Bird strikes are considered a significant threat to civil aviation flight safety and economic operations. Real-time accurate detection of flying birds in the low altitude region of airports is regarded as a key technology to prevent bird strike accidents. However, the existing algorithms are often limited by insufficient accuracy in detecting minimal flying bird instances. To address this issue, an algorithm named ABRT-DETR(AirBirds RT-DETR) was proposed, which is based on the RT-DETR-R18 framework. Firstly, a new BasicBlock-PCR module was constructed and used instead of the original BasicBlock to enhance the feature extraction capability of flying bird instances, the model complexity and computational burden were significantly reduced while the precision was improved. Second, HS-FPN (high-level screening-feature fusion pyramid) was introduced, the high-level features are used to screen and fuse the low-level feature information, which enhances the expression effect of multi-level features, thus further improving the detection accuracy and reducing the computational burden. Finally, the original GIoU loss function was replaced by EIoU, optimizing the localization accuracy and overall detection performance of flying bird targets. The experimental results show that compared with the original RT-DETR-R18 model, the precision, recall and mAP50 are improved by 3.8%,2.5% and 3.2%, respectively, and the model’s parameters and FLOPs are greatly reduced, which provides an efficient and reliable technological solution for low-altitude bird monitoring at airports.
选取主要通过精度(precision,P)、召回率(recall,R)和平均精度均值(mean average precision,mAP)来评估。精度指在模型识别的图片中正样本被正确预测的图片所占比例。召回率指被正确识别的图片占测试集全部图片的比例。mAP表示所有类别的平均精密度的平均值。精度、召回率和mAP的相应公式为
LiJ, LiY L, YuX C, et al. Design of aircraft structures against the threat of bird strikes[J]. Chinese Journal of Aeronautics, 2018, 31(7): 1535-1558.
[2]
DolbeerR A, BegierM J, MillerP R, et al. Wildlife strikes to civil aircraft in the United States (1990—2023)[M]. Washington: Federal Aviation Administration, 2023.
[3]
Lopez-LagoM, CasadoR, BermudezA, et al. A predictive model for risk assessment on imminent bird strikes in airport areas[J]. Aerospace Science and Technology, 2017, 62: 19-30.
WuChenxi, YingBaosheng, XuXiaowei, et al. Road small object detection algorithm based on improved single-step multi-box object detection[J]. Science Technology and Engineering, 2023, 23(5): 2051-2058.
ZhanChengxiang, MengQingyan, AnJianjian, et al. Weather adaptability of single-point multi-box aircraft target detection based on clear-SSD[J]. Science Technology and Engineering, 2020, 20(31): 12717-12723.
MouJiayu, NanXinyuan. Real-time detection algorithm for small traffic sign objects based on improved YOLOv7-tiny[J]. Science Technology and Engineering, 2024, 24(30): 13072-13079.
LiZhi’ang, XiaoXiaoling, ZhouShaofa. Small target ship detection in remote sensing images based on improved YOLOv5s[J]. Science Technology and Engineering, 2025, 25(2): 657-666.
ZhengHan, TianMeng, ZhaoYanfeng, et al. Hand pose estimation method based on improved Faster R-CNN[J]. Science Technology and Engineering, 2023, 23(3): 1160-1167.
[18]
LiangH, ZhangX, KongJ, et al. SMB-YOLOv5: a lightweight airport flying bird detection algorithm based on deep neural networks[J]. IEEE Access, 2024, 12: 84878-84892.
[19]
HongS J, HanY, KimS Y, et al. Application of deep-learning methods to bird detection using unmanned aerial vehicle imagery[J]. Sensors, 2019, 19(7). DOI:10.3390/s19071651.
[20]
VaswaniA. Attention is all you need[C]//Advances in Neural Information Processing Systems 30. Long Beach: Curran Associa-tes, Inc., 2017: 5998-6008.
[21]
CarionN, MassaF, SynnaeveG, et al. End-to-end object detection with transformers[C]//European Conference on Computer Vision. Cham: Springer International Publishing, 2020: 213-229.
[22]
Dai XChenY, YangJ, et al. Dynamic DETR: end-to-end object detection with dynamic attention[C]// Proceedings of the IEEE/CVF International Conference on Computer Vision. New York: IEEE, 2021: 2988-2997.
[23]
LiuS L, LiY L, QuJ Y, et al. Airport UAV and birds detection based on deformable DETR[J]. Journal of Physics: Conference Series, 2022, 2253(1): 1-11.
[24]
ZhaoY, LvW, XuS, et al. DETRs beat YOLOs on real-time object detection[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2024: 16965-16974.
[25]
ZhangH, MaZ, LiX, et al. RS-DETR: an improved remote sensing object detection model based on RT-DETR[J]. Applied Sciences, 2024, 14(22). DOI: 10.3390/app142210331.
[26]
GaoS, XueF, TanW, et al. RTPR-DETR: a real-time small object detection network based on partial-convolution with structural re-parameterization[C]//36th Chinese Control and Decision Conference (CCDC). New York: IEEE, 2024: 5513-5518.
[27]
ChenY, ZhangC, ChenB, et al. Accurate leukocyte detection based on deformable-DETR and multi-level feature fusion for aiding diagnosis of blood diseases[J]. Computers in Biology and Medicine, 2024, 170: 1-10.
[28]
ChenJ, KaoS H, HeH, et al. Run, Run, don’t walk: chasing higher FLOPS for faster neural networks[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2023: 12021-12031.
[29]
DingX, ZhangX, MaN, et al. RepVGG: making VGG-style convnets great again[C]//Proceedings of the IEEE/CVF Confe-rence on Computer Vision and Pattern Recognition. New York: IEEE, 2021: 13733-13742.
[30]
ZhangY F, RenW, ZhangZ, et al. Focal and efficient IoU loss for accurate bounding box regression[J]. Neurocomputing, 2022, 506: 146-157.
[31]
SunH, WangY, CaiX, et al. Airbirds: a large-scale challenging dataset for bird strike prevention in real-world airports[C]//Proceedings of the Asian Conference on Computer Vision. New York: IEEE, 2022: 2440-2456.