Abstract
<abstract>
<p>Among various aviation accidents, bird collision is one of the most common accidents for civil passenger aircraft in recent years. With the significant breakthrough of deep convolutional neural networks in the field of target detection, this paper proposes a target detection method to prevent bird collision accidents. The algorithm in this paper integrates different attention mechanisms on the YOLOv5s network to solve the problems of small target detection miss, false detection and insufficient feature extraction capability. The trend-aware loss (TAL) and trend factor (W<sub>i</sub>) are used to solve the drift of the prediction frame. After comprehensive ablation experiments, the improved algorithm shows significant improvement on the detection accuracy and speed. Results indicate that mean average precision (mAP) value reaches 99.8%, which is 6.3 percentage points higher than the original algorithm.</p>
</abstract>
Publisher
American Institute of Mathematical Sciences (AIMS)
Reference21 articles.
1. D. Aleksandra, I. Cavka, V. Cokorilo, Bird strikes on an aircraft and bird strike prevention, Tehnika, 2 (2014), 291–298. https://doi.org/10.5937/tehnika1402291D
2. A. T. Marques, H. Batalha, J. Bernardino, Bird displacement by wind turbines: assessing current knowledge and recommendations for future studies, Birds, 2 (2021), 460–475. https://doi.org/10.3390/birds2040034
3. T. Wu, X. Luo, Q. Xu, A new skeleton based flying bird detection method for low-altitude air traffic management, Chin. J. Aeronaut., 31 (2018), 2149–2164. https://doi.org/10.1016/j.cja.2018.01.018
4. X. Zhang, X. Wu, X. Zhou, X. Wang, Y. Zhang, Automatic detection and tracking of maneuverable birds in videos, in 2008 International Conference on Computational Intelligence and Security, (2008), 185–189. https://doi.org/10.1109/CIS.2008.46
5. K. A. Klein, R. Mino, M. J. Hovan, P. Antonik, G. Genello, MMW radar for dedicated bird detection at airports and airfields, in First European Radar Conference, (2004), 157–160.
Cited by
1 articles.
订阅此论文施引文献
订阅此论文施引文献，注册后可以免费订阅5篇论文的施引文献，订阅后可以查看论文全部施引文献