 目标检测, YOLOv5, 自动驾驶, 双向特征金字塔, 坐标注意力

Abstract: Real-time detection of vehicles, pedestrians, and other traffic participants is an important part of information interaction between the autonomous vehicle and the external environment. However, the low accuracy of multi-target detection in complex weather conditions is still a challenge. For this reason, this paper proposes an improved algorithm of YOLOv5 and its application in the multi-target detection of auto-driving systems. In this method, K-means++ algorithm is used to cluster the target samples in the data set to obtain anchor frames that are more suitable for different target scales and improve the accuracy of multi-target location and entity segmentation. Coordinate attention module is added to the backbone of the original YOLOv5 to improve the feature extraction capability of the model. The PANet（path aggregation network） structure in the original YOLOv5 network is replaced by the BiFPN（bidirectional feature pyramid） structure to realize the bidirectional fusion of deep and shallow features from top to bottom and from bottom to top, and improves the overall detection accuracy of the algorithm for targets with different scales. The experimental results show that the improved YOLOv5 algorithm achieves better performance, and the mAP of target detection reaches 92.2%, which is 8.47% higher than the results obtained by original YOLOv5 algorithm.

Key words: object detection, YOLOv5, automatic driving, bidirectional feature pyramid network, coordinate attention

宋绍剑, 夏海姐, 李刚. YOLOv5的改进算法及其在自动驾驶多目标检测的应用研究[J]. 计算机工程与应用, 2023, 59(15): 68-75.

SONG Shaojian, XIA Haijie, LI Gang. Research on Improved YOLOv5 Algorithm and Its Application in Multi-Object Detection for Automatic Driving[J]. Computer Engineering and Applications, 2023, 59(15): 68-75.

参考文献

[1] 陈虹，郭露露，宫洵，等.智能时代的汽车控制[J].自动化学报，2020，46（7）：1313-1332.
CHEN H，GUO L L，GONG X，et al.Automotive control in intelligent era[J].Acta Automatica Sinica，2020，46（7）：1313-1332.
[2] ZHAO Z Q，ZHENG P，XU S，et al.Object detection with deep learning：a review[J].IEEE Transactions on Neural Networks and Learning Systems，2019，30（11）：3212-3232.
[3] DALAL N，TRIGGS B.Histograms of oriented gradients for human detection[C]//2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition（CVPR’05），2005：886-893.
[4] LOWE D G.Object recognition from local scale-invariant features[C]//Proceedings of the Seventh IEEE International Conference on Computer Vision，1999：1150-1157.
[5] CORTES C，VAPNIK V.Support-vector networks[J].Machine Learning，1995，20：273-297.
[6] 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，2014：580-587.
[7] GIRSHICK R.Fast R-CNN[C]//Proceedings of the IEEE International Conference on Computer Vision，2015：1440-1448.
[8] 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.
[9] 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，2016：779-788.
[10] REDMON J，FARHADI A.YOLO9000：better，faster，stronger[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition，2017：7263-7271.
[11] REDMON J，FARHADI A.Yolov3：an incremental improvement[J].arXiv：1804.02767，2018.
[12] BOCHKOVSKIY A，WANG C Y，LIAO H Y M.Yolov4：optimal speed and accuracy of object detection[J].arXiv：2004.10934，2020.
[13] CARION N，MASSA F，SYNNAEVE G，et al.End-to-end object detection with transformers[C]//Proceedings 16th European Conference on Computer Vision，Glasgow，UK，August 23-28，2020：213-229.
[14] SAHIB F I A A L，TAHER H B，GHANI R F.Detection of the autonomous car robot using Yolo[C]//Journal of Physics：Conference Series，2021.
[15] SONG S J，LI Y C，HUANG Q B，et al.A new real-time detection and tracking method in videos for small target traffic signs[J].Applied Sciences，2021，11（7）：3061.
[16] HUU P N，PHAM THI Q，TONG THI QUYNH P.Proposing lane and obstacle detection algorithm using YOLO to control self-driving cars on advanced networks[J].Advances in Multimedia，2022：3425295.
[17] KASPER-EULAERS M，HAHN N，BERGER S，et al.Detecting heavy goods vehicles in rest areas in winter conditions using YOLOv5[J].Algorithms，2021，14（4）：114.
[18] ZHU X，LYU S，WANG X，et al.TPH-YOLOv5：improved YOLOv5 based on transformer prediction head for object detection on drone-captured scenarios[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision，2021：2778-2788.
[19] LIU S，QI L，QIN H，et al.Path aggregation network for instance segmentation[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition，2018：8759-8768.
[20] HOU Q，ZHOU D，FENG J.Coordinate attention for efficient mobile network design[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition，2021：13713-13722.
[21] LIN T Y，DOLLáR P，GIRSHICK R，et al.Feature pyramid networks for object detection[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition，2017：2117-2125.
[22] TAN M，PANG R，LE Q V.Efficientdet：scalable and efficient object detection[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition，2020：10781-10790.