Abstract:
Geocrowdsourced data (GcD), also known as volunteered geographic
information, has proven to be an invaluable resource to the geospatial science
community. From a United States National Security perspective, GcD has become a
force-multiplier for the Department of Defense aiding in nuclear counterproliferation
efforts; at a humanitarian level it was used to aid first responders reaching earthquake
victims in Haiti. Despite the potential benefits, research has shown GcD to be unreliable
unless moderated with quality assessment methods applied to the data. However,
circumstances may prevent moderation and new quality assessment methods must be
designed. This research demonstrates a correlation between the number of GcD
contributors and the level of positional accuracy of information contributed to the George
Mason University Geocrowdsourcing Testbed (GMU-GcT). A mobile-phone, imagebased
data contribution tool from the GMU-GcT was developed and distributed to
student volunteers at GMU who provided information regarding pre-defined locations on
campus. Findings showed that the positional accuracy characteristics of the data
contributions to the GMU-GcT improved with added contributors, reaching a level
comparable to previously-studied accuracy threshholds reached with a significantly more
detailed and heavily moderated data contribution workflow. Undermoderated reports
from single contributors averaged 8.55m in positional error. With an increasing number
of contributors, positional error of reports for the same item dropped to 3.89m (n=20).
The most common positional error threshold for geocrowdsourced data, referred to in
previous work as the Haklay distance (approximately 6.0 meters) was reached with two
contributors, and after four contributors, the positional error rate stayed fairly constant.
This research demonstrated that a fully moderated crowdsourced data contribution
process, used in previous incarnations of the GMU-GcT, is unnecessary for producing
data with adequate fitness-for-use, including common routing and obstacle avoidance
algorithms.