438
These datasets can roughly be divided into three categories: Raster datasets are
the representation of the world as a surface divided into a regular grid of cells
(e.g., pixels). Raster models are useful for storing data that varies continuously, as
in an aerial photograph, a satellite image, a surface of chemical concentrations, or
an elevation surface. For example, remote sensing imagery collected from UAVs
can be considered as raster data, and the pixel values may represent elevation above
sea level, chemical concentrations, or rainfall, etc. Vector datasets are the representation of the world using points, lines, and polygons. Vector models are useful for
storing data that has discrete boundaries (discrete data), such as country borders,
land parcels, and streets. For example, in land cover allocation problem, vector data
such as polygons may be used to identify the different types of lands. Graph datasets, which are often considered together with vector data, are graphs with vectors
and edges associated with spatial properties. A road network or electric grid in a city
can be considered as such type of data.
Due to the different nature of these datasets, the models, as well as data structures that are used for these, are different. For example, raster datasets where each
pixel (a square grid cell) represents information associated with a specific location
can be treated as a multidimensional image. The data structures for these datasets
are thus matrices where (x, y) coordinates represent a square area and the additional
dimensions represent the sensor readings. Thanks to the autocorrelation effect causing similar values at nearby locations, efficient image compression, and indexing
algorithms are used to handle these.
On the other hand, vector datasets are mostly mathematical equations that define
the geometric shapes which are relatively smaller datasets. However, depending on
the resolution, the details of each polygon may be represented in a more complicated manner, causing the size and rendering of these to increase by finer resolution.
As a simple example, a pentagon in finer resolution may be represented as a rectangle in a coarser resolution. To handle the rendering issues for different resolutions
and provide a data generalization, algorithms for vector simplification (e.g.,
Douglas–Peucker algorithm) were developed. However, it is worth noting that the
algorithms for vector data analytics are complex due to the inherent topological
relationships between features.
Graphs are studied for a long time due to their use in a variety of application
domains such as electronic circuit design, social network analysis, spatial analytics,
fluid dynamics, etc. Spatial network graphs have a variety of differences compared
to other graph datasets due to the varying planarity (e.g., bridges on roads, etc.),
temporal dimension, multi-constrained node structures (e.g., lights and stop signs
on roads) as well as their sparsity. Many data structures that were developed for
graph datasets are already used for spatial network graphs, but more specialized
tools and algorithms are also present such as Lagrangian shortest path framework
(Gunturi et al. 2011) for temporally aware routing (Evans et al. 2014), and shortest
path/travel time approaches using (Dijkstra 1959), and A∗ algorithms.
E. Eftelioglu and S. Shekhar
These datasets can roughly be divided into three categories: Raster datasets are
the representation of the world as a surface divided into a regular grid of cells
(e.g., pixels). Raster models are useful for storing data that varies continuously, as
in an aerial photograph, a satellite image, a surface of chemical concentrations, or
an elevation surface. For example, remote sensing imagery collected from UAVs
can be considered as raster data, and the pixel values may represent elevation above
sea level, chemical concentrations, or rainfall, etc. Vector datasets are the representation of the world using points, lines, and polygons. Vector models are useful for
storing data that has discrete boundaries (discrete data), such as country borders,
land parcels, and streets. For example, in land cover allocation problem, vector data
such as polygons may be used to identify the different types of lands. Graph datasets, which are often considered together with vector data, are graphs with vectors
and edges associated with spatial properties. A road network or electric grid in a city
can be considered as such type of data.
Due to the different nature of these datasets, the models, as well as data structures that are used for these, are different. For example, raster datasets where each
pixel (a square grid cell) represents information associated with a specific location
can be treated as a multidimensional image. The data structures for these datasets
are thus matrices where (x, y) coordinates represent a square area and the additional
dimensions represent the sensor readings. Thanks to the autocorrelation effect causing similar values at nearby locations, efficient image compression, and indexing
algorithms are used to handle these.
On the other hand, vector datasets are mostly mathematical equations that define
the geometric shapes which are relatively smaller datasets. However, depending on
the resolution, the details of each polygon may be represented in a more complicated manner, causing the size and rendering of these to increase by finer resolution.
As a simple example, a pentagon in finer resolution may be represented as a rectangle in a coarser resolution. To handle the rendering issues for different resolutions
and provide a data generalization, algorithms for vector simplification (e.g.,
Douglas–Peucker algorithm) were developed. However, it is worth noting that the
algorithms for vector data analytics are complex due to the inherent topological
relationships between features.
Graphs are studied for a long time due to their use in a variety of application
domains such as electronic circuit design, social network analysis, spatial analytics,
fluid dynamics, etc. Spatial network graphs have a variety of differences compared
to other graph datasets due to the varying planarity (e.g., bridges on roads, etc.),
temporal dimension, multi-constrained node structures (e.g., lights and stop signs
on roads) as well as their sparsity. Many data structures that were developed for
graph datasets are already used for spatial network graphs, but more specialized
tools and algorithms are also present such as Lagrangian shortest path framework
(Gunturi et al. 2011) for temporally aware routing (Evans et al. 2014), and shortest
path/travel time approaches using (Dijkstra 1959), and A∗ algorithms.
E. Eftelioglu and S. Shekhar