Bonerath, Annika: Geometric Algorithms for the Visual Exploration of Spatiotemporal Data. - Bonn, 2025. - Dissertation, Rheinische Friedrich-Wilhelms-Universität Bonn.
Online-Ausgabe in bonndoc: https://nbn-resolving.org/urn:nbn:de:hbz:5-81248
Online-Ausgabe in bonndoc: https://nbn-resolving.org/urn:nbn:de:hbz:5-81248
@phdthesis{handle:20.500.11811/12847,
urn: https://nbn-resolving.org/urn:nbn:de:hbz:5-81248,
doi: https://doi.org/10.48565/bonndoc-515,
author = {{Annika Bonerath}},
title = {Geometric Algorithms for the Visual Exploration of Spatiotemporal Data},
school = {Rheinische Friedrich-Wilhelms-Universität Bonn},
year = 2025,
month = feb,
note = {The visualization of spatial data is an important research field in geoinformation science. Especially nowadays, where positioning sensors are widely spread, many (large) data sets have spatial information. As an example, take a set of bird observations where a data point corresponds to the location of a bird sighting and possibly additional information (timestamp, photo, species specification, etc.). Often, such data sets are large and complex. Visualizations from cartography allow users to explore and analyze the data. This ranges from visualizations that give an overview of the data, to tools that enable a detailed analysis of data patterns. In this thesis, we develop methods that support such visualizations. To be more precise, we apply methods from theoretical computer science to existing visualizations from cartography to improve them, e.g., to decrease the time needed to produce a visualization.
This thesis has two parts. In the first part, we consider the spatiotemporal case where each data point is an event consisting of a point in space and time. For visualizing spatiotemporal data, it is common to use an interactive visualization. Here, we focus on filtering the data for time windows. Then, the data that temporally lies in the time window is visualized on a map. As visualizations, we consider three standard techniques for visualizing point sets.
• A standard visualization is the representation of the events with one or multiple polygons, i.e., we aggregate the points that lie in the time window into one or multiple polygons. In particular, we build on the existing representation technique α-shapes. It is parameterized by a value α. Depending on α the representation ranges from the convex hull of the point set, over multiple, detailed polygons, to no polygons at all. Typically, one chooses an α value that still reflects the point set distribution while at the same time simplifying it. We also discuss a modification of the standard α-shapes which produces schematized α-shapes.
• Another visualization of a point set is the spatial density map. Here, we overlay the map with a grid, and all grid cells that contain more data points than a given threshold are colored. Such a visualization can also color encode the number of points in the grid cells. As output, we receive a simplified and schematized aggregation of the points.
• We also look at the labeling of point sets. Labeling is a standard cartography technique to display additional data information. Therefore, we place an icon (e.g., a symbol) on the map over the data point. To achieve good legibility, only a selection of labels is displayed. To guarantee a pleasant interaction for the user, the visualization must be displayed in real-time. Especially for large data sets this is a challenge. In this work, we develop data structures that guarantee a fast response time. Such data structures are called time-windowed data structures. As a general idea, we break down the visualizations into their atomic geometric elements. Then, we pre-process the set of all time-window queries for which each atomic geometric element is displayed. Furthermore, we also look at consistency criteria between two successive time-window queries.
Especially, when the interaction is implemented in the user interface with a slider (time-window slider), an interaction with the time-window slider should not lead to flickering effects in the visualization.
In the second part of the thesis, we consider a static case where we have a complex spatial geometry without any temporal information as input. While displaying this complex geometry with all its details can be needed for a thorough analysis, it can be unclear and overwhelming for a user who wants a high-level overview. For example, think of a very detailed border of a country. Often, such a border is used as an underlying base map to give the user a spatial orientation. To not distract the reader from the data that lies over the base map, the base map should not be too detailed. Hence, simplifying the border can be necessary for clear and readable visualizations. We simplify polygons by hulls i.e., a polygon that contains the input polygon.},
url = {https://hdl.handle.net/20.500.11811/12847}
}
urn: https://nbn-resolving.org/urn:nbn:de:hbz:5-81248,
doi: https://doi.org/10.48565/bonndoc-515,
author = {{Annika Bonerath}},
title = {Geometric Algorithms for the Visual Exploration of Spatiotemporal Data},
school = {Rheinische Friedrich-Wilhelms-Universität Bonn},
year = 2025,
month = feb,
note = {The visualization of spatial data is an important research field in geoinformation science. Especially nowadays, where positioning sensors are widely spread, many (large) data sets have spatial information. As an example, take a set of bird observations where a data point corresponds to the location of a bird sighting and possibly additional information (timestamp, photo, species specification, etc.). Often, such data sets are large and complex. Visualizations from cartography allow users to explore and analyze the data. This ranges from visualizations that give an overview of the data, to tools that enable a detailed analysis of data patterns. In this thesis, we develop methods that support such visualizations. To be more precise, we apply methods from theoretical computer science to existing visualizations from cartography to improve them, e.g., to decrease the time needed to produce a visualization.
This thesis has two parts. In the first part, we consider the spatiotemporal case where each data point is an event consisting of a point in space and time. For visualizing spatiotemporal data, it is common to use an interactive visualization. Here, we focus on filtering the data for time windows. Then, the data that temporally lies in the time window is visualized on a map. As visualizations, we consider three standard techniques for visualizing point sets.
• A standard visualization is the representation of the events with one or multiple polygons, i.e., we aggregate the points that lie in the time window into one or multiple polygons. In particular, we build on the existing representation technique α-shapes. It is parameterized by a value α. Depending on α the representation ranges from the convex hull of the point set, over multiple, detailed polygons, to no polygons at all. Typically, one chooses an α value that still reflects the point set distribution while at the same time simplifying it. We also discuss a modification of the standard α-shapes which produces schematized α-shapes.
• Another visualization of a point set is the spatial density map. Here, we overlay the map with a grid, and all grid cells that contain more data points than a given threshold are colored. Such a visualization can also color encode the number of points in the grid cells. As output, we receive a simplified and schematized aggregation of the points.
• We also look at the labeling of point sets. Labeling is a standard cartography technique to display additional data information. Therefore, we place an icon (e.g., a symbol) on the map over the data point. To achieve good legibility, only a selection of labels is displayed. To guarantee a pleasant interaction for the user, the visualization must be displayed in real-time. Especially for large data sets this is a challenge. In this work, we develop data structures that guarantee a fast response time. Such data structures are called time-windowed data structures. As a general idea, we break down the visualizations into their atomic geometric elements. Then, we pre-process the set of all time-window queries for which each atomic geometric element is displayed. Furthermore, we also look at consistency criteria between two successive time-window queries.
Especially, when the interaction is implemented in the user interface with a slider (time-window slider), an interaction with the time-window slider should not lead to flickering effects in the visualization.
In the second part of the thesis, we consider a static case where we have a complex spatial geometry without any temporal information as input. While displaying this complex geometry with all its details can be needed for a thorough analysis, it can be unclear and overwhelming for a user who wants a high-level overview. For example, think of a very detailed border of a country. Often, such a border is used as an underlying base map to give the user a spatial orientation. To not distract the reader from the data that lies over the base map, the base map should not be too detailed. Hence, simplifying the border can be necessary for clear and readable visualizations. We simplify polygons by hulls i.e., a polygon that contains the input polygon.},
url = {https://hdl.handle.net/20.500.11811/12847}
}