請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/6011
標題: | 以軌跡資料剖析與倉儲進行物體移動行為分析 Trajectory Data Profiling and Warehousing for Behavior Analysis of Moving Objects |
作者: | Huey-Ru Wu 吳蕙如 |
指導教授: | 陳明憲 |
關鍵字: | 物體軌跡,移動模式,軌跡分割,行為剖析,軌跡倉儲, object trajectory,moving pattern,trajectory dividing,behavior profiling,trajectory warehousing, |
出版年 : | 2013 |
學位: | 博士 |
摘要: | Nowadays, devices attached with position detecting techniques are used on many places to track moving of objects. The collected time and position records, which constructed moving trajectories of objects, are in huge amount. Among the object trajectories, interesting moving behaviors are hidden and worth to be revealed through some processing. In this dissertation, we focus on analyzing object moving behaviors through trajectory data profiling and warehousing.
In an area where a set of objects moving around, there are some typical moving behaviors of objects at different regions in respect to the geographical nature or other spatiotemporal conditions. Not only paths that objects moving along, we also want to know how different groups of objects move with various speeds. Therefore, given a set of collected trajectories spreading in a bounded area, we are interested in discovering typical moving styles in different regions of all monitored moving objects. These regional typical moving styles are regarded as profile of the monitored moving objects, which may help reflect the geographical information of observed area and the moving behaviors of observed moving objects. However, an object can move with various speeds and arbitrarily changing directions. The changes cause difficulty in analyzing behaviors among object trajectories. Thus, we present DivCluST, an approach to finding regional typical moving styles by dividing and clustering trajectories in consideration of both spatial and temporal constraints. Different from existing works that considered only spatial properties or just some interesting regions of trajectories, DivCluST focuses more on finding typical regional spatiotemporal behaviors over a large area. It takes both spatial and temporal information into account when designing the criteria for trajectory dividing and the distance measurement for adaptive k-means clustering. Extensive experiments on three types of real data sets with specially designed visualization are presented to show the effectiveness of DivCluST. With huge amount of object moving trajectories collected continuously and boundlessly, we need a well designed data structure to analyze trajectory data and keep moving behavior information for further processes and applications. A trajectory data warehouse is an effective way to store organized moving patterns extracted from object trajectories, and can offer efficient information queries for event analysis and decision making. Different from existing works that stored only statistic values of trajectories or focused only on limited number of selected regions, we present a trajectory data warehouse storing moving patterns spreading in all areas. We design a proper data format for moving patterns to represent typical behaviors, containing main properties of trajectories, such as laying positions, moving directions and forward speeds. Also, we propose a corresponding table schema for keeping long-term moving patterns in our trajectory warehouse. A two-stage algorithm is proposed to online process the incoming trajectory data over a large area and extract the moving patterns from them based on multidimensional unit grids. Operations on moving patterns and related tables, such as spatial position relation and aggregation, based on multiple granularity of grids, are provided for flexible query requirements and warehouse maintenance. Experiments on real-world trajectory data sets show that our designs on storage and operations of trajectory patterns make our trajectory data warehousing effective and efficient for moving pattern analysis. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/6011 |
全文授權: | 同意授權(全球公開) |
顯示於系所單位: | 電機工程學系 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-102-1.pdf | 2.21 MB | Adobe PDF | 檢視/開啟 |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。