基于KNN和形态学的飞机尾涡区检测方法  

Aircraft Wake Detection Method Based on KNN and Morphology

作　　者：邓蕾蕾 潘卫军[1] 崔烁 潘璇 DENG Lei-Lei;PAN Wei-Jun;CUI Shuo;PAN Xuan(Civil Aviation Flight University of China,Guanghan 618000,China)

机构地区：[1]中国民用航空飞行学院,四川广汉618000

出　　处：《航空计算技术》2024年第5期48-52,共5页Aeronautical Computing Technique

基　　金：国家自然科学基金项目资助(U1733203);民航局安全能力建设项目资助(TM2018-9-1/3)。

摘　　要：为了提升空中交通管理系统的智能化水平,解决晴空条件下飞机尾流检测问题,提出了一种结合多普勒激光雷达技术的飞机尾流检测方法。基于K-最近邻算法(KNN)和图像形态学处理技术,旨在提高尾流检测的精度和可靠性。使用多普勒激光雷达对机场空域进行扫描,获取风场数据;通过动态滑动窗口生成可能包含尾流的候选区域;采用顶帽和黑帽形态学运算提取尾流特征,并将其输入KNN算法进行检测。实验结果表明,所提出的方法在精度、召回率和F1 Score方面,相较于基于尾流速度极差特征法的检测方法,分别提高了22.58%、9.29%和14.22%,有效提升了尾流检测性能,为管制员提供了更为可靠的辅助决策支持。In order to improve the intelligence of the air traffic management system,the problem of aircraft wake turbulence detection is solved.Combined with Doppler lidar technology,an aircraft wake turbulence detection method based on K Nearest Neighbour(KNN)algorithm and image morphological processing technology is proposed.Firstly,the radar is used to scan the inbound and outbound airspace of airport to obtain the wind field data of the target region.Secondly,the dynamic sliding window with different window sizes is used to generate the proposed candidate regions for the radial velocity field of the acquired aircraft wake stream.Finally,the morphological features of the wake stream in the candidate regions(ROIs)are extracted into the K Nearest Neighbour algorithm using the top hat and black hat morphology operations to detect them and are compared with those of the same conditions using the detection method based on the wake velocity extreme difference feature method.The experimental results show that the accuracy of the proposed aircraft wake detection method is 22.58%,9.29%and 14.22%higher than the accuracy,recall and F1 score of the speed polarimetric based method,respectively,and the method can provide decision support for the controllers.

关 键 词：尾涡检测 形态学 K最近邻算法 目标检测 多普勒激光雷达 可视化 

分 类 号：V328[航空宇航科学与技术—人机与环境工程]