机场跑道异物的存在不仅对飞机的滑行和起降构成潜在威胁，还可能导致航班延误、起飞中断等问题。因此，开发智能化的机场跑道异物检测系统对于保障航空安全以及确保航空运输的顺畅具有至关重要意义。当前，我国在此领域的技术研发尚处于初级阶段，而进口的检测系统不仅成本高昂，且关键技术不透明。此外，现阶段我国大多数机场依然采用人工定时巡视的方式对异物进行检查和清除。该方式不仅效率低下，耗时较长，而且可靠性较差。因此，开发一种可靠、高效且准确的异物检测系统，是提升国家空域安全保障能力的重要基石，也是我国民航科技发展的必然选择。与基于毫米波雷达的异物检测系统相比，基于图像的检测系统具有丰富的异物特征信息和部署灵活性强等优点。因此，本文主要研究基于图像的机场跑道异物检测系统。考虑到车载式检测系统具有高灵活性和强适应性的优点，本文在巡检车上安装工业相机及定位装置，构成一个移动式跑道异物检测系统，能在短时间内完成对预定区域的检查。本文的主要内容如下：

（1）搭建基于视觉感知的机场跑道异物检测系统平台。首先分析了机场跑道异物检测系统的工作环境，并结合实际巡检情况进行了需求分析，然后设计了系统巡检方案。在此基础上，构建了基于视觉感知的机场跑道异物检测系统平台，该平台为实现机场跑道异物的高效、准确检测提供了技术支撑。

（2）构建面向机场跑道的异物图像生成模型。针对异物图像数据稀缺的问题，构建一种面向机场跑道的异物图像生成模型(FOD-GAN)。该模型引入注意力机制模块和轻量级特征编码单元，能够有效捕获图像的关键特征信息，并迅速定位真实图像的重点区域，实现高质量异物图像的生成。此外，引入梯度惩罚损失函数以增强模型训练的稳定性。通过定性和定量评估，证明了所提方法相较于传统数据扩增技术具有显著优势，能够有效提升异物目标检测模型的检测精度。

（3）构建面向机场跑道的异物目标检测模型。针对异物检测模型检测性能与模型复杂度之间的矛盾，充分考虑入侵异物目标体积小，跑道背景复杂多变的特点，本文提出了一种面向机场跑道的异物目标检测模型。首先，设计了一种多尺度特征融合的轻量级主干网络，该网络能够在兼顾大尺度全局信息与小尺度局部信息的同时进行快速推理。其次，针对异物图像因背景复杂而难以有效表达异物目标特征的问题，通过嵌入无参注意力机制来增强模型的特征表达能力。进一步地，引入基于SIoU的损失函数以优化检测效果。最后，采用基于先验知识的训练策略进行训练，以提高模型的检测精度和泛化能力。实验结果表明本文模型在模型复杂度、检测精度之间取得了更好的均衡。

（4）针对内存和功耗有限的车载嵌入式平台，本文研究并实现了基于TensorRT的异物目标检测模型的部署与优化。在保持检测精度的同时显著提高了推理速度，为实现机场跑道异物的高效、可靠检测提供了重要支撑。

本文提出的基于视觉感知的机场跑道异物检测系统不仅为保障航空安全奠定了坚实的基础，并且推动了人工智能技术与航空领域的深度融合，对实现智慧机场建设具有重要的实际意义。

The presence of foreign objects on airport runways not only poses a potential threat to aircraft taxiing and landing, but also may lead to flight delays, take-off interruptions and other problems. Therefore, the development of intelligent airport runway foreign matter detection system is of vital significance to safeguard aviation safety and ensure the smoothness of air transport. Currently, China's technological research and development in this field is still in the primary stage, and the imported detection systems are not only costly, but also opaque in terms of key technologies. In addition, most of the airports in China still use manual inspection to check and remove foreign objects. This method is not only inefficient and time-consuming, but also less reliable. Therefore, the development of a reliable, efficient and accurate foreign object detection system is an important cornerstone for enhancing the safety and security capability of national airspace, and is also an inevitable choice for the development of civil aviation science and technology in China. Compared with the millimetre wave radar-based foreign object detection system, the image-based detection system has the advantages of rich foreign object feature information and high deployment flexibility. Therefore, this paper focuses on the image-based foreign object detection system for airport runways. Considering that the vehicle-mounted detection system has the advantages of high flexibility and strong adaptability, this paper installs an industrial camera and a positioning device on the inspection vehicle to constitute a mobile runway foreign object detection system, which can complete the inspection of the predetermined area in a short period of time. The main contents of this paper are as follows:

