By Chris McCahill
Researchers recently developed ways of combining the vast information available from Google Street View with custom AI programs to generate highly accurate inventories of street signs. The effort focused specifically on stop and yield signs, but these kinds of algorithms could be used to identify other types of infrastructure and even replace or augment time-consuming street audits.
For this particular study, researchers at RMIT in Melbourne developed fully automated systems based on open source tools that detect signs with 96 percent accuracy and identify their precise location. Manually collected GPS data, in comparison, can be up to 10 meters off target due to “human error.” Computer-aided street audits using Google Street View—still much faster than field audits—can take trained auditors up to 10 minutes per location. This newest study also relied on information from Google’s API, but the researchers note that it could incorporate other data sources, such as fleet vehicle cameras, and it can be easily scaled.
”Ours is one of several early applications for this to meet a specific industry need,” according to the authors, “but a whole lot more will emerge in coming years.”
Chris McCahill is the Deputy Director at SSTI.