自注意力特征下机场跑道周界异物入侵预警  

Airport Runway Perimeter Foreign Object Intrusion Warning Based on Self Attention Features

作　　者：王涛 冯凯 WANG Tao;FENG Kai(Guangzhou Baiyun International Airport Construction and Development Co.,LTD.Guangzhou Guangdong 510080,China;Heilongjiang Zhongyi Communication Technology Engineering Co.,LTD.Harbin Heilongjiang 150000,China)

机构地区：[1]广州白云国际机场建设发展有限公司,广东广州510080 [2]黑龙江中移通信技术工程有限公司,黑龙江哈尔滨150000

出　　处：《计算机仿真》2024年第12期88-92,共5页Computer Simulation

基　　金：广州白云国际机场三期扩建工程T3航站楼及货运区民航弱电系统工程(一标段)(23-37-0298-0)。

摘　　要：机场跑道上的异物种类繁多,且这些异物的形状、大小、颜色、材质各不相同,使得根据图像内容聚焦机场跑道的关键区域变得困难,难以捕捉到这些细微的差异,从而导致机场跑道周界异物入侵预警的精度较低。为此,提出机场跑道周界异物入侵自注意力特征预警仿真。通过生成对抗网络对机场跑道样本展开扩充处理,以解决数据稀疏性与不平衡问题;引入基于自注意力机制模块至机场跑道周界异物入侵检测YOLOv5模型内,动态加权不同区域的特征,以帮助模型更准确地识别出图像中的关键细微特征;利用YOLOv5模型对机场跑道实时监控,一旦检测到异物,立即发出预警,提高异物入侵检测的精度。实验结果表明,所提方法的机场跑道周界异物入侵检测精度更高、准确率更好、且更适合实际应用。There are many kinds of foreign objects on the airport runway,and the shape,size,color and material of these foreign objects are different,which makes it difficult to focus on the key areas of the airport runway according to the image content,and it is difficult to capture these subtle differences,thus leading to the low accuracy of foreign object intrusion early warning around the airport runway.For this reason,the self attention feature early warning simulation of foreign matter intrusion around the airport runway is proposed.The problem of data sparsity and imbalance is solved by generating countermeasures network to expand the airport runway samples.The self attention mechanism module is introduced into YOLOv5 model of airport runway perimeter foreign object intrusion detection,and the features of different regions are dynamically weighted to help the model more accurately identify the key subtle features in the image.The YOLOv5 model is used to monitor the airport runway in real time.Once a foreign object is detected,an early warning is sent immediately to improve the accuracy of foreign object intrusion detection.The experimental results show that the proposed method is more accurate,more accurate and more suitable for practical application.

关 键 词：异物入侵检测 自注意力机制 生成对抗网络 机场跑道 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]