Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2024 Mar 27;19(3):e0300120.
doi: 10.1371/journal.pone.0300120. eCollection 2024.

Multi-scale object detection in UAV images based on adaptive feature fusion

Siqi Tan  1 , Zhijian Duan  1 , Longzhong Pu  1
Affiliations

Multi-scale object detection in UAV images based on adaptive feature fusion

Siqi Tan et al. PLoS One. .

Abstract

With the widespread use of UAVs, UAV aerial image target detection technology can be used for practical applications in the military, traffic planning, personnel search and rescue and other fields. In this paper, we propose a multi-scale UAV aerial image detection method based on adaptive feature fusion for solving the problem of detecting small target objects in UAV aerial images. This method automatically adjusts the convolution kernel receptive field and reduces the redundant background of the image by adding an adaptive feature extraction module (AFEM) to the backbone network. This enables it to obtain more accurately and effectively small target feature information. In addition, we design an adaptive feature weighted fusion network (SBiFPN) to effectively enhance the representation of shallow feature information of small targets. Finally, we add an additional small target detection scale to the original network to expand the receptive field of the network and strengthen the detection of small target objects. The training and testing are carried out on the VisDrone public dataset. The experimental results show that the proposed method can achieve 38.5% mAP, which is 2.0% higher than the baseline network YOLOv5s, and can still detect the UAV aerial image well in complex scenes.

PubMed Disclaimer

Conflict of interest statement

The authors have declared that no competing interests exist.

Figures

Fig 1
Fig 1. The proposed UAV-YOLO network architecture.
Fig 2
Fig 2. AFEM module structure.
Fig 3
Fig 3. Deformable convolution process.
Fig 4
Fig 4. Feature extraction process of CA module.
Fig 5
Fig 5. SBiFPN network structure.
Fig 6
Fig 6. Multi-scale network detection process.
Fig 7
Fig 7. Comparison of training loss between UAV-YOLO and baseline methods: (a): Box loss comparison; (b): Objectness loss comparison; (c): Classification loss comparison.
Fig 8
Fig 8. Comparison of receptive fields of the two network convolution kernels.
Fig 9
Fig 9. Visualization of the focus areas of the two networks during the detection process.
Fig 10
Fig 10. Visualization of feature maps in two network detection processes.
Fig 11
Fig 11. The comparison of the detection results of the two methods in different scenarios.
Fig 12
Fig 12. Partial detection results on the VisDrone challenge test set.
Fig 13
Fig 13. Detection effect in real scenarios.
See this image and copyright information in PMC

References

    1. Girshick R. Fast R-CNN. Proceedings of 2015 IEEE International Conference on Comp-uter Vision [Internet]. 2015 April 01, 2015:[1440–8 pp.]. https://ui.adsabs.harvard.edu/abs/2015arXiv150408083G.
    1. Ren SQ, He KM, Girshick R, Sun J. Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks. IEEE Transactions on Pattern Analysis and Machine Intelligence. 2017;39(6):1137–49. doi: 10.1109/TPAMI.2016.2577031 - DOI - PubMed
    1. He KM, Gkioxari G, Dollar P, Girshick R. Mask R-CNN. IEEE Transactions on Pattern Analysis and Machine Intelligence. 2020;42(2):386–97. doi: 10.1109/TPAMI.2018.2844175 - DOI - PubMed
    1. Mittal P, Sharma A, Singh R, Dhull V. Dilated Convolution Based RCNN Using Feature Fusion for Low-Altitude Aerial Objects. Expert Syst Appl. 2022;199:14.
    1. Jin R, Lv JN, Li B, Ye JC, Lin DF. Toward Efficient Object Detection in Aerial Images Using Extreme Scale Metric Learning. IEEE Access. 2021;9:56214–27.