基于改进密度聚类与模式信息挖掘的异常轨迹识别方法

doi:10.11959/j.issn.1000-436x.2017287

通信学报 ›› 2017, Vol. 38 ›› Issue (12): 21-33.doi: 10.11959/j.issn.1000-436x.2017287

基于改进密度聚类与模式信息挖掘的异常轨迹识别方法

何明^1,²,仇功达¹,周波¹,柳强^1,³,曹玉婷¹

¹ 陆军工程大学指挥控制工程学院，江苏南京 210007
² 解放军第61所，北京 100000
³ 海军指挥学院，江苏南京210007

修回日期:2017-11-12 出版日期:2017-12-01 发布日期:2018-01-19
作者简介:何明（1978-），男，新疆石河子人，博士，陆军工程大学教授，主要研究方向为传感器网络。|仇功达（1992-），男，浙江余姚人，陆军工程大学硕士生，主要研究方向为机器学习。|周波（1982-），男，江苏淮安人，陆军工程大学博士生，主要研究方向智能信息处理。|柳强（1983-），男，辽宁锦州人，博士，海军指挥学院讲师，主要研究方向指挥控制系统工程。|曹玉婷（1989-），女，江苏南京人，陆军工程大学硕士生，主要研究方向为智能信息处理。
基金资助:
江苏省自然科学基金资助项目(BK20150721);江苏省自然科学基金资助项目(BK20161469);中国博士后基金资助项目(2015M582786);中国博士后基金资助项目(2016T91017);江苏省重点研发计划基金资助项目(BE2015728);江苏省重点研发计划基金资助项目(BE2016904);江苏省科技基础设施建设计划基金资助项目(BM2014391);国家重点研发计划基金资助项目(2016YFC0800606)

Abnormal trajectory detection method based on enhanced density clustering and abnormal information mining

Ming HE^1,²,Gong-da QIU¹,Bo ZHOU¹,Qiang LIU^1,³,Yu-ting CAO¹

¹ College of Command and Control Engineering，Army Engineering University of PLA，Nanjing 210007，China
² The 61st Research Institute of PLA，Beijing 100000，China
³ Naval Command College，Nanjing 210007，China

Revised:2017-11-12 Online:2017-12-01 Published:2018-01-19
Supported by:
The Natural Science Foundation of Jiangsu Province(BK20150721);The Natural Science Foundation of Jiangsu Province(BK20161469);China Postdoctoral Science Foundation(2015M582786);China Postdoctoral Science Foundation(2016T91017);The Primary Research ＆ Development Plan of Jiangsu Province(BE2015728);The Primary Research ＆ Development Plan of Jiangsu Province(BE2016904);The Engineering Research Center of Jiangsu Province(BM2014391);The National Key Research and Development Program of China(2016YFC0800606)

摘要/Abstract

摘要：

针对社会安全事件中异常行为信息识别挖掘难等问题，提出一种基于改进密度聚类与模式信息挖掘的异常轨迹识别方法。首先，针对采样问题，结合Hausdorff距离思想重新定义一种改进型DTW距离，用于描述轨迹具体行为，而MBR距离下的延伸定义，则用于描述轨迹覆盖区域热度。其次，在CFSFDP算法的密度关联与决策模型下，基于支持向量机回归（SVR，support vector regression）提出了特定支持向量机回归（SSVR，specific support vector regression），利用针对性改良下的回归差异非线性识别类中心，实现类的智能识别。最后，通过2种密度下的类识别，实现更多异常模式信息的挖掘与3种异常轨迹识别。结合上海市与北京市出租车轨迹集进行了仿真实验与数据分析，验证了算法在轨迹聚类异常识别方面的有效性。与传统方法相比，类发现能力提高了10%，异常轨迹信息得以区别与丰富。

关键词: 支持向量机回归, 密度聚类, 异常轨迹识别, 模式信息挖掘

Abstract:

Aiming at problems of low accuracy in the recognition and difficulty in enriching the information of abnormal behavior in the social security incidents，an abnormal trajectory detection method based on enhanced density clustering and abnormal information mining was proposed.Firstly，combined with Hausdorff distance，an enhanced DTW distance aiming at the problem of sampling to describe the behavior in detail was proposed.And based on the MBR distance， some definitions to describe the geographical distribution of trajectory were proposed.Secondly，with the density-distance decision model of CFSFDP algorithm，intelligent recognition of cluster was realized by using the difference of SSVR which was proposed based on SVR.Finally，based on the analysis of distribution under the two kinds of density，more abnormal information could be mined，three kinds of abnormal trajectories would be recognized.And the simulation results on trajectory data of Shanghai and Beijing verify that the algorithm is objective and efficient.Comparing to existing method，accuracy in the clustering is promoted by 10%，and the abnormal trajectories are sorted， abnormal information is enriched.

Key words: SVR, density clustering, abnormal trajectory detection, pattern information mining

中图分类号:

TP301

何明,仇功达,周波,柳强,曹玉婷. 基于改进密度聚类与模式信息挖掘的异常轨迹识别方法[J]. 通信学报, 2017, 38(12): 21-33.