Build the platform of airport runway foreign object detection system based on visual perception. Firstly, the working environment of the airport runway foreign matter detection system is analysed, and the demand analysis is carried out in combination with the actual inspection situation, and then the system inspection scheme is designed. On this basis, the visual perception-based airport runway foreign object detection system platform is constructed, which provides technical support for achieving efficient and accurate detection of airport runway foreign objects.

(2) Construct a foreign object image generation model for airport runway. Aiming at the problem of scarcity of foreign object image data, a foreign object image generation model for airport runways (FOD-GAN) is constructed. The model introduces an attention mechanism module and a lightweight feature coding unit, which can effectively capture the key feature information of the image and quickly locate the key areas of the real image to achieve high-quality foreign object image generation. In addition, a gradient penalty loss function is introduced to enhance the stability of model training. Through qualitative and quantitative evaluations, it is demonstrated that the proposed method has significant advantages over traditional data augmentation techniques and can effectively improve the detection accuracy of the foreign object target detection model.

(3) Construct a foreign object target detection model for airport runways. Aiming at the contradiction between the detection performance of foreign object detection model and model complexity, and taking into full consideration of the small size of intruding foreign object targets and the complexity and variability of the runway background, this paper proposes a foreign object target detection model for airport runways. Firstly, a lightweight backbone network with multi-scale feature fusion is designed, which is capable of fast inference while taking into account large-scale global information and small-scale local information. Secondly, to address the problem that foreign object images are difficult to effectively express foreign object target features due to the complex background, the feature expression capability of the model is enhanced by embedding a parameter-free attention mechanism. Further, a SIoU-based loss function is introduced to optimise the detection effect. Finally, a training strategy based on prior knowledge is used for training to improve the detection accuracy and generalisation ability of the model. The experimental results show that the model in this paper achieves a better balance between model complexity, and detection accuracy.

 (4) Aiming at the on-board embedded platform with limited memory and power consumption, this paper investigates and implements the deployment and optimisation of a foreign object target detection model based on TensorRT. It significantly improves the inference speed while maintaining the detection accuracy, which provides an important support for achieving efficient and reliable detection of foreign objects on airport runways.

The airport runway foreign object detection system based on visual perception proposed in this paper not only lays a solid foundation for safeguarding aviation safety, but also promotes the in-depth integration of artificial intelligence technology with the field of aviation, which is of great practical significance for realising the construction of smart airports.

[1]国家民航总局机场司民航安全技术中心. FOD 防范手册[R]. 2009.

[2]PATTERSON J R J. Foreign object debris (FOD) detection research [J]. International Airport Review,2008(2):22-27.

[3]国际民用航空组织.ICAO: Global Accident Rate Continues Downward Trend Report [EB/OL].https://www.mro-network.com/safety-regulatory/global-accident-rate-continues-downward-trend,2017-8-29.

[4]网易航空(北京)."天眼"FOD 系统实现无死角监测，跑道异物无处遁形[EB/OL].http://news.163.com/air/18/0103/10/D77H07P1000181O6.html.2018-01-03

[5]上海市人民政府.虹桥机场 FOD 检测设备处于国内领先[EB/OL].http://www.shanghai.gov.cn/nw2/nw2314/nw2315/nw31406/u21aw1039857.html.2015-8-3.

[6]S. Jain, M.S. Prasad, R. Gutpa, A Comparison of Manual and Automotive FOD Detection Systems at Airport Runways, in: 2022 10th International Conference on Reliability, Infocom Technologies and Optimization (Trends and Future Directions)(ICRITO), 2022, pp. 1-5.

