M.Sc Student | Massad Ido |
---|---|
Subject | Applying Kalman Filter in Modeling Digital Terrain Model from Crowdsourced-based GPS Observations |
Department | Department of Civil and Environmental Engineering | Supervisor | Dr. Sagi Dalyot |
Full Thesis text | ![]() |
Today, there
are a wide variety of online resources that provide public mapping
infrastructures and geographic information. Most web-based map services, such
as Google Maps, Yahoo! Maps and Bing Maps, are mostly based on data that is
collected by authoritative mapping agencies. Alternatively, some relatively new
web-map services, such as OpenStreetMap (OSM) and Wikimapia, are based mostly
on volunteered data collected by the public (e.g., crowdsourced mapping).
Though such volunteered-based map services platforms show an increasing planar
(2D) accuracy, completeness and update-rate of their mapping infrastructure,
surprisingly, there is a lack of comparable data and accuracy measures with
respect to the third dimension, i.e., elevation; more specifically, the
topographic representation that is based on the volunteered collected data.
Most volunteered-based web-services still rely on existing open-source
authoritative topographic infrastructures, and not on data collected by the
volunteers. Moreover, topographic information that is open to the public and is
free to use, e.g., ASTER and SRTM, is regularly available with relatively low
elevation accuracy (not better than 10m) and low planar resolution (over 30m). Volunteered
data, on the other hand, collected by individuals that are situated “all over”
the globe, can offer with new capabilities and data-characteristics having
potentially higher qualities. Since Volunteered Geographic Information (VGI)
can be a powerful tool for creating and sharing geographical data that is
accurate, reliable and current, that in some cases can even surpass a
commercial equivalent, a question arises as to why not use the third dimension
also.
One of the key elements that enables VGI was the introduction of mobile
“surveying” devices, such as smartphones and tablets, equipped with GNSS chips.
Current devices can reach a position accuracy of below 8m with the use
Assisted-GPS technologies; furthermore, recent tests show even greater position
accuracy that is below 5m in devices as the iPhone 5, Nokia Lumia, and Asus
Android tablet. In fact, some of the newer devices also combine GPS with other
similar services, such as GLONASS and soon Galileo, which will improve the
position accuracy even further.
This work proposes to examine the feasibility of using crowdsourced VGI working paradigm for the task of producing a reliable digital topographic representation (the bare earth) for general use. This is achieved by collecting GPS observations that are available from VG data sources, while applying a tailored 2D Kalman-filter based algorithm, aimed at reducing noise and ambiguities that exist in mobile GPS observations. This work presents this methodology, with analysis of results achieved by this implementation, showing the feasibility of this working methodology, having good results and accuracy of the generated topography.