基于改进AlexNet飞机尾流图像识别技术研究  

Research on Aircraft Wake Turbulence Identification Technology Based on Improved AlexNet

作　　者：段英捷 潘卫军[1] 王祺 DUAN Ying-jie;PAN Wei-jun;WANG Qi(Civil Aviation Flight University of China,Guanghan Sichuan 618307,China)

机构地区：[1]中国民用航空飞行学院,四川广汉618307

出　　处：《计算机仿真》2025年第1期25-29,73,共6页Computer Simulation

基　　金：国家自然科学基金(U1733203);民航局安全能力建设项目(青年基金)(QJ2022-63)。

摘　　要：在RECAT-CN的推行背景下,为解决飞机尾流间隔过于保守而导致的跑道利用率低、航班延误时间长等问题,提出了一种基于改进AlexNet神经网络的方法以识别近地阶段飞机尾流。首先根据探测需求设计了雷达布置及参数方案,然后针对激光雷达数据进行反演得到尾流云图,同时采取了数据增强对原始尾流图像特种进行处理,并保留其背景大气特征,最后通过Hash编码特征转换得到定长数据以减轻网络计算复杂度及时间成本。试验结果表明,改进后的网络模型在识别准确率和收敛时间均优于改进前网络水平,证明上述优化算法具备识别有效性。In the context of the implementation of RECAT-CN,in order to solve the problems of low runway utilization and long flight delays caused by conservative wake turbulence intervals,this paper proposes a method based on improved AlexNetto identify wake turbulence in the near ground.Firstly,the LiDar layout and parameter scheme were designed according to the detection requirements,and then the wake images were obtained by inversing LiDar data.Meanwhile,the original wake images were processed by data enhancement and the background atmosphere characteristics were retained.Finally,the fixed length data was obtained by Hash encoding to reduce the computational complexity and time cost.The experimental results show that the improved network model has better recognition accuracy and convergence time than the previous network level,proving that the above optimization algorithm has recognition effectiveness.

关 键 词：人工智能 飞机尾流 空中交通管制 激光雷达 

分 类 号：V271.4[航空宇航科学与技术—飞行器设计] TJ760[兵器科学与技术—武器系统与运用工程]