[7]Stratech, iFerret system [Online]. Available: http://www.stratechsystems.com/iv_iferret. asp.

[8]Tarsier-QinetiQ Automatic Runway Debris Detection 2009 [Online]. Available: http://www.tarsier.qinetiq.com/solution/Pages/default.aspx

[9]Xsight-Advanced Radar and Optic Sensors for FOD Detection and Homeland Security[Online]. Available: http://www.xsightsys.com

[10]FOD Finder system [Online]. Available:http://www.fodfinder.com

[11]赛英公司FOD监测雷达系统[OL]. http://www.docin.com/p-1454826112.html.

[12]高宏伟,王慧科,李卓.一种改进的机场跑道异物检测算法研究[J].计算机科学,2015,42(S1):205-208.

[13]赵军香,梁兴东,李焱磊.一种基于图像差值的跑道异物检测算法[J].国外电子测量技术,2016,35(09):17-23.

[14]于之靖,陶永奎,郑建文等.基于小波变换与形态学的机场跑道异物检测算法[J].科学技术与工程,2020,20(21):8690-8695.

[15]张怡,孙永荣,刘梓轩等.基于RTK定位的图像差分跑道异物检测[J].电子测量与仪器学报,2020,34(10):51-56.

[16]Jing Y, Zheng H, Lin C, et al. Foreign object debris detection for optical imaging sensors based on random forest[J]. Sensors, 2022, 22(7): 2463.

[17]吴静,王洪,汪学刚.机场跑道异物监测雷达的杂波图恒虚警率检测[J].计算机应用,2013,33(11):3288-3290+3308.

[18]王宝帅,兰竹,李正杰等.毫米波雷达机场跑道异物分层检测算法[J].电子与信息学报,2018,40(11):2676-2683.

[19]Feil P, Menzel W, Nguyen T P, et al. Foreign objects debris detection (FOD) on airport runways using a broadband 78 GHz sensor[C]//2008 38th European Microwave Conference. IEEE, 2008: 1608-1611.

[20]Qin F, Bu X, Zeng Z, et al. Small target detection for fod millimeter-wave radar based on compressed imaging[J]. IEEE Geoscience and Remote Sensing Letters, 2021, 19: 1-5.

[21]Wang Y, Song Q, Wang J, et al. Airport runway foreign object debris detection system based on arc-scanning SAR technology[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022, 60: 1-16.

[22]Zhong J, Zhang K, Zeng Q, et al. A False Alarm Elimination Algorithm of Foreign Objects Debris Detection Based on Duffing Oscillator[J]. IEEE Access, 2022, 10: 7588-7597.

[23]李小军,邓月明,陈正浩等.改进YOLOv5的机场跑道异物目标检测算法[J].计算机工程与应用,2023,59(02):202-211.

[24]兰庭信,蒋进,尚帅等.机场跑道异物检测与定位技术研究[J].电光与控制,2021,28(09):75-79.

[25]郭晓静,隋昊达.改进YOLOv3在机场跑道异物目标检测中的应用[J].计算机工程与应用,2021,57(08):249-255.

[26]Cao X, Gong G, Liu M, et al. Foreign object debris detection on airfield pavement using region based convolution neural network[C]//2016 International Conference on Digital Image Computing: Techniques and Applications (DICTA). IEEE, 2016: 1-6.

[27]Ren M, Wan W, Yu Z, et al. Bidirectional YOLO: improved YOLO for foreign object debris detection on airport runways[J]. Journal of Electronic Imaging, 2022, 31(6): 063047-063047.

[28]Jing Y, Zheng H, Zheng W, et al. A pixel-wise foreign object debris detection method based on multi-scale feature inpainting[J]. Aerospace, 2022, 9(9): 480.

[29]Noroozi M, Shah A. Towards optimal foreign object debris detection in an airport environment[J]. Expert Systems with Applications, 2023, 213: 118829.