Ming HE,Gong-da QIU,Bo ZHOU,Qiang LIU,Yu-ting CAO. Abnormal trajectory detection method based on enhanced density clustering and abnormal information mining[J]. Journal on Communications, 2017, 38(12): 21-33.

图/表 13

图1

图2

表1

图3

图4

图5

图6

图7

图8

图9

图10

图11

图12

参考文献 19

[1]	ZHENG Y . Trajectory data mining:an overview[J]. ACM Transactions on Intelligent Systems and Technology, 2015,6(3).
[2]	毛嘉莉, 金澈清, 章志刚 ,等. 轨迹大数据异常检测:研究进展及系统框架[J]. 软件学报, 2017,28(1): 17-34.
	MAO J L , JIN C Q , ZHANG Z G ,et al. Anomaly detection for trajectory big data:advancements and framework[J]. Journal of Software, 2017,28(1): 17-34.
[3]	ZHU J , JIANG W , LIU A ,et al. Time-dependent popular routes based trajectory outlier detection[C]// WISE. 2015: 16-30.
[4]	CHAWLA S , ZHENG Y , HU J . Inferring the root cause in road traffic anomalies[C]// IEEE International Conference on Data Mining. 2012: 141-150.
[5]	BIRANT D , KUT A . ST-DBSCAN:an algorithm for clustering spatial-temporal data[J]. Data＆Knowledge Engineering, 2007,60(1): 208-221.
[6]	TOOHEY K , DUCKHAM M . Trajectory similarity measures[J]. Sigspatial Special, 2015,7(1): 43-50.
[7]	魏龙翔, 何小海, 滕奇志 ,等. 结合Hausdorff距离和最长公共子序列的轨迹分类[J]. 电子与信息学报, 2013,35(4): 784-790.
	WEI L X , HE X H , TENG Q Z ,et al. Trajectory classification based on Hausdorff distance and longest common subsequence[J]. Journal of Electronics and Information Technology, 2013,35(4): 784-790.
[8]	YUAN G , SUN P , ZHAO J ,et al. A review of moving object trajectory clustering algorithms[J]. Artificial Intelligence Review, 2016: 1-22.
[9]	CHEN L , ?ZSU M , ORIA V . Robust and fast similarity search for moving object trajectories[C]// The 2005ACMSIGMOD International Conference on Management of Data. 2005: 491-502
[10]	CHEN L , NG R T . On the marriage of Lp-norms and edit distance[C]// VLDB. 2004: 792-803.
[11]	YUAN G , XIA S , ZHANG L ,et al. An efficient trajectory-clustering algorithm based on an index tree[J]. Transactions of the Institute of Measurement＆Control, 2012,34(7): 850-861.
[12]	NANNI M , PEDRESCHI D . Time-focused clustering of trajectories of moving objects[J]. Journal of Intelligent Information Systems, 2006,27(3): 267-289.
[13]	RODRIGUEZ A , LAIO A . Machine learning clustering by fast search and find of density peaks[J]. Science, 2014,344(6191): 1492-1496.
[14]	SHANG F , JIAO L C , SHI J ,et al. Fast density-weighted low-rank approximation spectral clustering[J]. Data Mining ＆ Knowledge Discovery, 2011,23(2): 345-378.
[15]	WANG J , WU J , WANG J ,et al. Multi-criteria decision-making methods based on the Hausdorff distance of hesitant fuzzy linguistic numbers[J]. Soft Computing, 2016,20(4): 1621-1633.
[16]	WEI L X , HE X H , TENG Q Z ,et al. Trajectory classification based on Hausdorff distance and longest common subsequence[J]. J Electron Inf Technol, 2013,35(4): 784-790
[17]	KHOSHAEIN V . Trajectory clustering using a variation of Fréchet distance[D]. Ottawa,Canada:University of Ottawa, 2014.
[18]	JEUNG H , YIUM L , JENSEN C S . Trajectory pattern mining[M]// Computing with Spatial Trajectories. Springer, 2011: 143-177.
[19]	郭虎升, 王文剑 . 动态粒度支持向量回归机[J]. 软件学报, 2013(11): 2535-2547.
	GUO H S , WANG W J . Dynamical granular support vector regression machine[J]. Journal of Software, 2013(11): 2535-2547.

采样点数	ENDTW		DTW
采样点数	变异系数	均值	变异系数	均值
100%	—	4.07×10^?6	—	1.63×10^?4
80%	0.128	5.20×10^?6	0.111	1.98×10^?4
65%	0.177	5.66×10^?6	0.189	2.65×10^?4
50%	0.162	5.58×10^?6	0.214	3.29×10^?4
40%	0.212	6.68×10^?6	0.186	4.25×10^?4
	均值变异系数	0.173		0.380

基于改进密度聚类与模式信息挖掘的异常轨迹识别方法

Abnormal trajectory detection method based on enhanced density clustering and abnormal information mining

在线阅读

PDF下载

可视化

摘要/Abstract

引用本文

使用本文

图/表 13

参考文献 19

相关文章 1

Metrics

推荐阅读 0