Header

UZH-Logo

Maintenance Infos

Exploring movement using similarity analysis


Dodge, S. Exploring movement using similarity analysis. 2011, University of Zurich, Faculty of Science.

Abstract

Movement is a vital aspect of almost all organisms and many spatio-temporal processes. Hence it is crucial to understand movement and gain knowledge about its patterns. Recent advances in positioning technologies provide an increasing access to massive repositories of movement data and hence challenges arise to develop new exploratory tools and knowledge discovery techniques in order to extract meaningful information, discover interesting patterns, and explore the dynamic behavior of moving objects (humans, vehicles, vessels, animals) or processes (hurricanes, oil spills). Among knowledge discovery techniques, the exploration of similarities in the movement of multiple objects is a key emerging interest. Learning about movement similarities can be beneficial in the prediction, modeling and simulation of collective behavior of dynamic phenomena.
This thesis intends to contribute to GIScience’s exploratory capacity to discover insights about patterns of movement as well as existing similarities between movement behaviors of different objects. Specifically, the aim is to develop concepts and methods for incorporating movement parameters such as speed, acceleration, or direction in the study and analysis of movement. Hence in this thesis, movement similarity is defined as the resemblance in the variations of movement parameters of objects over time.
This study, with a perspective on movement, undertakes a three-stage research process including the development of (a) a conceptual framework, (b) feature extraction and segmentation methods, and (c) similarity assessment techniques. The overall study involves an iterative research process integrating quantitative techniques from GIScience and knowledge discovery approaches in order to extract high-level information from low-level, raw movement data.
The core of the research process is presented in four scientific papers. Research Paper 1 proposes a conceptual framework for movement as well as a comprehensive classification of movement patterns. Research Paper 2 presents a segmentation technique in order to extract movement features from trajectories of moving objects. The segmentation process can be seen as a dimension reduction technique to simplify the structure of the movement data in order to facilitate knowledge discovery. Research Paper 3 proposes a novel similarity assessment approach relying on the segmentation technique. Finally, Research Paper 4 extends the dimensionality of the main approach towards the detection of relative movement patterns.
Furthermore, through a set of experiments this thesis shows that the proposed methods can be successfully applied in conjunction with data mining techniques in order to support knowledge discovery from various movement datasets in realworld applications (e.g. transportation, meteorology). Consequently, the outcomes of this thesis can contribute to knowledge discovery from movement data where the interest is to extract or group similar behaviors of dynamic objects.

Abstract

Movement is a vital aspect of almost all organisms and many spatio-temporal processes. Hence it is crucial to understand movement and gain knowledge about its patterns. Recent advances in positioning technologies provide an increasing access to massive repositories of movement data and hence challenges arise to develop new exploratory tools and knowledge discovery techniques in order to extract meaningful information, discover interesting patterns, and explore the dynamic behavior of moving objects (humans, vehicles, vessels, animals) or processes (hurricanes, oil spills). Among knowledge discovery techniques, the exploration of similarities in the movement of multiple objects is a key emerging interest. Learning about movement similarities can be beneficial in the prediction, modeling and simulation of collective behavior of dynamic phenomena.
This thesis intends to contribute to GIScience’s exploratory capacity to discover insights about patterns of movement as well as existing similarities between movement behaviors of different objects. Specifically, the aim is to develop concepts and methods for incorporating movement parameters such as speed, acceleration, or direction in the study and analysis of movement. Hence in this thesis, movement similarity is defined as the resemblance in the variations of movement parameters of objects over time.
This study, with a perspective on movement, undertakes a three-stage research process including the development of (a) a conceptual framework, (b) feature extraction and segmentation methods, and (c) similarity assessment techniques. The overall study involves an iterative research process integrating quantitative techniques from GIScience and knowledge discovery approaches in order to extract high-level information from low-level, raw movement data.
The core of the research process is presented in four scientific papers. Research Paper 1 proposes a conceptual framework for movement as well as a comprehensive classification of movement patterns. Research Paper 2 presents a segmentation technique in order to extract movement features from trajectories of moving objects. The segmentation process can be seen as a dimension reduction technique to simplify the structure of the movement data in order to facilitate knowledge discovery. Research Paper 3 proposes a novel similarity assessment approach relying on the segmentation technique. Finally, Research Paper 4 extends the dimensionality of the main approach towards the detection of relative movement patterns.
Furthermore, through a set of experiments this thesis shows that the proposed methods can be successfully applied in conjunction with data mining techniques in order to support knowledge discovery from various movement datasets in realworld applications (e.g. transportation, meteorology). Consequently, the outcomes of this thesis can contribute to knowledge discovery from movement data where the interest is to extract or group similar behaviors of dynamic objects.

Statistics

Downloads

3 downloads since deposited on 19 Mar 2012
1 download since 12 months
Detailed statistics

Additional indexing

Item Type:Dissertation
Referees:Weibel R, Fabrikant S I
Communities & Collections:07 Faculty of Science > Institute of Geography
Dewey Decimal Classification:910 Geography & travel
Language:English
Date:2011
Deposited On:19 Mar 2012 15:24
Last Modified:21 Nov 2017 15:57
Number of Pages:160
Free access at:Official URL. An embargo period may apply.
Official URL:http://opac.nebis.ch/ediss/20111260.pdf
Related URLs:http://opac.nebis.ch/F/?local_base=NEBIS&CON_LNG=GER&func=find-b&find_code=SYS&request=006785185

Download

Content: Published Version
Filetype: PDF - Registered users only
Size: 14MB