基于GMM 的单航空公司安检区旅客到达模型

中国民航大学学报

基于GMM 的单航空公司安检区旅客到达模型

于之靖¹，包邻淋^1,2，罗谦²

（1.中国民航大学电子信息与自动化学院，天津300300；2.中国民航局第二研究所，成都610041）

收稿日期:2016-09-18 修回日期:2016-10-30 出版日期:2017-04-22 发布日期:2017-06-14
作者简介:于之靖（1963—），男，河北沧州人，教授，博士，研究方向为自动化检测技术、光纤传感技术和计算机视觉测量等.
基金资助:
国家自然科学基金项目（U1333122，U1533203）；国家科技支撑计划（2012BAG04B02）；中央高校基本科研业务费专项（3122014P003）；四川省科技支撑计划项目（2016GZ0068）；成都市战略性新兴产品研发补贴项目（2015-CP01-00158-GX）

Single airline passenger arrival model in security area based on GMM

YU Zhijing¹, BAO Linlin^1,2, LUO Qian²

（1. College of Electronic Information and Automation, CAUC, Tianjin 300300, China;2. The Second Research Institute, CAAC, Chengdu 610041, China）

Received:2016-09-18 Revised:2016-10-30 Online:2017-04-22 Published:2017-06-14

摘要/Abstract

摘要：

运用高斯混合分布模型对某航空公司安检区旅客到达分布进行拟合分析。采用极值聚类的方法获取EM求解算法的初验分布，经过多次迭代后求得参数解。在实验环节采用某航空公司真实值机输出数据进行实验验证，结果表明GMM对单航班安检区旅客到达分布的拟合精度达到90%以上，比常用拟合方法提高了15%以上。在分峰不明显的情况下，E value-GMM与K means-GMM 拟合精度相比提高了5%左右。

关键词: 单航空公司, 高斯混合分布, EM算法, 极值聚类

Abstract:

GMM（Gaussian mixture distribution model）is used for a certain airline passenger arrival distribution fitting analysis. The extremum clustering method is adopted for the premiere distribution inspection of EM algorithm,parameter solution can be got after several iterations. In experimental part, a real airline check-in output data is adopted for experimental verification. Results show that the fitting precision of Gaussian mixture distribution is 90% or more. Compared with common fitting methods, the fitting precision is improved by more than 15%.Compared with extremum clustering and commonly used K means clustering of chi-square Gaussian mixture distribution model, the extremum clustering model has higher accuracy by about 5%.

Key words: single airline, Gaussian mixture distribution, EMalgorithm, extremum clustering

中图分类号:

V351.17

于之靖，包邻淋，罗谦. 基于GMM 的单航空公司安检区旅客到达模型[J]. 中国民航大学学报.

YU Zhijing, BAO Linlin, LUO Qian. Single airline passenger arrival model in security area based on GMM[J]. .

参考文献

[1] 王勇.经济新常态下我国民航业发展的趋势和对策[J].综合运输,2015, 10(37): 7-10.

[2] 中国民用航空局. 2014 年民航行业发展统计公报[N].中国民航报,2015-07-15(001).

[3] 邢志伟, 冯文星, 罗谦, 等. 基于航班离港时刻主导的单航班离港旅客聚集模型[J].电子科技大学学报, 2015, 44(5): 719-724.

[4] PROFILLIDIS V A, BOTZORIS G N. Fuzzy and Time series Models for the Forecast of Transport Demand[C]//Fuzzy Systems Conference IEEE International, 2007, 7(3): 23-26.

[5] BANDEIRA, MICHELLEC G S P. Key indicators that affect the perception of service quality in critical airport areas of passenger boarding[J]. J Transp Lit, 2014, 8(4): 7-36.

[6] GCL BEZERRA, CF GOMES. The effects of service quality dimensions and passenger characteristics on passenger’s overall satisfaction with an airport[J]. Journal of Air Transport Management, 2015, 3(1): 77-81.

[7] AKDERE M, ETINTEMEL U, TATBUL N. Plan-Based Complex Event DetectionAcrossDistributedSources[C]//VLDB, Auckland, New Zealand,2008: 66-77.

[8] TAHERSIMA H, TAHERSIMA F, SOHANI A M, et al. Prediction of Lorenz Chaotic Time Series via Genetic Algorithm [C]//2010 IEEE International Conference on Computational Intelligence for Measurement Systems and Applications(CIMSA), 2010: 13-17.

[9] 徐定杰, 沈忱, 沈锋, 等. 混合高斯分布的变分贝叶斯学习参数估计[J].上海交通大学学报, 2013, 47（7): 1119-1125.

[10] 熊坤来, 刘章孟, 柳征, 等. 基于耘酝算法的宽带信号DOA估计及盲分离[J].电子学报, 2015, 10（10): 2028-2033.

[11] 张亚平, 贾国洋, 程绍武, 等.基于Petri 网的航站楼安检流程建模及性能分析[J].武汉理工大学学报, 2015, 39(4): 688-691, 697.

[12] 徐捷,徐从富,耿卫东, 等. 基于粗糙集理论的动态目标识别及跟踪[J].电子学报, 2003, 30（4): 605-607.

[13] 杨朝辉.基于卡方检验的杂粤砸图像道路检测算法[J]. 计算机工程与设计, 2012, 5(33): 1923-1927.