摘要
针对轨迹数据概化大多只考虑空间位置因素,而忽略了其他轨迹特性,从而造成概化前后信息丢失过大的问题,提出轨迹数据的多因素概化方法.该方法综合考虑了空间位置、时间、速度、加速度和方向角等因素,首先,通过在轨迹点聚类的距离约束中加入多因素对DBSCAN算法进行改进.然后,用类簇的代表点进行区域划分,连接轨迹经过的区域生成概化轨迹.由于多因素下的轨迹概化缺少统一的衡量标准,从信息丢失和概化前后效果比较2个方面,提出多因素下轨迹概化效果的度量方法.飓风数据实验表明,所提方法生成的概化轨迹信息丢失明显减少,聚类结果更加接近原始轨迹.
In order to solve the problem that most of current trajectory data generalization methods only consider the spatial characteristic and ignore other trajectory features, thus resulting in informa- tion loss, a method of multiple-factor generalization of trajectory data （TRMGEN） is proposed, which considers trajectory features, such as spatial location, time, velocity, direction angle and ac- celeration, etc. First, the DBSCAN algorithm is improved by adding factors into distance constraint when clustering the points. Then, the representative points of clusters are used in the region divi- sion, and the regions are connected to generate the generalized trajectory. As the trajectory multi- factor generalization lacks unified measure standard, from the aspects of information loss and the similarity with original trajectory, a method for measuring generalization effect is presented. Experi- mental results on the Hurricane dataset show that the TRMGEN can greatly reduce the information loss of the generalized trajectory, and the cluster result of the generalized trajectory is closer to the original trajectory.
出处
《东南大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2012年第A02期271-275,共5页
Journal of Southeast University：Natural Science Edition
基金
国家教育部博士点基金资助项目(20110095110010)
关键词
轨迹概化
多因素
信息丢失量
轨迹聚类
trajectory generalization
multiple-factor
information loss
trajectory clustering
