改进YOLOv5s的无人机目标检测算法

doi:10.3778/j.issn.1002-8331.2205-0200

摘要/Abstract

摘要： 无人机在情报、侦察和监视领域，目标自动检测可为侦察等任务提供准确的目标位置及类别，为地面指挥人员提供详尽的目标信息。针对无人机图像背景复杂、分辨率高、目标尺度差异大等特点，提出一种改进YOLOv5s目标检测算法。将压缩-激励模块引入到YOLOv5s算法中，提高网络的特征提取能力；引入双锥台特征融合（bifrustum feature fusion，BFF）结构，提高算法对较小目标的检测检测精度；将CIoU Loss替换GIoU Loss作为算法的损失函数，在提高边界框回归速率的同时提高定位精度。实验结果表明，改进后的YOLOv5s取得了86.3%的平均均值精度（mAP），比原算法YOLOv5s提高了16.8个百分点，在复杂背景下仍能显著提升无人机图像目标检测性能。

关键词: 无人机检测, YOLOv5s, 压缩激励模块, CIoU Loss

Abstract: In the field of intelligence, reconnaissance and surveillance, automatic target detection can provide accurate target location and category for reconnaissance and other tasks, and provide detailed target information for ground commanders. Based on the complex background, high resolution and target scale difference of UAV images, a modified YOLOv5s object detection algorithm is proposed. Firstly, the compression-excitation module is introduced into the YOLOv5s algorithm to improve the feature extraction ability of the network. Secondly, the double-cone feature fusion（bifrustum feature fusion, BFF） structure is introduced to improve the detection accuracy of the algorithm for smaller targets. Finally, CIoU Loss replaces GIoU Loss as the loss function of the algorithm to improve the positioning accuracy of improving the bounding box regression rate. The experimental results show that the improved YOLOv5s achieves an average mean accuracy（mAP） of 86.3%, 16.8 percentage points higher compared to the original algorithm YOLOv5s, and can still significantly improve the UAV image target detection performance in a complex background.

Key words: UAV detection, YOLOv5s, compression excitation module, CIoU Loss

宋谱怡, 陈红, 苟浩波. 改进YOLOv5s的无人机目标检测算法[J]. 计算机工程与应用, 2023, 59(1): 108-116.

SONG Puyi, CHEN Hong, GOU Haobo. Improving UAV Object Detection Algorithm for YOLOv5s[J]. Computer Engineering and Applications, 2023, 59(1): 108-116.

参考文献

[1] 张瑶，卢焕章，张路平，等.基于深度学习的视觉多目标跟踪算法综述[J].计算机工程与应用，2021，57（13）：55-66.ZHANG Y，LU H Z，ZHANG L P，et al.Overview of visual muli-object tracking algorithms with deep learning[J].Computer Engineering and Applications，2021，57（13）：55-66.
[2] 张宸嘉，朱磊，俞璐.卷积神经网络中的注意力机制综述[J].计算机工程与应用，2021，57（20）：64-72.
ZHANG C J，ZHU L，YU L.Review of attention mechanism in convolutional neural networks[J].Computer Engineering and Applications，2021，57（20）：64-72.
[3] REDMONJ，FARHADI A.YOLOv3：An incremental improvement[J].arXiv：1804.02767，2018.
[4] ULTRALYTICS.YOLOv5[EB/OL].[2022-03-15）].https：//github.com/ultralytics/yolov5.
[5] ZHENG Z H，WANG P，LIU W，et al.Distance-IoU loss：Faster and better learning for bounding box regression[C]//Proceedings of the AAAI Conference on Artificial Intelligence，2020：12993-13000.
[6] ANWAR M Z，KALEEM Z，JAMALIPOUR A.Machine learning inspired sound-based amateur drone detection for public safety applications[J].IEEE Transactions on Vehicular Technology，2019，68（3）：2526-2534.
[7] 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.
[8] XUE T，SHAN T，FENG Y.Research on clutter suppression for low-altitude slow and small target detection[C]//Proceedings of the 2nd International Conference on Computer Engineering，Information Science and Internet Technolog，2017.
[9] BOCHKOVSKIY A，WANG C Y，LIAO H Y M.YOLOv4：Optimal speed and accuracy of object detection[J].arXiv：2004.10934，2020.
[10] 吴萌萌，张泽斌，宋尧哲，等.基于自适应特征增强的小目标检测网络[J/OL].激光与光电子学进展：1-14（2022-02-14）[2022-03-15].https：//kns.cnki.net/kcms/detail/31.1690.tn.20220212.1752.026.html.
WU M M，ZHANG Z B，SONG Y Z，et al.Small object detection network based on adaptive feature enhancement[J/OL].Laser &Optoelectronics Progress：1-14（2022-02-14）[2022-03-15].https：//kns.cnki.net/kcms/detail/31.1690.
tn.20220212.1752.026.html.
[11] 李永上，马荣贵，张美月.改进YOLOv5s+DeepSORT的监控视频车流量统计[J].计算机工程与应用，2022，58（5）：271-279.
Ll Y S，MA R G，ZHANG M Y.Trafic monitoring video vehicle volume statistics method based on improved YOLOv5s+DeepSORT[J].Computer Engineering and Applications，2022，58（5）：271-279.
[12] LI X P，LEI S Z，ZHANG B X，et al.Fast aerial UAV detection using improved inter-frame difference and SVM[J].Journal of Physics（Conference Series），2019，1187（3）：032082.
[13] LIU Y C，LIAO L C，WU H，et al.Trajectory and image-based detection and identification of UAV[J].The Visual Computer，2021，37（7）：1769-1780.
[14] 刘闪亮，吴仁彪，屈景怡，等.Bi-PPYOLO tiny：一种轻量型的机场无人机检测方法[J/OL].安全与环境学报（2021-12-21）[2022-03-15）].https：//kns.cnki.net/kcms/detail/11.4537.x.20211221.1700.001.html.
LIU S L，WU R B，QU J Y，et al.Bi-PPYOLO tiny：A lightweight airport UAV detection method[J/OL].The Journal of Safety and the Environment（2021-12-21）[2022-03-15）].https：//kns.cnki.net/kcms/detail/11.4537.x.20211221.1700.001.html.
[15] PHAM G N，NGUYEN P H.Drone detection experiment based onimage processing and machine learning[J].International Journal of Scientific & Technology Research，2020，9（2）：2965-2971.
[16] 王凤随，王启胜，陈金刚，等.基于注意力机制和Soft-NMS的改进Faster R-CNN目标检测算法[J].激光与光电子学进展，2021，58（24）：397-408.
WANG F S，WANG Q S，CHEN J G，et al.ImprovedFaster R-CNN target detection algorithm based on attentionmechanism and Soft-NMS[J].Laser & Optoelectronics Progress，2021，58（24）：397-408.
[17] LIN T Y，GOYAL P，GIRSHICK R，et al.Focal loss for dense object detection[C]//Proceedings of the IEEE International Conference on Computer Vision，2017：2999-3007.
[18] REN S Q，HE K M，GIRSHICK R，et al.Faster R-CNN：Towards real-time object detection with region proposal networks[J].arXiv：1506.01497，2015.
[19] WANG C Y，MARK LIAO H Y，WU Y H，et al.CSPNet：A new backbone that can enhance learning capability of CNN[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops，2020.