[30]Xu, Haoyu, Zhenqi Han, Songlin Feng, Han Zhou, and Yuchun Fang. "Foreign object debris material recognition based on convolutional neural networks." EURASIP Journal on Image and Video Processing 2018 (2018): 1-10.

[31]Sharma Vipul, Mir Roohie Naaz. A comprehensive and systematic look up into deep learning based object detection techniques: A review[J]. Computer Science Review, 2020, 38:100301.

[32]朱晓慧,钱丽萍,傅伟. 图像数据增强技术研究综述[J].软件导刊,2021,20(05):230-236.

[33]Kingma D P, Welling M. Auto-encoding variational bayes[J]. arXiv preprint arXiv:1312.6114, 2013.

[34]Saharia C, Chan W, Chang H, et al. Palette: Image-to-image diffusion models[C]//ACM SIGGRAPH 2022 conference proceedings. 2022: 1-10.

[35]Croitoru F A, Hondru V, Ionescu R T, et al. Diffusion models in vision: A survey[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023.

[36]林本旺,赵光哲,王雪平,李昊. 基于组残差块生成对抗网络的面部表情生成[J/OL].计算机工程与应用,2023:1-12.

[37]王通,陈延彬.基于改进生成对抗网络的动液面建模数据扩充[J].电子测量与仪器学报,2023,37(02):99-109.

[38]王泽,姜斌,程月华,张香燕,薛琪. 一种基于生成对抗网络的卫星异常检测方法[J].空间控制技术与应用,2023,49(01):113-120.

[39]姜世博,孙跃文,许硕,吴志芳. 基于生成对抗网络的CT图像金属伪影校正[J].哈尔滨工程大学学报,2022,43(12):1766-1771.

[40]范黎. 基于生成对抗网络的图像数据增强技术研究及应用[D]. 浙江：浙江大学, 2022.

[41]Jia Deng, Wei Dong, Socher Richard, Li Jia Li, Kai Li, Li Fei-Fei. ImageNet: A large-scale hierarchical image database[C]//2009 IEEE Conference on Computer Vision and Pattern Recognition. IEEE, 2009: 248-255.

[42]Francisco J. Moreno-Barea, Fiammetta Strazzera, Jerez JoseM. Daniel Urda, Leonardo Franco. Forward Noise Adjustment Scheme for Data Augmentation[C]// IEEE Symposium Series on Computational Intelligence (IEEE SSCI 2018). IEEE, 2018:728-734.

[43]Zhong Zhun, Zheng Liang, Kang Guoliang, Li Shaozi, Yang Yi. Random Erasing Data Augmentation[C]//Proceedings of the AAAI conference on artificial intelligence. 2020, 34(07): 13001-13008.

[44]Shorten C, Khoshgoftaar T M. A survey on image data augmentation for deep learning[J]. Journal of big data, 2019, 6(1): 1-48.

[45]Goodfellow I, Pouget-Abadie J, Mirza M, et al. Generative adversarial nets[J]. Advances in neural information processing systems, 2014, 27.

[46]Hinz T, Fisher M, Wang O, et al. Improved techniques for training single-image gans[C]//Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision. 2021: 1300-1309.

[47]Shaham T R, Dekel T, Michaeli T. Singan: Learning a generative model from a single natural image[C]//Proceedings of the IEEE/CVF international conference on computer vision. 2019: 4570-4580.

[48]Brock A, Donahue J, Simonyan K. Large scale GAN training for high fidelity natural image synthesis[J]. arXiv preprint arXiv:1809.11096, 2018.

[49]Zhu J Y, Park T, Isola P, et al. Unpaired image-to-image translation using cycle-consistent adversarial networks[C]//Proceedings of the IEEE international conference on computer vision. 2017: 2223-2232.

[50]Radford A, Metz L, Chintala S. Unsupervised representation learning with deep convolutional generative adversarial networks[J]. arXiv preprint arXiv:1511.06434, 2015.

