基于聚类算法的飞行航迹分析  

Flight Trajectory Analysis Based on Clustering Algorithm

作　　者：李琳丹 许雅玺[1] 张榆薪 刘坤 LI Lin-dan;XU Ya-xi;ZHANG Yu-xin;LIU Kun(College of Airport,China Civil Aviation Flight College,Guanghan 618300;School of Computer,China Civil Aviation Flight Institute,Guanghan 618300;Institute of Aeronautical Engineering,China Academy of Civil Aviation Flight,Guanghan 618300)

机构地区：[1]中国民用航空飞行学院机场学院,广汉618300 [2]中国民用航空飞行学院计算机学院,广汉618300 [3]中国民用航空飞行学院航空工程学院,广汉618300

出　　处：《现代计算机》2018年第19期3-7,共5页Modern Computer

基　　金：国家自然科学基金民航联合基金(No.U1233105);中国民航飞行学院大学生创新创业训练计划项目(No.201710624049)

摘　　要：为了实现空域资源高效利用,增强空中交通智慧化管理,减轻管制员负荷。利用AP算法基于历史飞行数据对航迹进行聚类,传统聚类相似性度量采用欧氏距离,针对航迹非等长特性,利用动态时间规整解决非等长距离输入。利用离散余弦变换对航迹时序列降噪,以得到更佳的聚类效果。结果显示:291条飞行航迹数据集,共划分出7个聚类,其聚类效果也较好。得出结论:基于近邻传播算法的航迹聚类能进行快速有效地数据处理,同时解决航迹非等长问题,增强航迹聚类的有效性和适用性。In order to realize efficient utilization of airspace resources, enhance intelligent management of air traffic and reduce the load of controllers. Track to make use of the AP algorithm based on the history of flight data clustering, traditional clustering similarity measure using Euclide - an distance, in vie,＊- of the track is the isometric characteristics, using the Dynamic Time neat （Dynamic Time Warping） long distance with unequal input. The Discrete Cosine Transform ,＊-as used to reduce the noise of the track sequence so as to get a better clustering effect. The results showed that 291 flight path data sets were divided into 7 clusters, and the clustering results were better. It is concluded that the near neighbor propagation algorithm can be used for the rapid and effective data processing, and it can solve the problem of non-equal track length, and enhance the validity and applicability of track clustering.

关 键 词：飞行航迹 AP算法 动态时间规整 离散余弦变换 

分 类 号：V355.1[航空宇航科学与技术—人机与环境工程] TP391.41[自动化与计算机技术—计算机应用技术]