[51]Woo S, Park J, Lee J Y, et al. Cbam: Convolutional block attention module[C]//Proceedings of the European conference on computer vision (ECCV). 2018: 3-19.

[52]Wang Q, Wu B, Zhu P, et al. ECA-Net: Efficient channel attention for deep convolutional neural networks[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2020: 11534-11542.

[53]Arjovsky M, Chintala S, Bottou L. Wasserstein generative adversarial networks[C]//International conference on machine learning. PMLR, 2017: 214-223.

[54]Chen C, Liu M Y, Tuzel O, et al. R-CNN for small object detection[C]//Computer Vision–ACCV 2016: 13th Asian Conference on Computer Vision, Taipei, Taiwan, November 20-24, 2016, Revised Selected Papers, Part V 13. Springer International Publishing, 2017: 214-230.

[55]Torralba A, Fergus R, Freeman W T. 80 million tiny images: A large data set for nonparametric object and scene recognition[J]. IEEE transactions on pattern analysis and machine intelligence, 2008, 30(11): 1958-1970.

[56]冯彦卿. 基于深度学习的视觉目标检测与跟踪技术研究[D].湖南:国防科技大学,2020.

[57]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.

[58]Girshick R. Fast r-cnn[C]//Proceedings of the IEEE international conference on computer vision. 2015: 1440-1448.

[59]Ren S, He K, Girshick R, et al. Faster r-cnn: Towards real-time object detection with region proposal networks[J]. Advances in neural information processing systems, 2015, 28.

[60]张帆,郭思媛,任方涛,张新红,李结平.基于改进YOLOv3的玉米叶片气孔自动识别与测量方法[J].农业机械学报,2023,54(02):216-222.

[61]白一睿,方辉,张泽.基于YOLO神经网络模型的花生智能精选系统设计[J].机械,2023,50(02):1-6.

[62]陈德海,孙仕儒,王昱朝,雷志军.改进SSD算法的小目标检测研究[J].传感器与微系 统,2023,42(03):65-68+72.

[63]范明凯,肖满生,胡一凡,吴宇杰.基于改进SSD算法的胃部息肉图像检测[J].软件工程,2023,26(03):30-35.

[64]沈翔. 基于RetinaNet的小目标检测提升方法研究[D].南京：南京邮电大学,2022.

[65]宋建熙,李兴科,于哲,李西兵.改进Retina-Net的草坪杂草目标检测[J].中国农机化学报,2022,43(12):170-177.

[66]Joseph Redmon,Santosh Kumar Divvala,Ross B. Girshick,Ali Farhadi. You Only Look Once: Unified, Real-Time Object Detection.[J]. CoRR,2015,abs/1506.02640.

[67]Gao S H, Cheng M M, Zhao K, et al. Res2net: A new multi-scale backbone architecture[J]. IEEE transactions on pattern analysis and machine intelligence, 2019, 43(2): 652-662.

[68]Wang C Y, Liao H Y M, 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: 390-391.

[69]Gevorgyan Z. SIoU loss: More powerful learning for bounding box regression[J]. arXiv preprint arXiv:2205.12740, 2022.

[70]Yang L, Zhang R Y, Li L, et al. Simam: A simple, parameter-free attention module for convolutional neural networks[C]//International conference on machine learning. PMLR, 2021: 11863-11874.

[71]Han K, Wang Y, Tian Q, et al. Ghostnet: More features from cheap operations[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2020: 1580-1589.

[72]Ma N, Zhang X, Zheng H T, et al. Shufflenet v2: Practical guidelines for efficient cnn architecture design[C]//Proceedings of the European conference on computer vision (ECCV). 2018: 116-131.

[73]Howard A, Sandler M, Chu G, et al. Searching for mobilenetv3[C]//Proceedings of the IEEE/CVF international conference on computer vision. 2019: 1314-1324.

[74]陈胤杰. 基于嵌入式与YOLOv4算法的无人机实时目标识别系统研究[D].电子科技大学,2022.

[75]施一飞.对使用TensorRT加速AI深度学习推断效率的探索[J].科技视界,2017,(31):26-27.